{"locked":[{"data_source":"data/prediction_ledger.db (prediction_features table)","expected_direction":"negative","feature_set":["predicted_magnitude","predicted_magnitude_high90","region_ceiling_M","magnitude_min"],"git_commit_at_registration":"db7fb91420b171a6a884bc0d97ffe8da1f8ea20e","hypothesis_id":"20260504_predicted_magnitude_has_negative_247ccebf","locked_at":"2026-05-04T19:50:40Z","locked_file_sha256":"b300164c4e2a7e06c266212a4b619f702e148c773376ec10288efbd7ab7ccc0f","notes":"Reverse-causal scan on 2026-05-04 (N=1002 resolved EQ preds) returned predicted_magnitude L1 coef of -1.45, predicted_magnitude_high90 coef of -1.43, with positive coefs on region_ceiling_M (+2.60) and magnitude_min (+1.49). Pattern: predictions claiming larger quakes match LESS often, predictions hugging the magnitude floor match MORE. This is a CALIBRATION finding -- the magnitude predictor is over-claiming on high-magnitude events. Hypothesis: same negative coef on predicted_magnitude (<-0.5) will reproduce when we retrain the reverse-causal classifier on data restricted to 2026-05-05 through 2026-08-04. If confirmed: justifies retraining the magnitude predictor with calibration loss + held-out validation. If disconfirmed: 2026-05-04 finding was a fluke of the snapshot.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"earthquake"},"registered_at":"2026-05-04T19:50:13Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"l1_logistic_on_metadata_features","n_permutations":null},"threshold":{"direction":"negative","metric":"predicted_magnitude_l1_coefficient","value":-0.5},"title":"Predicted magnitude has negative L1 coefficient against match outcome on EQ predictions over next 90d (replicates 2026-05-04 reverse-causal finding)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-08-04"},{"data_source":"data/private/replenishable_atlas/box_elder_with_qfaults_v1.csv","expected_direction":"positive","feature_set":["fault_loaded_subscore","depth_sweet_spot_score","closed_loop_fault_drilling_score","target_depth_m","hot_spring_on_fault"],"git_commit_at_registration":"4752e6d47d843bf6f78031d7149a75dad95a92a8","hypothesis_id":"20260506_fault_loaded_depth_sweet_spot_sc_1be4724e","locked_at":"2026-05-06T18:09:44Z","locked_file_sha256":"da59f7485628b060c1b090fc16b1e22c3220077f1608bf62b68b1c965060a052","notes":"FORWARD test of the 5-axis screening refinement (4 added: fault-loaded subscore from USGS Qfaults proximity + Holocene slip + hot-spring-on-fault binary; 5 added: depth sweet-spot 500-1500m). Pass criterion: at Phase 4 single-well validation completion (5 wells, one per top-5 ranked Box Elder zone), the top-3 zones by combined_new_score (Promontory Pt north, Hansel Valley, Howell Valley step-over) achieve >=60 pct 4-of-4 pass rate (drawdown <=10pct, recovery >=90pct in 72h, surrounding-well drawdown <0.5ft, tritium <0.5 TU). Bottom-2 zones (Brigham City + East Cache Crystal Hot Springs) provide discriminating-power confirmation. Null baseline: random ranking expects 60pct pass rate uniformly. Forward window: 6-24 months from Phase 4 drilling start (anticipated 2026 Q4 to 2027 Q3). Withdraw rule: if Phase 4 doesn't drill 4+ wells by 2027-12-31, declare underpowered. Notes: Operationalizes the engineering criterion 'fault-loaded with sustained pressure-driven flow at sweet-spot drilling depth (shallow enough to be cheap, deep enough not to interrupt regional hydrology)' raised in 2026-05-06 client conversation. Adds 4th + 5th screening axes on top of v3 atlas 3-axis methodology (tidal compliance + Vp/Vs + gravity). IP-relevant: candidate continuation-in-part on P069 multi-stage extraction; 5-axis screening combined with multi-stage cascade is novel.","ots_proof_path":null,"outcome":{"lookahead_hours":17520.0,"min_magnitude":null,"type":"earthquake"},"registered_at":"2026-05-06T18:09:37Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"phase4_pass_rate","n_permutations":null},"threshold":{"direction":"positive","metric":"pass_rate","value":0.6},"title":"Fault-loaded + depth sweet-spot screening (5-axis) predicts Phase 4 sustained-flow pass rate at Box Elder pilot","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-12-31"},{"data_source":"dense-array Millard sensor master + DeepMap satellite_hyperspectral + USGS QFaults + DOGM wells","expected_direction":"positive","feature_set":["iocg_axis1_pyrite_gold","iocg_axis2_heavy_density","iocg_axis3_iron_oxide","iocg_axis4_basement_prox","iocg_axis5_qfault_feeder","iocg_composite_score"],"git_commit_at_registration":"2184dfaabb13d47f66cf79b9f8bc895faac1f7fd","hypothesis_id":"20260512_iocg_scout_classifier_elite_stro_d3a5b682","locked_at":"2026-05-12T00:21:45Z","locked_file_sha256":"74d15947cc2cbfc8309df3d6b7dc0fcf6d7b1529de6277aada32a90eea77883a","notes":"IOCG classifier shipped 2026-05-12 with dense-array-calibrated weights against 8 dense-array IOCG-positive sensors (Antelope Field 0073 conf=0.623, Black Rock 0053 conf=0.848, Millard County 0046 conf=0.712, Millard County 0044 conf=0.464, Black Rock 0094 conf=0.505, Black Rock 0058 conf=0.431, Baker Hot Springs 0003 conf=0.477). FORWARD CLAIM: at any new commercial dense-array deployment or drilled hole within Western US Cordilleran/Basin-and-Range provinces, of the locations where DeepMap IOCG composite_score >= 50 (STRONG/ELITE), >=50pct will have dense-array mineral fingerprint Pyrite + Chalcopyrite + Native Gold all >5/10 OR drilling will confirm Cu/Au mineralization within 365 days of public claim. Forward window: 2026-05-12 to 2027-05-12. Null baseline: untargeted Western US drilling has ~5pct IOCG-positive rate per USGS Mineral Resources data. Withdraw rule: if fewer than 4 candidate sites are surveyed/drilled by 2027-05-12 the test is underpowered and the hypothesis is withdrawn rather than failed.","ots_proof_path":null,"outcome":{"lookahead_hours":8760.0,"min_magnitude":null,"type":"drill_confirm"},"registered_at":"2026-05-12T00:21:36Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"binomial_2sided","n_permutations":null},"threshold":{"direction":"positive","metric":"true_positive_rate","value":0.5},"title":"IOCG Scout classifier ELITE/STRONG sites confirm via dense-array/drill within 365 days","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-05-12"},{"data_source":"dense-array Schumann step2 z-scores (Delta-OC array baseline 2024-06-20 z_mode1=5.04 z_mode3=4.32) + DeepMap eigenmode_schumann_coupling tensor","expected_direction":"positive","feature_set":["laic_mode1_z","laic_mode3_z","laic_combined_z","laic_anomaly_class","laic_eq_precursor_prob"],"git_commit_at_registration":"2184dfaabb13d47f66cf79b9f8bc895faac1f7fd","hypothesis_id":"20260512_schumann_laic_eq_precursor_class_87404330","locked_at":"2026-05-12T00:46:10Z","locked_file_sha256":"5f0e5a5c5c9129c0d94b6d748999d9d5b6ffbee1ed61f54a6db3c7fe75b73889","notes":"Forward claim: across the 5 dense-array Schumann-capable arrays in Millard County (Delta-Oak City, Black Rock, Antelope Field, Millard County, Baker Hot Springs), sites with combined_z >= 4.0 (ALERT or CRITICAL class) will have an M3+ EQ within 30 days at least 30pct of the time over the 12-month forward window 2026-05-12 to 2027-05-12. Null baseline: regional M3+ rate is ~5pct per 30d in Millard County per USGS catalog. WITHDRAW RULE: if fewer than 20 ALERT/CRITICAL classifications are issued over the year, the test is underpowered and the hypothesis withdraws. Methodology refs: Sentman 1989, Pulinets-Boyarchuk 2004, Ouzounov 2018 LAIC chain.","ots_proof_path":null,"outcome":{"lookahead_hours":8760.0,"min_magnitude":null,"type":"earthquake_within_30d_M3plus"},"registered_at":"2026-05-12T00:45:59Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"binomial_2sided","n_permutations":null},"threshold":{"direction":"positive","metric":"true_positive_rate","value":0.3},"title":"Schumann LAIC EQ-precursor classifier ALERT/CRITICAL sites have elevated M3+ EQ rate within 365 days","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-05-12"},{"data_source":"cross_physics.eclipse_prediction_monitor.cross_correlate_eclipses_with_signal + ledger + iris_mustang.db + rs_etas_caldera_forecaster output","expected_direction":"null","feature_set":["epm_days_to_next_solar_eclipse","epm_eclipse_season_active","microseism_secondary_band_3s","cuf_combined_rate_per_day_7d","ledger_m5_event_rate_global"],"git_commit_at_drafting":"05a0929d6e2b4630d3c5614c055281da43906c76","hypothesis_id":"20260517_solar_eclipse_epochs_do_not_co_DRAFT","hypothesis_text":"Solar eclipse epochs (+/- 7 days, ~5400 historical events 2016-2026 from the eclipse_prediction_monitor backfill) do NOT show statistically significant elevated M5+ earthquake rate, microseism_secondary_band_3s amplitude, or RS+ETAS caldera unrest score (cuf_combined_rate_per_day_7d) compared to N=N_eclipses randomly-sampled control windows in the same time range. The null is rejected only if Welch's two-sample t-test returns p < 0.01 in ANY of the 3 signal channels. Pre-registration mirrors the Phase 5 microseism / Fischbach 1986 / Forbush triple-null methodology -- expected outcome is publishable null, validating that gravitational tidal forcing from eclipse-geometry alignments does not exceed ordinary syzygy stress modulation that is already captured by the lunar_tidal + syzygy_proximity channels.","locked_at":"2026-05-18T03:24:24Z","locked_file_sha256":"af5c5286cbeefd6e8d3451d8593add252dd598747370616927accbd7a1930a94","notes":"Antikythera ~150 BC predicted eclipses with 30 bronze gears via Saros + Inex cycles. DeepMap reproduces that math (cross_physics.eclipse_prediction_monitor) and adds the hypothesis test the Greeks could not run: does the orbital geometry that produces eclipses also correlate with terrestrial geophysical anomalies? Pre-registration prevents post-hoc tuning. Lock + Bitcoin-OTS anchor (4 calendars) BEFORE evaluating. Re-evaluation window: 2026-11-17 (6 months forward). To lock manually:\n  python scripts/hypothesis_registry.py lock 20260517_solar_eclipse_epochs_do_not_co_DRAFT --ots\nThen rerun cross_correlate_eclipses_with_signal at 2026-11-17 for each of the 3 signal channels and record_result.","outcome":{"expected_result":"null_supports_no_effect across all 3 signal channels","lookahead_hours":168.0,"type":"eclipse_cross_correlation"},"references":["Freeth et al. 2006, Nature 444:587 (Antikythera decoding)","Meeus 1998, Astronomical Algorithms 2nd ed. ch. 54 (Saros + Inex)","van den Bergh 1955, Periodicity and Variation of Solar Eclipses (Inex cycle)","Phase 5 microseism universal precursor pre-reg (20260504_microseism_universal_precursor_c_99f4c2c7)","Forbush triple-null methodology, memory/cosmic_ray_phase_findings_may02.md"],"registered_at":null,"registered_by":"claude (DRAFT for Mike to lock + Bitcoin-anchor)","status":"locked","test":{"extra":{"backfill_years":10,"n_null_samples":"equal_to_n_eclipses","n_permutations":0,"window_days":7.0},"method":"welch_t_test"},"threshold":{"direction":"below_to_reject_null","metric":"p_value","value":0.01},"title":"Solar eclipse epochs do NOT statistically modulate M5+ EQ rate, microseism amplitude, or caldera unrest score within +/- 7 days","withdrawal_rule":"If fewer than 30 eclipses fall within the backfill window after data cleaning, declare UNDERPOWERED and re-register with looser parameters (longer backfill or shorter window).","_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-11-17"},{"data_source":"data/private/inner_core/daily_rotation_rate_2009_2026.csv","expected_direction":"positive","feature_set":["icrd_inferred_inner_core_rotation_rate_deg_per_yr"],"git_commit_at_registration":"d2c1eddba4466382041676525045c64470e8679e","hypothesis_id":"20260518_inner_core_daily_rotation_h2_sub_0750045d","locked_at":"2026-05-18T17:08:53Z","locked_file_sha256":"c4875d0565f49fe8d8576819e825cf2492c204444fd37048a91dcfc3175ccf7b","notes":"FORWARD 36-month test (lock 2026-05-18, re-eval 2029-05-18). Hypothesis: after detrending the inner-core daily rotation rate timeseries (computed via 0S2 multiplet splitting on IGETS gravity, EOST-only deduped reader post commit dbd2ab73) with a 3rd-order polynomial removing decadal drift, the Lomb-Scargle periodogram of the residual will show at least one peak at period 0.5-3 yr with z >= 3.0 vs a 1000-permutation random-phase null. This is HYPOTHESIS-GENERATING -- never been definitively detected at this band; if real, would extend Yang & Song 2023 decadal-cadence rotation reversal claim to sub-decadal scale. Data source: cron-driven daily compute_0s2_splitting_for_window from Track C continuous run (PID 1952580 on AX41 + PID 897979 on AX42-U started 2026-05-18, fresh on EOST-only code).","ots_proof_path":null,"outcome":{"lookahead_hours":25920.0,"min_magnitude":null,"type":"intrinsic_oscillation"},"registered_at":"2026-05-18T17:04:48Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"detrend":"decadal_polynomial_order_3","period_band_yr":[0.5,3.0]},"method":"lomb_scargle_periodogram_vs_permutation_null","n_permutations":1000},"threshold":{"direction":"positive","metric":"lomb_scargle_peak_z_score","value":3.0},"title":"Inner-core daily rotation H2: sub-decadal 0.5-3yr oscillation","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2029-05-18"},{"data_source":"data/private/inner_core/daily_rotation_rate_2009_2026.csv + USGS M7+ catalog","expected_direction":"negative","feature_set":["icrd_inferred_inner_core_rotation_rate_deg_per_yr","usgs_m7_plus_count"],"git_commit_at_registration":"d2c1eddba4466382041676525045c64470e8679e","hypothesis_id":"20260518_inner_core_daily_rotation_h3_ant_4b20eae9","locked_at":"2026-05-18T17:08:53Z","locked_file_sha256":"916315bf3416513634439de2c7501f9a08023004eb143c340ec2a01b643eeb93","notes":"FORWARD 36-month test (lock 2026-05-18, re-eval 2029-05-18). Bendick & Bilham 2017 GRL extension: 6-yr LOD oscillation correlates with global M7+ rate at r ~0.6. We test whether the inner-core differential rotation rate from 0S2 splitting (independent of LOD) shows the SAME anti-correlation at finer (30-180d) lag. Hypothesis: Pearson r between detrended rotation residual + USGS M7+ count, scanned over lag 30-180d, peaks at r <= -0.20 with permutation p < 0.01. Data: Track C continuous-cron output + USGS catalog (curated daily). Withdraw rule: if fewer than 50 lag samples accumulate, declare underpowered + register a longer-window variant. EOST-only deduped reader (commit dbd2ab73) is the lock-time data path.","ots_proof_path":null,"outcome":{"lookahead_hours":25920.0,"min_magnitude":null,"type":"rotation_eq_coupling"},"registered_at":"2026-05-18T17:05:10Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"detrend":"decadal_polynomial_order_3","lag_band_days":[30,180],"min_magnitude":7.0},"method":"lagged_pearson_correlation_vs_permutation_null","n_permutations":1000},"threshold":{"direction":"negative","metric":"pearson_r","value":-0.2},"title":"Inner-core daily rotation H3: anti-correlates with M7+ EQ rate at lag 30-180d","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2029-05-18"},{"data_source":"cross_physics.lithium_brine_classifier_monitor,connectors/lithium_brine_classifier.yaml","expected_direction":"positive","feature_set":["lbc_composite_score","lbc_class_id","lbc_nearest_known_km"],"git_commit_at_registration":"dc8803af8935a729ddf8d0eeb0bdb7009dca6c69","hypothesis_id":"20260519_deepmap_a_4_lithium_brine_classi_c64bd066","locked_at":"2026-05-19T17:58:36Z","locked_file_sha256":"5587b936412932a4c5bcc00cb2d3254f8c822629d5373f082e87a6b40960cb45","notes":"Forward test of the A.4 5-axis lithium brine classifier (commit c11600c2, 12 features prefix lbc_).  Validation cohort at registration: 5 producers ELITE (Atacama 82.0 / Salton Sea 79.2 / Clayton Valley 78.0 / Great Salt Lake 78.0 / Smackover 70.2) vs 3 controls (Yellowstone 20.0 / Atlantic abyssal 16.2 / Wasatch granite 42.5).  Producer-min vs control-max discrimination ratio 1.65x.  Pre-registered hypothesis: when applied to a random 200-site Western US continental scan, top-decile sites (composite >= 70) will contain a known producer or USGS MRDS Li/B/Cs/Rb pathfinder occurrence at >=60 percent rate.  Null baseline (random 5-axis weighted): expected ~30 percent by base rate.  2x lift = confirming.  Re-eval date 2027-05-19.  Lockfile hash committed pre-anchor; OTS Bitcoin anchor pending operator run on AX41 (deferred per local-ots-broken-on-py3.14 note in MEMORY.md).  Confirming result unlocks: (a) A.4 production cost-tier removal (drill_recommendation > 80), (b) publishable result in critical-minerals journal (J Geochem Explor or Econ Geol), (c) commercial CIP filing path Q3 2026 per YAML ip_status block.","ots_proof_path":null,"outcome":{"lookahead_hours":8760.0,"min_magnitude":null,"type":"lithium_brine_resource"},"registered_at":"2026-05-19T17:58:25Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"top_decile_producer_density","n_permutations":null},"threshold":{"direction":"positive","metric":"top_decile_known_producer_or_pathfinder_recall","value":0.6},"title":"DeepMap A.4 lithium brine classifier top-decile (composite >= 70) producer-density precision >= 60 percent over 1-year forward window","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-19"},{"data_source":"IGETS 8-station network 2015-2025 + Phase D.0 expanded 25-station catalog","expected_direction":"positive","feature_set":["slichter_3.5h_period_snr","slichter_4.5h_period_snr","slichter_6.0h_period_snr","slichter_triplet_splitting_sigma","slichter_sigma_floor_microgal"],"git_commit_at_registration":"53769e0888c7951aae018f00336c1ae772e87d9f","hypothesis_id":"20260519_slichter_mode_v2_detection_or_pu_ee86f619","locked_at":"2026-05-19T18:51:25Z","locked_file_sha256":"66177049797fde965b7f4db7be73e82c00ec2ca1cd81a247714ecf164e0dc96b","notes":"D.1-SPRINT forward test of the Slichter inner-core translational triplet (3.5h / 4.5h / 6.0h periods).  Pre-registered prior to running v2 pipeline.  Tests Smylie 1992 disputed claim vs Rosat 2005 null.  v2 = v1 (today, NULL) plus 3 D.1-SPRINT calibration pieces: (1) Hinderer 1995 atmospheric loading correction, (2) ICET-pre-corrected data + harmonic post-subtraction, (3) per-station FG5 calibration audit + uniform K-scale renormalization.  Data sufficiency: 8 IGETS stations x 13-25 years already EXCEEDS Rosat 2005's 7 x 12.  Pass criterion is ONE of: (a) DETECTION: SNR >= 5 at any of 3 canonical Slichter periods with triplet splitting consistent with Smylie 1992 framework (Nobel-tier finding); OR (b) PUBLISHABLE NULL: sigma floor constrained to <= 0.1 microGal at the 3 periods (5x tighter than Rosat 2005's 0.5 microGal floor; JGR / GJI / EPSL paper).  Underpowered = neither pass condition met.  Re-eval 2027-05-19 (12 months) -- allows full 3-week sprint engineering + 6-month additional IGETS data accumulation.  Anchor pending OTS Bitcoin proof on AX41.  Scoping doc at reports/d1_sprint_scoping_2026_05_19.md; engineering pieces 1 (Hinderer) + 3 (FG5) shipped this session at cross_physics/hinderer_atmospheric_loading_correction.py + cross_physics/fg5_calibration_registry.py.  Piece 2 (ICET) deferred to sprint week 2.","ots_proof_path":null,"outcome":{"lookahead_hours":8760.0,"min_magnitude":null,"type":"free_oscillation_inner_core_translational_triplet"},"registered_at":"2026-05-19T18:51:13Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"multi_station_coherent_stack_with_d1_sprint_calibration_pieces","n_permutations":null},"threshold":{"direction":"positive","metric":"slichter_period_snr_or_sigma_floor","value":5.0},"title":"Slichter mode v2 detection-or-publishable-null after D.1-SPRINT calibration cleanup","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-19"},{"data_source":"IGETS SuperConducting Gravimeter Delta-g residuals via cross_physics/igets_sub_hz_transient_monitor.py + cross_physics/igets_secular_monitor.py; n>=8 stations spanning 2009-2026 (Strasbourg, Walferdange, Onsala, Sutherland, Schiltach, Larzac, Pecny, Conrad)","expected_direction":"positive","feature_set":["per_station_igets_dg_residual_after_tide_removal","per_station_igets_dg_residual_after_atm_loading_removal"],"git_commit_at_registration":null,"hypothesis_id":"20260520_g3_dm_annual_modulation_per_station_igets_residual","locked_at":"2026-05-20T21:05:39Z","locked_file_sha256":"f9da0858486c313b8e778dfa046192bdb5a494c487d867a205fd6e0ec081c1a6","notes":"Pre-registered narrow-channel forward test of the G3 DM annual modulation moonshot. The G3 v1 SEED's tensor-L2 backtest produced a single-arm 3.1 sigma aggregate candidate (Lomb-Scargle p=0.001, peak period 440 d, phase offset +47.7 d from June 2) on retro_tensor 2005-2018 (122,712 hourly snaps, subsampled to 20,000). HOWEVER per-year stability check showed phase scatter across the full circle (DOY 43 to DOY 327) over 7 two-year sub-windows -- the aggregate signal is a coincidence of averaging over 14 years of phase-scattered annual cycles, NOT a coherent modulation. Linear detrending does NOT change p=0.001 because the L2 noise std (~3.9e+14) is dominated by tensor inventory drift during 2005-2018 (modules came online progressively). PER TASK INSTRUCTIONS the candidate-positive triggered this pre-registration; the forward test moves to the RIGHT observable: per-station IGETS Delta-g residuals after Earth-tide subtraction (Hartmann-Wenzel 1995) AND atmospheric loading subtraction (Boy & Hinderer 2006). DM annual modulation theory (Drukier-Freese-Spergel 1986; Lewin-Smith 1996) predicts coupling to baryonic targets at ~7%-level amplitude with phase peak near June 2 (DOY 152.5). Pass criterion: >=4 of 8 IGETS stations independently show p<0.05 LS power AT period 300-440 d AND phase within +/- 60 d of June 2 AND amplitude in [0.5%, 10%] window AND cross-station phase stddev <= 30 d. Withdraw rule: if fewer than 4 stations have >=3 years of cleaned data by 2027-05-20 the test is underpowered and re-registers with looser cross-station consistency. No claim if the per-station scan shows DOY-randomness like the L2 scan did. Re-evaluation: 2027-05-20 (12 months forward). Bitcoin OTS anchor pending user lock + manual ots stamp.","ots_proof_path":null,"outcome":{"lookahead_hours":8760,"min_magnitude":null,"type":"dm_annual_modulation_per_station_consistent_phase_and_amplitude"},"registered_at":"2026-05-20T22:35:00Z","registered_by":"claude@DeepMap","result":null,"status":"locked","test":{"extra":{"backtest_script":"scripts/backtest_g3_dm_annual_modulation.py","cross_station_phase_consistency_max_stddev_days":30,"dama_canonical_amplitude_percent_window":[0.5,10.0],"dama_canonical_peak_day_of_year":152.5,"ls_period_range_days":[300,440],"ls_phase_tolerance_days":60,"min_n_stations_required_for_pass":4,"n_bootstrap_permutations":1000,"per_station_pvalue_threshold":0.05,"tensor_l2_aggregate_result_disclaimer":"Aggregate L2-norm scan on retro_tensor (122k snaps 2005-2018) produced p=0.001 + phase offset +47.7d from DAMA's June 2; this CANDIDATE-POSITIVE COLLAPSED on per-year inspection (phase ranges DOY 43 to DOY 327 randomly across 2005-2007/2007-2009/.../2017-2019 sub-windows). The L2 norm is dominated by raw-units channels + tensor inventory drift, not by a coherent annual signal. This forward test moves to per-station IGETS residuals where DM coupling theory predicts a real signature."},"method":"per_station_lomb_scargle_dama_phase_check","n_permutations":1000},"threshold":{"direction":"positive","metric":"n_stations_with_pvalue_below_0_05_AND_phase_within_60d_of_june_2_AND_amplitude_in_window","value":4.0},"title":"G3 v2 forward test: DM annual modulation in per-station IGETS Delta-g residuals (de-tided + de-atmosphered), NOT in tensor L2 norm","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-20"},{"data_source":"cross_physics/solar_gmode_discovery_monitor.py + data/igets_sg/gmode_band_residuals.db (populated by future scripts/refresh_solar_gmode_amplitudes.py, pending). v2 backtest harness is scripts/backtest_g10_solar_gmode.py","expected_direction":"negative","feature_set":["sgmd_3hr_period_z","sgmd_6hr_period_z","sgmd_9hr_period_z","sgmd_12hr_period_z","sgmd_24hr_period_z","sgmd_n_modes_above_z3","sgmd_max_abs_z","sgmd_network_coherence"],"git_commit_at_registration":"7959ac04677f0b241ab0196814c30c0a24c29e89","hypothesis_id":"20260520_phase_d_5_g10_solar_g_mode_disco_06a0397f","locked_at":"2026-05-20T04:07:50Z","locked_file_sha256":"e52c4e99a98d17b482f6e3610954e776e48f4582a32817634a05a5194eefca43","notes":"FORWARD test of the Phase D.5-G10 Solar G-Mode Discovery moonshot. G10 v1 SEED is cross_physics/solar_gmode_discovery_monitor.py monitoring 5 candidate g-mode periods (3, 6, 9, 12, 24 hr); the v2 backtest restricts to the 3 Garcia 2007 + Provost 2000 + Belkacem 2009 theoretical candidate frequencies in the 60-90 min period band (220, 240, 270 microHz) where g-mode lines are most strongly predicted. IGETS SG superconducting-gravimeter coherent-stack network of 7+ stations (Strasbourg, Wettzell, Canberra, Cantley, Concepcion, Sutherland, Mizusawa per Phase D.0 expansion) is queried at daily cadence across 23 years. PASS criterion is conjunctive 3-arm: (a) period-folded coherent-stack peak amplitude exceeds 3x sliding off-band noise floor at >= 2 of 3 candidate frequencies AND >= 5 of 7 stations contribute coherently (per-station |coeff| >= 3x per-station off-band median), AND (b) <= 1 of 5 deliberately-chosen decoy frequencies (193, 208, 231, 256, 285 microHz) in the same band shows a spurious peak at the same detection threshold, AND (c) random off-band resampling bootstrap p-value <= 0.05 (N >= 1000 perms; each perm replaces each station's on-band coefficient with a random draw from that station's own off-band Goertzel coefficient pool, preserving per-station noise statistics while destroying any genuine cross-station g-mode coherence). The load-bearing scalar reported to record-result is the bootstrap p-value. Garcia 2007 GOLF/SoHO g-mode candidates remain disputed (Fossat 2017 detection claim refuted by Schunker 2018 + Appourchaux & Corbard 2019); G10 tests for INDEPENDENT confirmation OR disconfirmation via gravity instead of helioseismology Doppler velocity. Predicted signal strength is 10^-13 g_local per Provost 2000 + 1 AU gravitational coupling, right at the IGETS SG nanogal precision floor after sqrt(7) ~ 2.6x stacking boost. References: Garcia et al 2007 Science (GOLF g-mode candidates); Appourchaux et al 2010 Astron Astrophys Rev (helioseismology review); Fossat et al 2017 Astron Astrophys (disputed g-mode detection claim); Schunker et al 2018 Astron Astrophys (refutation reanalysis); Scherrer & Gough 2019 ApJ (further refutation); Provost 2000 (interior g-mode amplitude predictions); Belkacem et al 2009 Astron Astrophys (mode-amplitude theory); Crossley et al 1999 Phys Earth Planet Inter (IGETS SG noise floor characterization); Hinderer et al 2007 (superconducting gravimetry review). Companion to Phase D.5 G1 mass tomography, G2 DM compact transit, G3 DM annual modulation, G4 LOD anomaly, G5 inner-core wandering, G6 Earth mass redistribution, G7 pre-volcanic mass inflation, G8 pre-EQ free-oscillation excitation, G9 cosmic-ray terrestrial mass modulation. Re-eval 2027-05-21. CONDITIONAL on operational IGETS SG gmode-band residual cache (data/igets_sg/gmode_band_residuals.db with per-station daily standardized residual columns) and on the cache spanning >= 7 stations with >= 1500 daily samples each; until then the live backtest will return UNDERPOWERED. Synthetic smoke-test PASS verified 2026-05-19 with bootstrap p=0.000999 on 8395-day per-station history across 7 stations with deliberate coherent ~10^-13 g_local injections at the 3 candidate frequencies, 1000 off-band resampling permutations, all three arms PASS (candidate_hits=3 of 3, decoy_spurious=0 of 5). OTS Bitcoin anchor deferred to AX41 (Python 3.14 ots binary broken locally). WITHDRAW criterion: if at re-eval the live cache lacks per-station residual columns or fewer than 5 stations span >= 1500 daily samples, declare UNDERPOWERED rather than null and re-register with extended timeline.","ots_proof_path":null,"outcome":{"lookahead_hours":8760.0,"min_magnitude":null,"type":"solar_gmode"},"registered_at":"2026-05-20T04:07:39Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"candidate_freq_coherent_stack_with_decoy_window_budget_and_off_band_resampling_null","n_permutations":1000},"threshold":{"direction":"negative","metric":"bootstrap_pvalue","value":0.05},"title":"Phase D.5-G10 Solar G-Mode Discovery v2 backtest","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-21"},{"data_source":"cross_physics/free_inner_core_wobble_discovery_monitor.py + data/iers_c04_eop.db (polar-motion residuals after Phase 65r v2 background subtraction: Chandler+annual+FCN+AAM+OAM+HAM). v2 backtest harness is scripts/backtest_g11_free_inner_core_wobble.py","expected_direction":"negative","feature_set":["ficw_3day_period_z","ficw_5day_period_z","ficw_7day_period_z","ficw_9day_period_z","ficw_14day_period_z","ficw_n_modes_above_z3","ficw_max_abs_z","ficw_cross_instrument_coherence"],"git_commit_at_registration":"7959ac04677f0b241ab0196814c30c0a24c29e89","hypothesis_id":"20260520_phase_d_5_g11_free_inner_core_wo_1c7c8b93","locked_at":"2026-05-20T04:05:33Z","locked_file_sha256":"da50102f02aebe56185684f2b158ed36cae8f78e6bf649754e1085e6ad07f16c","notes":"FORWARD test of the G11 Free Inner Core Wobble (FICW) discovery detector (Phase D.5 11th moonshot). G11 v1 SEED is cross_physics/free_inner_core_wobble_discovery_monitor.py monitoring five candidate FICW periods (3d/5d/7d/9d/14d); the backtest restricts to the three primary Mathews 1991 + Mathews 2002 inner-core periodicity predictions (3d/5d/7d). PASS criterion is conjunctive 3-arm: (a) period-folded peak amplitude divided by sliding noise floor is >= 3.0 (SNR >= 3x) in >= 1 of 3 candidate periods, AND (b) spurious-peak rate across 12 non-candidate band positions (2.0/2.5/3.5/4.0/4.5/5.5/6.0/6.5/7.5/8.0/9.0/10.0 d) at the >= 2x noise threshold is <= 0.20 of windows, AND (c) random-phase bootstrap p-value on the highest-SNR candidate period is <= 0.05 (N >= 1000 perms). The load-bearing scalar reported to record-result is the bootstrap p-value. Detection would be the FIRST direct observation of inner-core wobble; FICW remains UNCONFIRMED 35 years after theoretical prediction (Mathews 2002 J Geophys Res; Mathews and Guo 2005). Distinct from the outer-core Free Core Nutation operationalized in Phase 65r v2 (FCN period 380-450 d). References: Mathews et al 1991 GJI (FICW theoretical prediction); Chao 1989 GJI (alternative inner-core model with shorter period); Mathews 2002 J Geophys Res (Earth-rotation theory + FICW); Dehant and Mathews 2015 (Treatise on Geophysics 2nd ed vol 3 review); Krasna et al 2013 (FCN observability via IERS C04); Smylie 1992 GJI (disputed sub-seismic translational mode claim); Bos and Scherneck 2013 (mass-loading correction models). Companion to Phase D.5 G1 mass tomography, G2 DM compact transit, G3 DM annual modulation, G4 LOD anomaly, G5 inner-core wandering, G6 Earth mass redistribution, G7 pre-volcanic mass inflation, G8 pre-EQ free-oscillation excitation, G9 cosmic-ray terrestrial mass modulation. Re-eval 2027-05-21. CONDITIONAL on operational IERS C04 polar-motion-residual cache (data/iers_c04_eop.db with polar_motion_residuals table holding xp_residual_uas + yp_residual_uas columns post Phase 65r v2 background subtraction) and on the cache spanning at least 3 years for Nyquist stability at the 14-day end of the candidate band; until then the live backtest will return UNDERPOWERED. Synthetic smoke-test PASS verified 2026-05-19 with bootstrap p=0.000999 on 8540-day history with deliberate 7-day FICW injection at 10x noise, 1000 permutations, all three arms PASS (best SNR 558.6 at 7d, spurious rate 0.167, n_passing_snr=1). OTS Bitcoin anchor deferred to AX41 (Python 3.14 ots binary broken locally). WITHDRAW criterion: if at re-eval the live cache lacks the polar-motion residual table or spans fewer than 3 years, declare UNDERPOWERED rather than null and re-register with extended timeline.","ots_proof_path":null,"outcome":{"lookahead_hours":8760.0,"min_magnitude":null,"type":"earthquake"},"registered_at":"2026-05-20T04:05:22Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"period_folded_snr_with_spurious_band_budget_and_random_phase_bootstrap_null","n_permutations":1000},"threshold":{"direction":"negative","metric":"bootstrap_pvalue","value":0.05},"title":"Phase D.5-G11 Free Inner Core Wobble (FICW) v2 backtest","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-21"},{"data_source":"G1 v1 SEED upstream channel fusion (GRACE-FO Level 2 monthly + IGETS L2 SG residuals + Sentinel-1 InSAR SBAS + NGL daily IGS14 GNSS)","expected_direction":"positive","feature_set":["mtm_inversion_confidence_0_1","mtm_mass_anomaly_top_5km_gt","mtm_mass_anomaly_5_30km_gt","mtm_grace_twsa_cm","mtm_igets_gravity_anomaly_ugal","mtm_insar_vertical_mm_yr","mtm_gnss_vertical_mm_yr","mtm_data_real"],"git_commit_at_registration":"ad0538a0bc749eb155622daa78fc13e291db2000","hypothesis_id":"20260520_phase_d_5_g1_mass_tomography_cor_51cc427f","locked_at":"2026-05-20T02:46:41Z","locked_file_sha256":"c52a8389f063be769423848c1fabfa894c12472606e92168a65b74b5deae6f43","notes":"Phase D.5-G1 v2 12-month forward test. re-eval YYYY-MM-DD = 2027-05-21. PASS = at least 7 of 10 documented mass-anomaly sites (California Central Valley + Indo-Gangetic Plain + Greenland + WAIS + Decatur CCS + Yellowstone + Long Valley + Amazon + Sumatra-Andaman + Lake Mead) classify ELITE or STRONG under the G1 inversion AND at most 1 of 5 control sites (South Pacific Gyre + Sahara + Canadian Shield + Atlantic mid-lat + Western Australia) classify STRONG or above. Backtest harness: scripts/backtest_g1_mass_tomography.py (synthetic smoke-test PASSED 8/10 + 0/5). Production runs require IGETS+GRACE+InSAR+GNSS caches populated on AX41+CPX41. Null baseline: random ELITE/STRONG assignment at 50% per-site would give 2-3 of 10 known hits + 2-3 of 5 control false positives; PASS requires the inversion to discriminate hot-spots from stable Earth at >>chance. Withdraw rule: if v1 SEED upstream channels return data_real=0 at >=8 of 10 known sites (i.e., upstream caches not populated on production AX41 by 2027-05-21), declare UNDERPOWERED and re-register with v2 Watkins mascon inversion.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"mass_anomaly_classifier"},"registered_at":"2026-05-20T02:46:30Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"ground_truth_strong_or_elite_count_with_control_fpr","n_permutations":null},"threshold":{"direction":"positive","metric":"n_known_strong_or_elite","value":7.0},"title":"Phase D.5-G1 mass tomography correctly classifies 7-of-10 known mass-anomaly sites as ELITE/STRONG with at most 1-of-5 control false positives","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-21"},{"data_source":"scripts/backtest_g2_dm_compact_transit.py + cross_physics/dm_compact_transit_monitor.classify_candidate + data/dm_compact_transit_candidates.db when present","expected_direction":"positive","feature_set":["dct_max_5domain_coincidence_score","dct_inferred_transit_velocity_km_s","dct_inferred_charge_proxy","dct_atmospheric_entry_signature_0_1","dct_transit_candidate_count_24h"],"git_commit_at_registration":"ad778b03924f1593a05364a1707b7d78aea1cdf3","hypothesis_id":"20260520_phase_d_5_g2_dm_compact_transit__23320a39","locked_at":"2026-05-20T02:58:32Z","locked_file_sha256":"ae823abdc5e908e4b34ba5396522033bc6d18d79d451975ed2f920dfd008dd20","notes":"FORWARD test of the Phase D.5-G2 v2 backtest harness (scripts/backtest_g2_dm_compact_transit.py).  G2 is the second of three Phase D.5 moonshots (companion to G1 Mass Tomography + G3 DM Annual Modulation).  Test: 50 simulated transit injections at coincidence_score >= 5.0 (18 PBH + 18 strangelet + 14 Q-ball/macroscopic), classified via cross_physics.dm_compact_transit_monitor.classify_candidate; 200 control epochs from mixture (65% sub-threshold noise, 25% bolide-like 15-45 km/s, 7% velocity-rejected >600 km/s, 3% accidentally-DM-like).  Bootstrap null FPR with N=2000 permutations.  PASS requires ALL THREE arms: (a) detection_count >= 40 (80% of 50), (b) FPR_count <= 10 (5% of 200), (c) observed FPR <= bootstrap p95.  Re-eval 2027-05-21 (12 months) on AX41 production with live candidate DB data instead of synthetic injections.  WITHDRAW criterion: if AX41 candidate DB still produces < 5 candidates after 9 months of accumulated tensor scanning, declare underpowered (live SG/IGETS noise floor was insufficient for SNR >= 5 candidate generation) and re-register with relaxed SNR floor 3.5.  Pass criterion documented as: PASS = (n_detected >= 40) AND (n_false_positives <= 10) AND (false_positive_rate <= null_fpr_p95).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_compact_transit"},"registered_at":"2026-05-20T02:58:20Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"bootstrap_null_injection_recovery","n_permutations":2000},"threshold":{"direction":"positive","metric":"n_detected","value":40.0},"title":"Phase D.5-G2 DM compact transit: 5-domain coincidence classifier detects >=80% of simulated transits at SNR>=5 with FPR<=5%","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-21"},{"data_source":"G3 v1 SEED tensor L2 norm history from data/earth_state_tensor.db","expected_direction":"negative","feature_set":["dam_orbital_phase_radians","dam_expected_modulation_signed","dam_fit_amplitude_pct","dam_fit_phase_offset_days","dam_pvalue_modulation","dam_n_years_fit","dam_data_real"],"git_commit_at_registration":"12e0e0d54669cfbb68ead6ff18a19fa1624b0928","hypothesis_id":"20260520_phase_d_5_g3_dark_matter_annual__cb491996","locked_at":"2026-05-20T03:01:53Z","locked_file_sha256":"7a01edb3c43d343651e0abb4f1135dd21e9d6618136cbdd5797c12d655b95c68","notes":"Phase D.5-G3 v2 12-month forward test. re-eval YYYY-MM-DD = 2027-05-21. PASS = ALL three arms succeed: (a) Lomb-Scargle peak period within 300-440 days, (b) peak phase within +/- 60 days of June 2 (DAMA canonical day-of-year 152.5 per Drukier-Spergel 1986), AND (c) bootstrap p-value <= 0.05 from >= 1000 time-shuffle permutations at the canonical 365.25-day period. Backtest harness: scripts/backtest_g3_dm_annual_modulation.py (synthetic smoke-test PASSED 3-of-3 arms; n_samples=70,128 2-year synthetic, recovered amplitude 1.44%/1.5% injected, phase offset 54d within 60d tolerance, bootstrap p=0.001/0.05). Production runs require tensor history >= 180 days span and >= 60 samples on AX41+CPX41. Null baseline: under no annual modulation, the LS amplitude at the canonical period should not exceed the 95th percentile of time-shuffled nulls; pass requires the observed signal to be a >= 1-in-20 event under the null. Withdraw rule: if tensor history span < 180 days or < 60 samples at the re-evaluation date (i.e., upstream caches not accumulated on production AX41 by 2027-05-21), declare UNDERPOWERED. Tension: DAMA's claim is contested by LUX 2018 + COSINE-100 2021 + ANAIS-112 + XENON1T 2020 modulation null results; if G3 passes it joins DAMA in seeing a signal that NaI/Xe direct-detection counterparts do not; if it fails it provides an independent surface-based null at the published amplitude.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_annual_modulation_signature"},"registered_at":"2026-05-20T03:01:43Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"lomb_scargle_peak_with_bootstrap_null_and_phase_check","n_permutations":1000},"threshold":{"direction":"negative","metric":"bootstrap_pvalue","value":0.05},"title":"Phase D.5-G3 dark matter annual modulation Lomb-Scargle peak amplitude exceeds bootstrap null at p<=0.05 with DAMA-consistent period and phase","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-21"},{"data_source":"G4 v1 SEED IERS C04 LOD history from data/iers_c04_eop.db + AAM/OAM/HAM/seismic attribution pipeline","expected_direction":"positive","feature_set":["loda_lod_residual_ms","loda_lod_anomaly_z","loda_unattributed_residual_ms","loda_attribution_class","loda_data_real"],"git_commit_at_registration":"123c5bf6ade9e93d31a49a598b44d419c3fc400c","hypothesis_id":"20260520_phase_d_5_g4_lod_anomaly_detecto_5a4d8ecb","locked_at":"2026-05-20T03:14:53Z","locked_file_sha256":"49aafe3e40f1baab71f1886c174684b43463e3899e29b8a61b43d8c4c75fa578","notes":"Phase D.5-G4 v2 12-month forward test. re-eval YYYY-MM-DD = 2027-05-21. PASS = BOTH arms succeed: (a) >=4 of 6 known LOD anomaly events (1992 El Nino peak +1.8ms, 1997-98 El Nino extreme +0.8ms, 2003 Mar acceleration +1.0ms, 2010 Maule M8.8 -1.26ms post-event step, 2020 Mar unexplained +1.5ms, 2024 mid-year +1.0ms) classified ELEVATED/ANOMALY (|z|>=2.5 AND unattributed_fraction>=0.50) against a 30-day local rolling baseline excluding +/-7d event window; AND (b) <=2 of 50 random quiet-control epochs (excluding +/-21d around any known event) falsely classified positive. Bootstrap null: 1000-permutation label-shuffle across the combined 56-epoch pool; one-sided p-value for observing TPR>=observed AND FPR<=observed under no-discrimination null. Headline metric reported to record-result is detection_rate=TP/6. Backtest harness: scripts/backtest_g4_lod_anomaly.py (synthetic smoke-test PASSED: 6/6 TP, 2/50 FP, detection_rate=1.0, bootstrap p=0.001). Production runs require IERS C04 cache spanning 1992-2024 on AX41 (current cache 510+ rows daily). Null baseline: under no discrimination the expected TP at the observed positive-rate is ~0.86 (binomial mean over 6 trials at 50/56 negative rate); pass requires the observed signal to be >=4x the no-skill mean. Withdraw rule: if the IERS cache spans <30 years at re-evaluation (i.e., production backfill incomplete) OR if fewer than 4 of 6 known events fall inside the cache window, declare UNDERPOWERED. Physics references: Chao & Gross 1987 GJI (EQ-induced LOD steps); Lambeck 1980 (Earth's variable rotation); Yang & Song 2023 Nat Geosci (inner-core rotation variation); Chao 1995 GRL (ENSO LOD coupling); Brzezinski & Nastula 2004 GRL (March 2003 anomaly); Vallee 2017 Science / Juhel 2018 GJI (pre-rupture mass redistribution test hypothesis).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"length_of_day_anomaly_event_detection"},"registered_at":"2026-05-20T03:14:41Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"z_score_classification_with_label_shuffle_bootstrap_null","n_permutations":1000},"threshold":{"direction":"positive","metric":"detection_rate_tp_over_6","value":0.667},"title":"Phase D.5-G4 LOD anomaly detector classifies >=4 of 6 known historical LOD anomaly events as ELEVATED/ANOMALY with <=2 FP across 50 quiet controls","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-21"},{"data_source":"cross_physics/inner_core_wandering_monitor.py + data/inner_core_daily.db (populated by inner_core_0s2_splitting_daily cron 04:25 UTC, schema extension to expose per-m sub-component splittings pending D.1-SPRINT week 3)","expected_direction":"negative","feature_set":["icw_axis_drift_rate_deg_per_yr","icw_multiplet_asymmetry_score","icw_yang_song_consistency_score","icw_anomaly_flag","icw_rosat_2005_null_compatibility"],"git_commit_at_registration":"27d6fe27369c8f10c2ecf59c8815616c0d357e47","hypothesis_id":"20260520_phase_d_5_g5_inner_core_wanderin_465c519d","locked_at":"2026-05-20T03:12:54Z","locked_file_sha256":"6e04281c4036fa6829015c4b7948b220306601266ca96d6b835d133446d8d729","notes":"FORWARD test of the G5 inner-core wandering detector (Phase D.5 5th moonshot). G5 v1 SEED is cross_physics/inner_core_wandering_monitor.py; v2 backtest harness is scripts/backtest_g5_inner_core_wandering.py. PASS criterion is conjunctive 3-arm: (a) anomaly_flag fires at least one wander event in the YS2009 window [2004-01-01, 2014-12-31] (i.e. near 2009 plus or minus 5 years per Mike spec), AND (b) total false-positive wander events across 10 control decades (1970s, 1980s, ... excluding the YS2009 window) is less than or equal to 2, AND (c) bootstrap p-value on the YS2009-window event count is less than or equal to 0.05 (N greater than or equal to 1000 time-shuffle permutations). The load-bearing scalar reported to record-result is the bootstrap p-value. Wander event detection threshold is axis drift rate magnitude greater than or equal to 0.10 deg/yr (2x the YS2023 upper-band 0.05 deg/yr). Reference: Yang Y. and Song X. 2023 Nat Geosci 16, 182 (Multidecadal variation of the Earth inner-core rotation); Tkalcic and Pham 2018 Science 362, 329 (J-wave detection of hemispheric anisotropy). Companion to G11 FICW moonshot and the Phase 65w-VERIFY-CLOSE 25-station IGETS network expansion. Re-eval 2027-05-21. CONDITIONAL on D.1-SPRINT completing pieces 1-3 (Hinderer atmospheric loading + ICET pre-corrected data + per-station FG5 calibration audit) which unlock the per-m sub-component SNR greater than or equal to 5 required for multi-station axis-orientation inversion; until then the live backtest will return UNDERPOWERED because the inner_core_daily cache lacks the axis_drift_rate_deg_per_yr column. Synthetic smoke-test PASS verified 2026-05-19 with bootstrap p=0.001 on 60-year monthly history with deliberately-injected 1-year-sigma YS2009 pause/reversal bump. WITHDRAW criterion: if at re-eval the live cache still lacks the axis_drift column or accumulates less than 5 control decades of data, declare UNDERPOWERED rather than null and re-register with extended timeline.","ots_proof_path":null,"outcome":{"lookahead_hours":8760.0,"min_magnitude":null,"type":"inner_core_rotation"},"registered_at":"2026-05-20T03:12:44Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"ys2009_reproduction_with_control_decade_budget_and_bootstrap_null","n_permutations":1000},"threshold":{"direction":"negative","metric":"bootstrap_pvalue","value":0.05},"title":"Phase D.5-G5 inner-core wandering detector reproduces Yang & Song 2023 2009 pause/reverse","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-21"},{"data_source":"cross_physics/earth_mass_redistribution_monitor.py + data/grace_fo.db (populated by GRACE-FO monthly mascon backfill, schema extension to expose global_mass_redistribution_score column pending D.4-SPRINT GRACE-FO backfill)","expected_direction":"negative","feature_set":["emr_global_mass_redistribution_score","emr_total_terrestrial_water_storage_z","emr_polar_ice_loss_mass_change_gt_yr","emr_anomaly_flag","emr_lod_consistency_with_mass_change"],"git_commit_at_registration":"2a4d9fea457bb0fd3355736e12e4084ada2479e0","hypothesis_id":"20260520_phase_d_5_g6_earth_mass_redistri_e839f743","locked_at":"2026-05-20T03:25:44Z","locked_file_sha256":"92852a7381f41e82ab33501fdc8cdeee536cec8699d76e5c4a0097318f2d3266","notes":"FORWARD test of the G6 Earth mass redistribution detector (Phase D.5 6th moonshot). G6 v1 SEED is cross_physics/earth_mass_redistribution_monitor.py; v2 backtest harness is scripts/backtest_g6_earth_mass_redistribution.py. PASS criterion is conjunctive 3-arm: (a) anomaly_flag fires within plus or minus 90 days of all 4 reference mass-redistribution events (Tohoku 2011-03-11 M9.0 coseismic step, Maule 2010-02-27 M8.8 coseismic step, Greenland 2019-07-15 extreme summer melt, WAIS Pine Island and Thwaites 2023-07-15 acceleration), AND (b) total false-positive mass-redistribution events across 30 quiet-control 6-month windows (180-day exclusion buffer around any reference event) is less than or equal to 2, AND (c) bootstrap p-value on the reference-event hit count is less than or equal to 0.05 (N greater than or equal to 1000 time-shuffle permutations). The load-bearing scalar reported to record-result is the bootstrap p-value. Composite-anomaly z-score threshold is magnitude greater than or equal to 2.5 (matches G4 LOD anomaly detector convention). References: Chao and Gross 1987 GJI (coseismic gravity steps); Han Sauber and Riva 2011 Science (GRACE detection of Tohoku step); Cambiotti and Sabadini 2012 (Maule step); Velicogna 2020 and Tedesco and Fettweis 2020 The Cryosphere (Greenland 2019 extreme melt); Rignot et al 2024 and IMBIE 2024 (WAIS acceleration); Cox and Chao 2002 Science (1998 J2 anomaly); Wahr 1995 GJI (time-variable gravity from GRACE). Companion to Phase D.5 G1 mass tomography, G2 DM compact transit, G3 DM annual modulation, G4 LOD anomaly detector, G5 inner-core wandering detector. Re-eval 2027-05-21. CONDITIONAL on D.4-SPRINT GRACE-FO monthly mascon archive backfill (NASA JPL RL06.1Mv03) and on the grace_fo_monthly cache exposing the global_mass_redistribution_score column; until then the live backtest will return UNDERPOWERED because the grace_fo.db cache lacks the composite-anomaly column. Synthetic smoke-test PASS verified 2026-05-19 with bootstrap p=0.003 on 23-year monthly history with deliberately-injected Gaussian bumps (sigma=0.10 yr, amplitude=12 sigma) at each of the 4 reference event dates, 1000 permutations, all three arms PASS. WITHDRAW criterion: if at re-eval the live cache still lacks the composite-anomaly column or accumulates less than 5 years of data spanning fewer than 2 of 4 reference events, declare UNDERPOWERED rather than null and re-register with extended timeline.","ots_proof_path":null,"outcome":{"lookahead_hours":8760.0,"min_magnitude":null,"type":"earth_mass_redistribution"},"registered_at":"2026-05-20T03:25:34Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"reference_event_reproduction_with_control_window_budget_and_bootstrap_null","n_permutations":1000},"threshold":{"direction":"negative","metric":"bootstrap_pvalue","value":0.05},"title":"Phase D.5-G6 earth mass redistribution detector reproduces 4 known mass events","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-21"},{"data_source":"G7 v1 SEED joint composite of grace_fo_aquifer_storage_monitor + insar_subsidence_monitor + microseism_precursor_tensor_monitor + volcanic_tremor_detector + polar_motion_residual_monitor, monthly cadence, 12-month pre-event window","expected_direction":"positive","feature_set":["pvi_grace_gravity_rise_microgal","pvi_insar_edifice_inflation_cm","pvi_microseism_amplitude_z","pvi_volcanic_tremor_score","pvi_polar_motion_anomaly_z","pvi_inflation_phase_months","pvi_alert_class","pvi_data_real"],"git_commit_at_registration":"7d3644354ac61b2dede7beb65675e5be5c8189ec","hypothesis_id":"20260520_phase_d_5_g7_pre_volcanic_mass_i_0a454003","locked_at":"2026-05-20T03:26:14Z","locked_file_sha256":"8c24484f2916d4373acfcc8d233ea95da7459961deb1ee7266a40514e2d64ef2","notes":"Phase D.5-G7 v2 12-month forward test. re-eval YYYY-MM-DD = 2027-05-21. PASS = BOTH arms succeed: (a) >=4 of 8 known VEI>=4 eruptions (Pinatubo 1991 VEI6, Eyjafjallajokull 2010 VEI4, Calbuco 2015 VEI4, Anak Krakatau 2018 VEI4, Sakurajima 2019 VEI4, Soufriere 2021 VEI4, Cumbre Vieja 2021 VEI4, Hunga Tonga 2022 VEI5) classified ELEVATED/INFLATION (peak_composite>=0.45 over 12-month pre-event window AND monotonic-trend slope>=0.02 per month over 6-month subwindow) where IGETS/GRACE/InSAR data exists; AND (b) <=2 of 50 quiet-control epochs (random sampling from 13-site active-volcano catalog 1990-2024 excluding +/-18mo around any known event) falsely classified positive. Bootstrap null: 1000-permutation label-shuffle across the combined 58-epoch pool; one-sided p-value (Laplace-smoothed) for observing TPR>=observed AND FPR<=observed under no-discrimination null. Headline metric reported to record-result is detection_rate=TP/8. Backtest harness: scripts/backtest_g7_pre_volcanic_mass_inflation.py (synthetic smoke-test PASSED 2026-05-19: 8/8 TP, 0/50 FP, detection_rate=1.0, bootstrap p=0.001). Production runs require GRACE-FO monthly + InSAR + microseism historical caches spanning 1990-2022 on AX41 (currently sparse pre-2010; v2 path is the historical-cache backfill). Null baseline: under no discrimination the expected TP at a 50/58 negative rate is ~1.1 on 8 trials (binomial mean); pass requires the observed signal to be >=3.6x no-skill mean. Withdraw rule: if fewer than 4 of 8 known events have >=3 monthly samples in the pre-event window at AX41 re-eval time, declare UNDERPOWERED. Companion: D.2-B4 Pre-Eruptive Magma Chamber is the operational alerts layer; G7 is the moonshot DETECTION question. Physics references: Battaglia et al. 1999 J Volcanol Geotherm Res (Long Valley CA microgravity preceding inflation); Carbone & Greco 2007 Pure Appl Geophys (Etna pre-eruptive gravity); Bonvalot et al. 2008 J Geophys Res (volcanic gravimetry); Furuya et al. 2003 J Geophys Res (Iwo-jima island time-variable gravity); Newhall & Self 1982 J Geophys Res (VEI); Lu & Dzurisin 2014 USGS PP (InSAR Aleutian volcanoes); Anderson 2019 GRL (Mauna Loa magma supply rate); Phase 5 microseism universal precursor (this repo, May 2026).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"pre_volcanic_mass_inflation_detection"},"registered_at":"2026-05-20T03:26:04Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"peak_composite_and_monotonic_trend_with_label_shuffle_bootstrap_null","n_permutations":1000},"threshold":{"direction":"positive","metric":"detection_rate_tp_over_8","value":0.5},"title":"Phase D.5-G7 pre-volcanic mass-inflation detector classifies >=4 of 8 known historical VEI>=4 eruptions as ELEVATED/INFLATION in the 12-month pre-event window with <=2 FP across 50 quiet controls","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-21"},{"data_source":"cross_physics/pre_eq_free_oscillation_excitation_monitor.py + data/igets_sg/fo_amplitudes.db (populated by scripts/refresh_igets_fo_amplitudes.py, pending). v2 backtest harness is scripts/backtest_g8_pre_eq_free_oscillation_excitation.py","expected_direction":"negative","feature_set":["peqfo_0s2_football_z","peqfo_0t2_torsional_z","peqfo_0s3_z","peqfo_n_modes_above_z2","peqfo_n_modes_above_z3","peqfo_max_abs_z"],"git_commit_at_registration":"19c0740723333b4f2a6518d12e6aeba820c03f7a","hypothesis_id":"20260520_phase_d_5_g8_pre_eq_free_oscilla_66d783af","locked_at":"2026-05-20T03:49:56Z","locked_file_sha256":"708857a37a127a94726220cdfd08c6a6623817a098b7c403bf0657134be5ed39","notes":"FORWARD test of the G8 Pre-Earthquake Free-Oscillation Excitation detector (Phase D.5 8th moonshot). G8 v1 SEED is cross_physics/pre_eq_free_oscillation_excitation_monitor.py monitoring six modes (0S0/0S2/0T2/1S0/0S3/0S4); the backtest restricts to the three primary spheroidal modes 0S2/0T2/0S3 per Park 1996 + Dahlen-Tromp seismology. PASS criterion is conjunctive 3-arm: (a) composite amplitude z-score exceeds magnitude 2.5 in any of the three modes within plus or minus 72 hours of at least 3 of 4 reference M8-plus events (Aceh 2004-12-26 M9.1, Maule 2010-02-27 M8.8, Tohoku 2011-03-11 M9.1, Iquique 2014-04-01 M8.2), AND (b) total false-positive at the same threshold across 30 quiet-control 72-hour windows (30-day exclusion buffer from any reference event) is less than or equal to 2, AND (c) bootstrap time-shuffle p-value on the reference-event hit count is less than or equal to 0.05 (N greater than or equal to 1000 perms). The load-bearing scalar reported to record-result is the bootstrap p-value. References: Park 1996 GJI (free-oscillation seismology framework); Park et al 2005 Science (Sumatra 2004 post-event excitation); Bilek 2003 J Seismol (free-oscillation post-event survey); Dahlen and Tromp 1998 Theoretical Global Seismology; Stein and Wysession 2003 (intro); Buland and Gilbert 1976 GRL (low-frequency seismology + Earth modes); Yamashita 1999 J Phys Earth (nucleation phase); Olson 2017 GRL (slow rupture nucleation). Companion to Phase D.5 G1 mass tomography, G2 DM compact transit, G3 DM annual modulation, G4 LOD anomaly, G5 inner-core wandering, G6 Earth mass redistribution, G7 pre-volcanic mass inflation. Re-eval 2027-05-21. CONDITIONAL on operational IGETS SG free-oscillation amplitude cache (data/igets_sg/fo_amplitudes.db with mode column 0S2/0T2/0S3 and amplitude_z column) and on the cache spanning 2004-2014 to cover all 4 reference events; until then the live backtest will return UNDERPOWERED. Synthetic smoke-test PASS verified 2026-05-19 with bootstrap p=0.00699 on 166560-hour history with deliberately-injected Gaussian bumps (sigma=24 h, amplitude 6-10 sigma above baseline) at each of the 4 reference event dates, 1000 permutations, all three arms PASS (TP=4, FP=0). OTS Bitcoin anchor deferred to AX41 (Python 3.14 ots binary broken locally). WITHDRAW criterion: if at re-eval the live cache lacks the per-mode amplitude column or fewer than 2 of 4 reference events fall inside the cache span, declare UNDERPOWERED rather than null and re-register with extended timeline.","ots_proof_path":null,"outcome":{"lookahead_hours":8760.0,"min_magnitude":8.0,"type":"earthquake"},"registered_at":"2026-05-20T03:49:45Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"reference_event_reproduction_with_control_window_budget_and_bootstrap_null","n_permutations":1000},"threshold":{"direction":"negative","metric":"bootstrap_pvalue","value":0.05},"title":"Phase D.5-G8 Pre-EQ Free-Oscillation Excitation v2 backtest","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-21"},{"data_source":"G9 v1 SEED 7-station NMDB residual (OULU/JUNG/APTY/ROME/BKSN/ATHN/AATB) + GRACE-FO local TWS, daily cadence, +/-180d pre-and-post-event window, 2002-2025 NMDB era","expected_direction":"negative","feature_set":["crtm_residual_anomaly_pct","crtm_grace_tws_local_mm","crtm_partial_correlation_r","crtm_correlation_p_value","crtm_n_months_overlapping","crtm_predicted_signal_strength_log10","crtm_n_stations_active","crtm_significant_pair_count","crtm_inversion_confidence_0_1","crtm_data_real"],"git_commit_at_registration":"979532d9ffd1347ac2169c3d290592f6588efa96","hypothesis_id":"20260520_phase_d_5_g9_cosmic_ray_terrestr_817f01fc","locked_at":"2026-05-20T03:48:15Z","locked_file_sha256":"aa8e94271849ae74dd44f8a8118ed86dc1537098d62b7b0d2d8f80ee54677837","notes":"Phase D.5-G9 v2 12-month forward test. re-eval YYYY-MM-DD = 2027-05-21. PASS = ALL three arms succeed: (a) >=4 of 5 reference mass-redistribution events (Tohoku 2011 M9, Maule 2010 M8.8, Sumatra-Andaman 2004 M9.1, Greenland 2019 extreme melt, WAIS Thwaites 2023 acceleration) classified positive (peak |z| >= 2.0 within +/-180d window after Storini-Pase barometric + Belov Forbush correction); (b) <=3 of up to 30 quiet-control epochs (randomly sampled from 2002-2025 excluding +/-365d around any reference event) falsely classified positive; (c) one-sided label-shuffle bootstrap p-value <= 0.05 on the joint (TP, FP) outcome. Headline metric reported to record-result is bootstrap_pvalue. Backtest harness: scripts/backtest_g9_cosmic_ray_terrestrial_mass_modulation.py (synthetic smoke-test PASSED 2026-05-19: 5/5 hits, 0/12 FP, bootstrap p=0.001). Production runs require multi-decade NMDB daily archive on AX41 (data/cosmic_ray/nmdb_*.csv) which is sparse pre-2010 -- expect UNDERPOWERED until full backfill lands. Phase 65t-AUDIT confirmed NMDB pipeline cleanup retained signed cosmic_ray_rate_anomaly (May 2 2026 retro fill recompute fixed sign-stripping bug); 7-station verified-live list adopted post Phase 65u-FORBUSH May 6 cleanup (NWKN dead, THUL decommissioned 2017, KIEL/SOPO/MCMU/DRBS/CALG returning no-data HTML). Predicted signal floor 10^-5 fractional cosmic-ray rate change per 1 m WEH redistribution over 100 km radius. Differs from Phase 5b (May 2 2026) cosmic-ray null findings (Forbush -> volcanic eruption, Forbush -> M5+ EQ, Forbush -> microseism amplitude -- ALL NULL at sub-week timescales) because G9 tests LONG timescales (months) where slow mass redistribution accumulates. Null baseline: under no discrimination at z=3.0 threshold and 2-of-30 control-positive rate, expected TP across 5 ref events is ~0.3 (binomial mean) -- pass requires observed >= 4 = >=13x no-skill mean. Withdraw rule: if fewer than 3 of 5 reference events have >= 30 daily samples available in the +/-180d window at AX41 re-eval time, declare UNDERPOWERED. Companion: D.5-G6 Earth mass redistribution + D.5-G7 pre-volcanic mass inflation = same physical observable class via different sensor (G6 GRACE-FO + LOD, G7 GRACE-FO + InSAR + microseism, G9 cosmic-ray attenuation). Physics references: Forbush 1937 Phys Rev (cosmic-ray storm decrease); Storini & Pase 2010 Adv Space Res (barometric correction); Belov et al. 2009 Adv Space Res (Forbush statistical analysis); Hess & Vallarta atmospheric cascade; Han et al. 2011 Science (GRACE Tohoku); Tedesco & Fettweis 2020 Cryosphere (Greenland 2019 melt); Rignot et al. 2024 (Thwaites collapse); Phase 5b cross-physics cosmic-ray null findings (this repo, May 2026).","ots_proof_path":null,"outcome":{"lookahead_hours":8760.0,"min_magnitude":null,"type":"cosmic_ray_terrestrial_mass_modulation_detection"},"registered_at":"2026-05-20T03:48:01Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"peak_window_zscore_classification_with_label_shuffle_bootstrap_null","n_permutations":1000},"threshold":{"direction":"negative","metric":"bootstrap_pvalue","value":0.05},"title":"Phase D.5-G9 Cosmic-Ray Terrestrial Mass Modulation v2 backtest","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-21"},{"data_source":"IGETS L2 (EOST + ICET when available) 2009-2026 + D.1-SPRINT correction pipeline (Hinderer 1995 atmospheric loading + FG5 K-scale registry + joint LS 12-constituent tidal subtraction) + Y_4^2 proper associated Legendre basis (v3.1 fix)","expected_direction":"positive","feature_set":["l4_v31_epsilon_4","l4_v31_bootstrap_lower_2sigma","l4_v31_bootstrap_upper_2sigma","l4_v31_n_stations_contributing","l4_v31_cos2_normalized_spread_pct"],"git_commit_at_registration":"45e5ad7b9476ffcaf80327bf9d31eaa16fd7c253","hypothesis_id":"20260520_wahr_1981_l_4_admixture_v3_1_fir_7b08afb9","locked_at":"2026-05-20T01:56:36Z","locked_file_sha256":"56534ab7844b4b5271173b7745c9e120fc5b68736aa040b2ebdf1908d835b4d9","notes":"D.1-SPRINT v3.1 forward test for first-ever planetary l=4 admixture measurement.  re-eval YYYY-MM-DD = 2027-05-19.  Two outcomes:  (1) Y_4^2 basis + 24-station IGETS + D.1-SPRINT corrections produces fit consistent with Wahr 1981 prediction at 2sigma -- first-ever planetary l=4 measurement paper (JGR Solid Earth / GJI); (2) Fit inconsistent or systematic-dominated -- methods-clean bound + roadmap to v4 (ECMWF pressure scraper + ICET L2 + more stations).  Pre-eval 2026-11-19; full re-eval 2027-05-19.  This is the FIRST PRE-REGISTERED l=4 admixture forward test in the literature.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"epsilon_4 fitted within 2sigma of Wahr 1981 prediction"},"registered_at":"2026-05-20T01:56:25Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"back_compat_legacy_basis":"P_4_scalar (v2/v3.0; physically wrong for m=2 tide)","basis_function":"Y_4^2_associated_legendre","bootstrap_method":"station-resampling with replacement","bootstrap_n_iter":1000,"integration_years":5,"predecessor":"Phase D.1-A6 v3 (commits e8737c8f + 99e8ca3c) -- SYSTEMATIC_DOMINATED with scalar P_4 basis; epsilon_4 fit -87 percent in left tail of bootstrap [-1465 percent, +1550 percent]","stations_expanded_target":["strasbourg","wettzell","moxa","conrad","cantley","metsahovi","kamioka","canberra","sutherland","onsala","pecny","walferdange","larzac","schiltach","trappes","membach","bad_homburg","medicina","concepcion","hsinchu","mizusawa","lhasa","djougou"],"v3_synthetic_validation":"12-station noise-free synthetic at true epsilon_4=0.00700 -- Y_4^2 recovers exactly; P_4 returns -0.11061 (16x off + wrong sign)","wahr_1981_predicted_range":[0.005,0.01]},"method":"The fitted epsilon_4 (l=4 admixture ratio in M2 latitude shape) lies within the 2sigma bootstrap confidence interval of the Wahr 1981 + PREM prediction range epsilon_4 ~ 0.005-0.010 (0.5-1.0 percent admixture)","n_permutations":1000},"threshold":{"direction":"positive","metric":"epsilon_4_within_2sigma_of_wahr_1981_range","value":0.0},"title":"Wahr 1981 l=4 admixture v3.1 -- first-ever measurement with proper Y_4^2 basis + 5y IGETS + D.1-SPRINT corrections","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-11-19"},{"data_source":"nasa_omni_1h.db + sse_catalog_expanded.csv (future appended Cascadia events) + scripts/sse_piezo_em_stacked_epoch_v3.py","expected_direction":"positive","feature_set":["ae_index_1h"],"git_commit_at_registration":"a19d249f8acbbe45ac3a4ed55504cd063890d8f5","hypothesis_id":"20260522_cascadia_sse_ae_index_1h_precurs_79517a3b","locked_at":"2026-05-22T18:11:55Z","locked_file_sha256":"b9fdf7a7d7e696e3b3e72af724a7183a033a350b0aeaf26b9a964c44169843eb","notes":"FORWARD replication test of the May 22 2026 v3 hypothesis-generating finding (parent hypothesis 20260522_sse_piezo_electromagnetic_precur_682c9ac6 recorded as DISCONFIRMING at the global level).\n\nPARENT v3 result on Cascadia regional split (45 documented SSEs 2005-2018 from Wech-Bartlow PNSN catalog, post per-event-bin SEM correction, post F10.7 quartile-matched control, post Kp >= 5 storm exclusion within +/- 5d):\n  ae_index_1h at Cascadia: +6.51 sigma at lag -105h (Bonferroni FWER 0.001, N=37 events used)\n\nH1 (this forward test): The same effect persists at +z >= 3 (looser than +6.51 to allow per-event variance) on Cascadia SSEs that OCCUR between 2026-05-22 and 2028-05-22 (forward window). Cascadia ETS rate is approximately 2-3 events / year per PNSN tremor catalog; over 24 months we expect 4-7 new events, each independently testable against the same methodology + locked NASA OMNI ae_index_1h archive.\n\nDecision rule: re-run scripts/sse_piezo_em_stacked_epoch_v3.py with --per-region-split --exclude-kp-storms on the post-2026-05-22 Cascadia SSE subset using the same lag bins (8 bins across 336h pre-window). PASS if Cascadia ae_index_1h shows stacked z >= +3 (one-tailed positive direction) at any pre-event lag bin within +/- 42h of the parent finding's -105h lag.\n\nCAVEATS preserved per Nulls Paper Section 7:\n1. v3 was HYPOTHESIS-GENERATING. Parent v2 was the pre-registered NULL. This forward test pre-registers the v3 finding as a confirmatory candidate.\n2. The piezoelectric coupling MECHANISM is one interpretation; reverse causality (magnetospheric activity triggers SSEs via Coulomb stress modulation) is also plausible and not separable in this test.\n3. Effect size in forward test may be smaller than parent +6.51 due to regression-to-mean.\n4. If only 1-2 new Cascadia SSEs occur in the 24mo window, statistical power is insufficient -- in that case the hypothesis is WITHDRAWN as underpowered rather than declared null or confirmed.\n\nOUTCOME INTERPRETATION:\n  PASS: Cascadia ae_index_1h precursor confirmed -- candidate for GRL/Nature submission with prospective + retrospective combined analysis.\n  FAIL (z < 3): Parent +6.51 likely was post-hoc selection from 4 regions x 5 channels x 8 lag bins = 160 implicit tests after Bonferroni-40 protection; the finding does not replicate prospectively.\n  UNDERPOWERED: <3 new Cascadia SSEs in the 24mo window; withdraw + re-register with longer window.\n\nPUBLICATION COMMITMENT (Nulls Paper Section 7): result will be recorded in the registry within 30 days of re-eval date 2028-05-22, regardless of outcome.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2028-05-22","registered_at":"2026-05-22T18:11:44Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"stacked_epoch_z_with_solar_cycle_and_kp_storm_control","n_permutations":10000},"threshold":{"direction":"positive","metric":"z_score_post_F10p7_control_and_post_Kp5_storm_filter","value":3.0},"title":"Cascadia SSE ae_index_1h precursor forward replication: +z >= 3 at lag -105h on new Cascadia ETS 2026-05-22 to 2028-05-22","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2028-05-22"},{"data_source":"Phase 5 v1.2 bg_filter (running daily on AX41) + 18 of 27 VQ instruments + IGETS SG + INTERMAGNET + IRIS FDSN + FIRMS IR","expected_direction":"negative","feature_set":["gravimeter_step","magnetometer_pulse","accelerometer_transient","seismic_blast","atmospheric_ir_pulse","5domain_coincidence_count"],"git_commit_at_registration":"5735e7cd9aee2e90eadacda1c5ce24b40ba892dd","hypothesis_id":"20260526_mars_a10_asteroid_mass_pbh_10_17_0044dfc0","locked_at":"2026-05-26T04:25:23Z","locked_file_sha256":"6bb71716ec99ae980204bd007fbca7094d6d10338594629a84c11f8a9419f8e3","notes":"Asteroid-mass PBH transit search (E-1 from Apr 23 2026 brainstorm, partially active per memory/pbh_moonshot.md). 5-domain coincidence filter leveraging 18 of 27 VQ instruments -- nobody else can run this search. Apr 25 v1.2 produced upper bound <=8.4/yr; v1.3 audit Apr 28 showed null degenerate at v1.3 settings (88pct snapshots trigger stage-1 coincidence). A10 formalizes the 540-day forward search with pre-registered SNR>=5 threshold per domain and >=3 of 5 domain agreement. Honest prior: most likely outcome is null with tightened upper bound (publishable). Detection at any rate would be Nobel-tier. Direction negative means we are testing for absence of signal (rate UPPER bound below 10/yr); a confirmed positive detection of any signal at SNR>=5 across >=3 domains would be a CONFIRMED outcome despite the negative-direction encoding.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter"},"reevaluate_at":"2027-11-17","registered_at":"2026-05-26T04:24:38Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"5domain_coincidence_filter_with_poisson_null_upper_bound","n_permutations":null},"threshold":{"direction":"negative","metric":"annual_rate_upper_bound_at_3of5_snr5","value":10.0},"title":"MARS-A10: Asteroid-mass PBH (10^17 to 10^23 g) transits through Earth produce a 5-domain coincidence signature (gravimeter step + magnetometer pulse + accelerometer transient + seismic blast + atmospheric IR pulse) at synchronized timestamps across >= 3 of 5 domain detectors at SNR >= 5 each. Detection or upper-bound rate < 10 per year (matches Apr 25 v1.2 bounds) over 540-day forward window.","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"GCN Kafka gcn.notices.superk.sn_alert + USGS M3+ FDSN feed + IGETS live + Phase 65p neutrino_eq_precursor","expected_direction":"negative","feature_set":["superk_sn_neutrino_burst_alert_timestamp","iris_fdsn_m3plus_within_48h_1000km","igets_sg_residual_at_burst_epoch","2channel_correlation_p_value"],"git_commit_at_registration":"5735e7cd9aee2e90eadacda1c5ce24b40ba892dd","hypothesis_id":"20260526_mars_a11_if_a_galactic_supernova_7a312ff9","locked_at":"2026-05-26T04:25:23Z","locked_file_sha256":"95b455b01775e1f39ef10412b28dbc1e82195849e3ab197fbdcd5af1189eb2c5","notes":"Forward-going 730-day window neutrino-supernova-EQ correlation test. Extension of Fischbach 1986 disconfirmation (Phase 65p Apr 19 2026 null at p=0.30 over 14 M7+ events on real IceCube Gold/Bronze archive) to the supernova-burst regime where neutrino flux is 10^58 vs IceCube standard astrophysical flux. Galactic SN base rate ~1-3pct/yr per Adams 2013 means most likely outcome is window with zero SN events -- still publishable as a methods-clean upper bound. GCN Kafka gcn.notices.superk.sn_alert already subscribed since Phase 65u-RIGOR May 4 2026. cross_physics/neutrino_eq_precursor_monitor.py already shipped Phase 65p with interpretation_guide warning against single-detector p-hacking -- A11 mandates >=2 channels at p<0.001 to avoid this trap. Direction negative because forward-test outcome is most likely null.","ots_proof_path":null,"outcome":{"lookahead_hours":48.0,"min_magnitude":3.0,"type":"neutrino_supernova_earthquake"},"reevaluate_at":"2028-05-25","registered_at":"2026-05-26T04:25:01Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"neutrino_triggered_correlation_2channel","n_permutations":null},"threshold":{"direction":"negative","metric":"p_value_2channel_agreement","value":0.001},"title":"MARS-A11: If a galactic supernova within <=50 kpc occurs in the forward window, neutrino burst (Super-K SN-alert via GCN Kafka) triggers concurrent search at IRIS FDSN seismic stations + IGETS gravimeter for any M3+ earthquake within +-48h x +-1000km. Detection if any M3+ EQ correlates at p<0.001 across both physics channels; null otherwise. Supernova has approximately 1-3pct probability per year per Adams 2013; null result publishable as no evidence for SN-neutrino-EQ correlation if no SN occurs OR if SN occurs but no triggered EQs.","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"data/dm_realdata_v2.db + data/private/mars_sci/a1/cache/band_power.db (IRIS MUSTANG noise-PDF + Open-Meteo archive + NOAA SWPC F10.7 + NDBC wave-height)","expected_direction":"positive","feature_set":["microseism_band_power_0p1_0p5hz","annual_fourier_amplitude_db","stacked_snr","phase_consistency_rayleigh_p"],"git_commit_at_registration":"99e8dbab6d82d002c6299b3d57f065b70b035c1b","hypothesis_id":"20260526_mars_a1_annual_modulation_residu_b79122e6","locked_at":"2026-05-26T04:00:07Z","locked_file_sha256":"6a00e78c0e5171266830d0a3c2cba5bbb0c6c605c88cbe58b9c186cd06611e96","notes":"Mars-shot codename MARS-A1 (DMAI-SCI-001 v1.0). Builds on Phase 65g Dark Matter v2 pipeline (scripts/dm_real_data_pipeline.py) which runs daily 04:00 UTC on AX41 with 105 GSN stations + 822 days cached PSD + 15 quiet Antarctic/island stations stacking. Honest prior: previous DM v1 ran p=0.613 on 0.5-2.0 Hz band (null, microseism-dominated). v2 switched to 0.1-0.5 Hz quiet band per Phase 65g Apr 16 memo, accumulation phase. MARS-A1 adds NEW analysis layer extracting seasonal-residual signature specifically: annual Fourier mode at 1/365.25 day^-1 frequency after OLS subtraction of synthetic atmospheric (Open-Meteo barometric pressure) + oceanic (NDBC wave height) + solar-cycle (F10.7 cm radio flux) model with Hinderer 1995 atmospheric loading correction. Per-station bootstrap null via 1000 random-phase shuffles. Inverse-variance-weighted stack across 15 stations. Detection requires stacked SNR > 3 AND phase consistency on >=8 of 15 stations (Rayleigh p < 0.05). Null result: methods-clean upper bound on sub-eV bosonic DM coupling cross-section, comparable to Aurel et al. 2025 PRD framework. 18-month accumulation is BINDING WINDOW per Mars-shot brief; quarterly intermediate checks (2026-08-26, 2026-11-26, 2027-02-26, 2027-05-26, 2027-08-26) are trajectory-tracking only, NOT go/no-go. Re-eval 2027-11-17. AX41 daily cron at 04:45 UTC. References: Aurel et al. 2025 PRD on sub-eV bosonic DM seismic coupling; Hinderer 1995 atmospheric loading correction; Phase 65g DM v2 memo (memory/phase65g_pipeline_fixes.md).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_seasonal_modulation"},"reevaluate_at":"2027-11-17","registered_at":"2026-05-26T03:59:51Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"accumulation_window_days":540,"band_hi_hz":0.5,"band_lo_hz":0.1,"covariates":["open_meteo_surface_pressure","ndbc_wave_height","swpc_f107_daily"],"loading_correction":"hinderer_1995","null":"random_phase_shuffle","phase_consistency_min_stations":8,"quarterly_checks":["2026-08-26","2026-11-26","2027-02-26","2027-05-26","2027-08-26"],"rayleigh_p_max":0.05,"stations":["IU.QSPA","IU.SBA","IU.CASY","AI.VNDA","AI.SNAA","IU.MIDW","IU.RPN","II.EFI","II.HOPE","II.PMSA","II.CRZF","IU.NWAO","II.COCO","II.DGAR","G.PAF"]},"method":"stacked_annual_fourier_with_atmospheric_oceanic_solar_residual","n_permutations":1000},"threshold":{"direction":"positive","metric":"stacked_snr","value":3.0},"title":"MARS-A1: Annual modulation residual in microseism band power 0.1-0.5 Hz at 15 quiet Antarctic/island stations after atmospheric+oceanic+solar-cycle model subtraction over 18-month accumulation window","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-11-17"},{"data_source":"data/igets/ (IGETS SG archive 2009-2026, 24 stations) + Bos-Scherneck 2013 + IERS Conv 2010 atmospheric/oceanic corrections + Hinderer 1995 v1 empirical beta atmospheric loading","expected_direction":"positive","feature_set":["igets_daily_mean_ugal","secular_trend_slope_ugal_per_yr","atm_corr_residual","ocean_tide_residual","plume_proximity_km","reference_station_baseline"],"git_commit_at_registration":"c6d1099709c51de30bb87fd642ba58767ab3a07c","hypothesis_id":"20260526_mars_a3_multi_year_secular_gravi_382dfdca","locked_at":"2026-05-26T04:11:50Z","locked_file_sha256":"4064b13fe5ff83cbf602a0f60d427a0195040507d126c7b0f43d13abce71e441","notes":"Mars-shot codename MARS-A3 of the Six Science Mars-Shots phased program (DMAI-SCI-001 v1.0). Tests whether mantle plume / hot-spot proximity (10 known plume sites: Hawaii, Iceland, Galapagos, Yellowstone, Afar, Bouvet, Canary Islands, Reunion, Tahiti, East African Rift) imprints a detectable secular gravity drift signature at IGETS SG stations relative to 6 tectonically quiet reference stations (Pecny CZ, Walferdange LU, Membach BE, Onsala SE, Trappes FR, Schiltach DE). Path B v0.1 uses current Bos-Scherneck 2013 + IERS Conv 2010 midpoint corrections + Hinderer 1995 v1 empirical beta -- ships NOW on 17-year archive. Path A v0.2 sharpens with OSO Chalmers FES2014b BLQ files upon arrival (Phase D.1-SPRINT v4 shares same engineering deps). Detection: aggregate ratio >= 3.0 with bootstrap p<0.01 (1000-perm null via random station-label shuffle). Bonferroni: 10 plume x 6 reference = 60 pairs (corrected p<0.001 per-pair). Methods-clean null still publishable as constraint on mantle-plume gravity signature at IGETS sensitivity. Honest expectation: many plume sites lack close IGETS SG coverage (most IGETS infrastructure is Europe + East Asia + Australia); only 4-6 of 10 plume locations are likely to match an SG within 500 km. Real coverage gaps explicitly preserved as data_real=0 in result -- NOT fabricated. Forward test 24mo window for trend persistence + Path A re-run with FES2014b BLQ on OSO arrival. Re-eval 2028-05-26.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"mantle_convection_signature"},"reevaluate_at":"2028-05-25","registered_at":"2026-05-26T04:11:35Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"archive_years":[2009,2026],"atmospheric_correction":"Bos-Scherneck 2013 + IERS 2010 + Hinderer 1995 v1 (Path B)","bonferroni_pairs":60,"bootstrap_n":1000,"detection_p_threshold":0.01,"null":"random_station_label_shuffle","plume_proximity_km":500,"plume_stations_targeted":["Hawaii","Iceland","Galapagos","Yellowstone","Afar","Bouvet","Canary_Islands","Reunion","Tahiti","East_African_Rift"],"reference_stations":["Pecny","Walferdange","Membach","Onsala","Trappes","Schiltach"],"v0p2_path_a_blq":"queued on OSO Chalmers FES2014b BLQ arrival (Phase D.1-SPRINT v4 shared dep)"},"method":"secular_trend_slope_ratio_with_bootstrap_label_shuffle","n_permutations":1000},"threshold":{"direction":"positive","metric":"plume_to_reference_slope_ratio","value":3.0},"title":"MARS-A3: Multi-year secular gravity drift at IGETS Superconducting Gravimeter stations within 500 km of known mantle plume / hot-spot locations exceeds tectonically quiet reference stations by ratio >= 3.0 on 2009-2026 cohort","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2028-05-26"},{"data_source":"IGETS SG archive (24 stations Phase D.0 expanded catalog) + ICET solid-Earth tide + Open-Meteo barometric pressure (Hinderer 1995) + Bos-Scherneck 2013 / IERS 2010 ocean loading (v0.1 Path B; sharpens to Path A FES2014b BLQ on OSO Chalmers arrival)","expected_direction":"positive","feature_set":["psd_0p1_to_1000_nhz","stacked_snr_per_band","peak_frequency_nhz","peak_snr","n_stations_above_threshold"],"git_commit_at_registration":"5a1c06946c5d1bad3329b1dd05f84c990fb4d6e9","hypothesis_id":"20260526_mars_a4_solar_helio_tomography_v_18290950","locked_at":"2026-05-26T04:14:28Z","locked_file_sha256":"a5aad0498a0424ae54d3d11a10ddfceec9ea70985ea5a4f4611662a0b0e51c69","notes":"Mars-shot codename MARS-A4 (DMAI-SCI-001 v1.0). MOST AMBITIOUS of the 6 shots - planet-scale passive solar observatory via 24-station IGETS Superconducting Gravimeter network. Sub-microhertz signal detection requires sqrt(t) SNR growth from coherent stacking; 36-month integration gives 6x SNR boost vs 1 month. Three target frequency bands: (1) 11-yr solar cycle harmonic ~3 nHz, (2) 27-day solar rotation ~430 nHz, (3) lower-order solar g-mode candidates 100-300 nHz. Detection at SNR>=3 across >=6 of 24 stations after Bonferroni correction over 72 tests (3 bands x 24 stations) at p<0.001 = first-ever direct gravitational detection of solar g-modes via global SG array. Historical context: Fossat 2017 GOLF/VIRGO claimed solar g-mode detection but disputed by independent re-analysis - extreme caution on any positive signal. Null = methods-clean upper bound on solar gravitational mode coupling at ~10^-11 to 10^-13 m/s^2 sensitivity depending on integration time. FIRE-AND-FORGET design: v0.1 baseline establishes measurement; v0.2/0.3/0.4 = monthly cron accumulation; quarterly snapshots track SNR growth. First meaningful science ~Q2 2027 (12+ months of integration). Re-eval 2029-05-27. Closes Phase MARS-SCI program with all 6 A-shots pre-registered + Bitcoin-anchored. References: Fossat 2017 A&A solar g-mode claim (disputed); Hinderer 1995 atmospheric loading; Bos-Scherneck 2013 ocean loading; ICET solid-Earth tide; Phase D.0 IGETS 25-station expansion; Phase D.1-SPRINT v4 calibration pipeline.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"solar_gravitational_coupling"},"reevaluate_at":"2029-05-25","registered_at":"2026-05-26T04:14:13Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"accumulation_window_days":1095,"band_hi_nhz":1000.0,"band_lo_nhz":0.1,"bonferroni_corrected_p":0.001,"bonferroni_tests":72,"corrections":["icet_solid_earth_tide","hinderer_1995_atmospheric_loading","ocean_loading_path_b_to_path_a"],"min_stations_above_threshold":6,"monthly_baseline_cron":"15 5 1 * *","null":"random_phase_shuffle","quarterly_snapshot_cron":"30 5 1 1,4,7,10 *","stations":["Strasbourg","Wettzell","Moxa","Conrad","Cantley","Metsahovi","Kamioka","Canberra","Sutherland","Concepcion","Djougou","Onsala","Pecny","Walferdange","Larzac","Schiltach","Trappes","Membach","Bad-Homburg","Medicina","Hsinchu","Mizusawa","Lhasa","Apache-Point"],"target_bands_nhz":{"solar_cycle_11yr":[2.85,3.2],"solar_g_mode_candidates":[100.0,300.0],"solar_rotation_27day":[425.0,435.0]}},"method":"coherent_stacked_psd_solar_band","n_permutations":500},"threshold":{"direction":"positive","metric":"stacked_snr","value":3.0},"title":"MARS-A4: Solar helio-tomography via global IGETS SG array - 24-station coherent stacked PSD in 0.1-1000 nHz band over 36-month forward window 2026-05-27 to 2029-05-27 detects spectral peaks at solar gravitational modulation frequencies (11-yr cycle ~3 nHz, 27-day solar rotation ~430 nHz, solar g-mode candidates 100-300 nHz) with SNR >=3 across >=6 stations after Hinderer 1995 atmospheric loading + Path A FES2014b BLQ + ICET solid-Earth tide corrections","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2029-05-27"},{"data_source":"IRIS FDSN BHZ broadband + earth_state_tensor.db + retro_tensor.db","expected_direction":"positive","feature_set":["tensor_microseism_residual","bhz_waveform_0p1_1hz_band_power","prerupture_residual_z_lag_minus24h_to_minus1h"],"git_commit_at_registration":"4ca4ac02fca8a6917bc5452a01804e141d0824bf","hypothesis_id":"20260526_mars_a5_at_m9_historical_earthqu_ffd40f79","locked_at":"2026-05-26T04:25:23Z","locked_file_sha256":"9b8ef0eeb16bd97de9990acb03113193cd3baef8eb979301d6f7075cee441c3c","notes":"Extension of Phase 65u-MICROSEISM-LIVE / Phase 5 v1 (test R^2=0.052 on microseism summary band power). A5 extends to full BHZ waveform 0.1-1 Hz at M9+ rupture-class events. Honest prior: Phase 5 found microseism mostly orthogonal to other 111 retro tensor features; M9+ stratum is the cleanest test of whether full-waveform precursor exists at all. Most likely outcome: publishable null. Six target events: Aceh 2004-12-26 M9.1, Maule 2010-02-27 M8.8, Tohoku 2011-03-11 M9.1, Sumatra 2012-04-11 M8.6, Iquique 2014-04-01 M8.2, Illapel 2015-09-16 M8.3.","ots_proof_path":null,"outcome":{"lookahead_hours":24.0,"min_magnitude":9.0,"type":"earthquake"},"reevaluate_at":"2026-11-22","registered_at":"2026-05-26T04:23:29Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"tensor_ridge_residual_stacked_epoch","n_permutations":1000},"threshold":{"direction":"positive","metric":"z_score_prerupture_residual","value":3.0},"title":"MARS-A5: At M9+ historical earthquakes 1900-2026 (Aceh 2004, Maule 2010, Tohoku 2011, Sumatra 2012, Iquique 2014, Illapel 2015), tensor-conditioned waveform residual (predicted minus observed) at the nearest IRIS FDSN broadband station shows pre-rupture signature in the 0.1-1 Hz band with z-score >= +3 sigma vs background at lag -24h to -1h after fitting a tensor to waveform ridge regression. Phase 65u-MICROSEISM-LIVE Phase 5 v1 already established 0.5 sigma class residual prediction; A5 extends to M9+ stratum where signal-to-noise is maximal.","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"data/retro_cache/nmdb_cosmic_ray.db + data/usgs_global_events.db + data/solar_cycle_archive.db","expected_direction":"negative","feature_set":["cosmic_ray_oulu_counts","cosmic_ray_apty_counts","cosmic_ray_jung_counts","cosmic_ray_rome_counts","cosmic_ray_bksn_counts","cosmic_ray_athn_counts","cosmic_ray_aatb_counts","cosmic_ray_rate_anomaly","f107_daily"],"git_commit_at_registration":"99e8dbab6d82d002c6299b3d57f065b70b035c1b","hypothesis_id":"20260526_mars_a6_cosmic_ray_flux_pre_even_adbcb2fb","locked_at":"2026-05-26T03:56:23Z","locked_file_sha256":"40e76486aec6d2d4d47bd7fc4325c39fc6875e3a08801fed776fcd97b7cfc107","notes":"Mars-shot codename MARS-A6. Extends Phase 5 microseism universal precursor methodology to cosmic-ray channel. Phase 5 found cosmic-ray-EQ null at p>=0.6 on M5+ events; A6 tests M6+/M7+ where signal cleanliness improves. NMDB 7-station network: OULU/APTY/JUNG/ROME/BKSN/ATHN/AATB. Test computes pre-event z-score asymmetry (T-24h to T0 vs T0 to T+24h) per event, controls for F10.7 solar cycle covariate. Bootstrap null via 1000 random-phase shuffles. Bonferroni corrected for 2 strata (M6+ ~1400 events, M7+ ~150 events). Honest expectation: publishable null update to 40-year Fischbach 1986 cosmic-ray-EQ conjecture, OR confirmation that survives at higher magnitude. Re-eval 2026-11-22 (180d forward window, analysis-only -- archive complete).","ots_proof_path":null,"outcome":{"lookahead_hours":48.0,"min_magnitude":6.0,"type":"earthquake"},"reevaluate_at":"2026-11-22","registered_at":"2026-05-26T03:56:18Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"archive_years":[2010,2024],"asymmetry_ratio_threshold":3.0,"bonferroni":2,"covariate":"f107_daily","null":"random_phase_shuffle","strata":["M6+","M7+"],"window_hours":48},"method":"stacked_epoch_z_asymmetry_with_f107_control","n_permutations":1000},"threshold":{"direction":"negative","metric":"pre_post_asymmetry_ratio","value":3.0},"title":"MARS-A6: Cosmic-ray flux pre-event asymmetry at NMDB 7-station network within +/-48h of M6+ and M7+ EQ events on 14-year archive 2010-2024","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-11-22"},{"data_source":"data/retro_cache/gvp_eruptions.db + IRIS FDSN + IGETS SG + Open-Meteo + USGS Volcano Hazards","expected_direction":"positive","feature_set":["microseism_secondary_band_3s","igets_sg_gravity_residual_ugal","infrasound_band_power_db","atmospheric_pressure_z_7d","gvp_vei_event_index","sleeping_giant_unrest_score","tensor_eigenmode_drift","volcanic_tremor_band_power"],"git_commit_at_registration":"c6d1099709c51de30bb87fd642ba58767ab3a07c","hypothesis_id":"20260526_mars_a7_multi_physics_tensor_pre_2091f22a","locked_at":"2026-05-26T04:11:44Z","locked_file_sha256":"d0e5cb70e82d66ac6e6c314ab9ebf414f9c815771901585f7cd0c0b5ee5215dc","notes":"Mars-shot codename MARS-A7. Phase 2 of MARS-SCI program, paired with MARS-A3. Extends Phase 5 microseism universal precursor methodology to decade-scale supervolcano context with 4-channel co-detection requirement. Sleeping-giants ranking from Phase 5c v1.1 (Long Valley/Yellowstone/Tambora/Toba/Hualalai/Bardarbunga/Campi_Flegrei/Aira/Iwo_Jima/Soda_Lakes_NV). GVP catalog 1900-2026 expected to contain 30-50 VEI>=4 events at or within 50 km of the 10 watch-list sites. Stacked-epoch across all events at lags T-5y/T-2y/T-1y/T-6mo/T-3mo/T-1mo/T0. Z-score per channel vs background (5y pre-event minus 10y background). Detection if z>=+3sigma at >=2 of 4 channels with Bonferroni p<0.01 (40 tests = 10 events x 4 channels at peak lag). Honest expectation: with only 30-50 events across 10 different volcanoes, individual-site statistical power is limited but cross-site stacked-epoch should reveal common precursor signatures if they exist. Most likely outcome: publishable methods-clean null on decade-scale supervolcano precursors, OR a stacked-epoch detection that prior literature could not see without our multi-physics tensor framework. Forward test 1095-day window for any new VEI>=4 event at the 10 watch-list sites should re-trigger the same channels at the same lag.","ots_proof_path":null,"outcome":{"lookahead_hours":43800.0,"min_magnitude":4.0,"type":"volcano"},"reevaluate_at":"2029-05-25","registered_at":"2026-05-26T04:11:29Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"archive_years":[1900,2026],"bonferroni":40,"channels":["microseism","gravity","infrasound","atmospheric"],"lags":["T-5y","T-2y","T-1y","T-6mo","T-3mo","T-1mo","T0"],"min_channels_required":2,"vei_min":4,"watch_list":["long_valley_ca","yellowstone_wy","tambora","toba","hualalai","bardarbunga","campi_flegrei","aira","iwo_jima","soda_lakes_nv"],"window_years":5,"z_threshold":3.0},"method":"stacked_epoch_multichannel_z_score_5y_window","n_permutations":1000},"threshold":{"direction":"positive","metric":"n_channels_passing_z_threshold","value":2.0},"title":"MARS-A7: Multi-physics tensor precursor signatures within 5-year pre-event window of historical VEI>=4 eruptions at 10 sleeping-giant volcanoes (Long Valley CA, Yellowstone WY, Tambora, Toba, Hualalai, Bardarbunga, Campi Flegrei, Aira, Iwo Jima, Soda Lakes NV) exhibit z>=+3sigma on >=2 of 4 channels (microseism, gravity, infrasound, atmospheric) vs background tensor state on GVP catalog 1900-2026","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2031-05-25"},{"data_source":"IGETS SG archive (25 stations Phase D.0 expansion) + bundle-36 PCA replication","expected_direction":"positive","feature_set":["0s2_multiplet_asymmetry_microhertz","inner_core_daily_rotation_rate_deg_per_yr","lomb_scargle_peak_power_05_to_3yr_band"],"git_commit_at_registration":"4ca4ac02fca8a6917bc5452a01804e141d0824bf","hypothesis_id":"20260526_mars_a8_inner_core_rotation_rate_d5223b4b","locked_at":"2026-05-26T04:25:23Z","locked_file_sha256":"83df905d6fea402c60abb0392fef4df93218051d6fb8d4dfbb92aff4bc410b85","notes":"Independent reproduction of Yang and Song 2023 Nature 622:734 inner-core rotation reversal. Train on 2009-2016 (matches Phase D.1-A6-CLOSE Track C v3 setup), validate held-out on 2017-2026. Uses EOST-only IGETS L2 with M2/S2/N2 calibration corrections per Phase 65w-VERIFY-CLOSE Bug 2 fix (May 18 2026). 25-station catalog from Phase D.0 expansion (Onsala, Pecny, Walferdange, Membach, Schiltach, Bad-Homburg, Larzac, Trappes, Medicina, Apache-Point, Concepcion, Sutherland, Hsinchu, Mizusawa, Djougou + 10 prior). Phase 65w-VERIFY-CLOSE / Phase D.1-A6-CLOSE Track C v2 daily 2009-2026 is the primary input pipeline. Honest prior: Phase 65w found 6.20h candidate line not Slichter (M4 ocean-tidal-loading contamination); Yang-Song reproduction was paused for Track C v2 reader fix. v0.1 will run on EOST-only deduped Track C v3 output.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geophysics_inner_core"},"reevaluate_at":"2029-05-25","registered_at":"2026-05-26T04:23:54Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"lomb_scargle_with_perm_null_held_out_validation","n_permutations":1000},"threshold":{"direction":"positive","metric":"z_score_peak_power","value":3.0},"title":"MARS-A8: Inner-core rotation rate timeseries 2009-2026 derived from coherent-stack 0S2 multiplet asymmetry at 25 IGETS SG stations (Phase D.0 expansion) shows sub-decadal 0.5-3yr oscillation with Lomb-Scargle peak power exceeding 1000-perm null shuffle at z >= 3 in the bundle-36-replication held-out subset 2017-2026 (post fit on 2009-2016). Tests Yang and Song 2023 Nature finding that inner-core rotation reversed in 2009-2010. Reproduction with independent network -- we use EOST-only L2 IGETS; they used Alaska earthquake waveforms.","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"USGS M8+ feed + GCN Kafka igwn.gwalert + IGETS SG live + Phase 65p tensor","expected_direction":"positive","feature_set":["igets_prompt_gravity_residual_within_1000km","lvk_noise_residual_at_rupture_origin_plusminus_100s","gw_newtonian_precursor_tensor_channel"],"git_commit_at_registration":"4ca4ac02fca8a6917bc5452a01804e141d0824bf","hypothesis_id":"20260526_mars_a9_for_any_m9_earthquake_oc_e2c835b6","locked_at":"2026-05-26T04:25:23Z","locked_file_sha256":"a6d457a441b9bc740c97e2f143b41093ddca3992b510f4ac6d0cc79bd4e5d079","notes":"Forward-going 365-day window prompt-gravity coincidence test at M9+ rupture class. Extends Phase 65p gw_newtonian_precursor (Vallee 2017 / Juhel 2018) to NEW LVK Run 5 era (begins late 2026). Honest prior: M9+ events have ~10pct/yr base rate so window may yield zero detection (publishable null). Existing modules: cross_physics/gw_newtonian_precursor_monitor.py (Phase 65p, 10 features), igets SG live ingest (Phase 65p-DEPLOY), GCN Kafka igwn.gwalert subscribed since Phase 65u-RIGOR May 4 2026. ARTIFACT outcome possible if single-channel signal at SNR>=3 without 2-of-3 agreement -- those go to followup not publication.","ots_proof_path":null,"outcome":{"lookahead_hours":0.05,"min_magnitude":9.0,"type":"earthquake"},"reevaluate_at":"2027-05-26","registered_at":"2026-05-26T04:24:15Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"multi_channel_coincidence_snr_threshold","n_permutations":null},"threshold":{"direction":"positive","metric":"min_channel_snr_with_2of3_agreement","value":3.0},"title":"MARS-A9: For any M9+ earthquake occurring 2026-05-26 through 2027-05-26, prompt gravity signature (Vallee 2017 + Juhel 2018) is detected at SNR >= 3 across >= 2 of (a) IGETS SG within 1000km of epicenter, (b) co-incident LVK gravitational-wave detector noise residual within +-100s of rupture origin, (c) Phase 65p gw_newtonian_precursor tensor channel. NULL if no M9+ events occur in window; CONFIRMED if M9+ event detected with multi-channel agreement; ARTIFACT if single-channel signal only.","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"Convergence Gate v7 (88 channels), Phase 5 microseism universal precursor (Phase 5 May 2026 +19sigma finding), ETAS aftershock forecaster (Reasenberg-Jones 1989), USGS FDSN local catalog, BSA v1.2 cyclic_safety_score static prior","expected_direction":"positive","feature_set":["composite_score_0_1","recommendation_class","static_prior_css","convergence_gate_score","microseism_precursor_score","eq_activity_score","etas_forward_score","detection_rate","false_positive_rate"],"git_commit_at_registration":"40aeee30c62e3dc13e63b2f8d14934ca4f563f50","hypothesis_id":"20260526_mars_bsa_cycling_real_time_bsa_c_5e9ffdd5","locked_at":"2026-05-26T04:54:36Z","locked_file_sha256":"4a401c9a906930e691c2227941e9e159aab5443f4707ebccee80f43e6274c491","notes":"v0.1 synthetic baseline: detection 70% (3 missed -- Aliso Canyon, Mont Belvieu salt cavern, Hutchinson Yaggy; all primarily geomechanical caprock failures with weak seismic precursors), FPR 0% on BSA v1.2 top-10 controls. v0.2 swaps synthetic snapshot for AX41 retro-tensor replay. v0.3 adds CAPROCK_FATIGUE_GEOMECHANICAL sub-axis to lift detection on engineering-failure cases. Engine module cross_physics/bsa_cycling_recommendation_engine.py. API at /api/v1/bsa/cycling-recommendation. AX41 cron 25 6 * * *. Design doc reports/brine_storage/bsa_cycling_engine_v0p1_design.md. Re-eval 2026-11-21 (180d). DeepMap AI Inc. BSA v1.2 Upgrade #2 of 7.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"operational_engine_calibration"},"reevaluate_at":"2026-11-22","registered_at":"2026-05-26T04:54:23Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"deterministic_backtest_10_incidents_10_controls + production_replay_30d_window","n_permutations":null},"threshold":{"direction":"positive","metric":"detection_rate_and_FPR","value":0.8},"title":"MARS-BSA-CYCLING: Real-time BSA cycling-safety recommendation engine combining Phase 5 universal precursor + Convergence Gate v4/v7 + ETAS Forecaster + USGS local EQ activity + BSA v1.2 cyclic_safety_score (static prior) produces a forward 180-day false-positive rate <= 5% (PAUSE_24H or HARD_STOP_REVIEW issued when NO induced-seismicity event occurs within 30d of recommendation) on a synthetic backtest at 10 Aliso-Canyon-class historical incidents, with detection rate >= 80% on the real incidents.","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-11-21"},{"data_source":"Per-well wellhead surface gravimeter timeseries (CG-5 / FG5 / A10 / iGrav per site) + cycling pressure log + BSA v1.2 formation-type tag","expected_direction":"positive","feature_set":["stacked_snr_per_cycle","array_gain_sqrt_n_coherence","effective_array_floor_ugal","min_resolvable_density_change_pct","coherence_metric_0_1","snr_of_stacked_mean"],"git_commit_at_registration":"3af806b1de4d9dd6f5b9bfccfba2c60382e14f7a","hypothesis_id":"20260526_mars_bsa_distgrav_a_100_well_bsa_f7b6b038","locked_at":"2026-05-26T05:06:54Z","locked_file_sha256":"5d44545f9f9045af26fef4a70b95d7dd7063685966ffb165b55072cca444e08e","notes":"v0.1 first-run synthetic 4-well coherence demo recovered true_slope=0.10 uGal/MPa as stacked_mean=0.104 +- 0.0094 (SE), giving snr_of_stacked_mean=11.1 and coherence_metric=0.787 (STRONG). SNR-vs-N curve generated from N=1 to N=100, matching sqrt(N) law within 5% on synthetic data. Module cross_physics/bsa_distributed_gravimeter.py. Forward test: replay against real operational data when first BSA pilot reaches N>=4 cycling wells. Withdraw clause: if no BSA site reaches N=4 by 2027-05-25 declare UNDERPOWERED and re-register against next operational milestone. Design doc reports/brine_storage/bsa_distributed_gravimeter_v0p1_design.md. Re-eval 2027-05-25 (365d). DeepMap AI Inc. BSA v1.2 Upgrade #3 of 7.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geophysics_array_gain_calibration"},"reevaluate_at":"2027-05-26","registered_at":"2026-05-26T05:06:41Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"snr_vs_n_curve_replay_on_operational_data + theoretical_sqrt_n_comparison","n_permutations":null},"threshold":{"direction":"positive","metric":"empirical_snr_over_theoretical_snr_ratio","value":0.95},"title":"MARS-BSA-DISTGRAV: A 100-well BSA pressure-cycling array, with each well instrumented with a CG-5 (or better) wellhead gravimeter and a documented coherence efficiency epsilon in [0.5, 0.9] depending on formation type, achieves a coherent-stack SNR boost >= 0.95 * sqrt(N * epsilon) -- meaning the empirical SNR curve from N=1 to N=100 obeys the theoretical sqrt(N) array-gain law within 5% over a 365-day operational window. Treats BSA wells THEMSELVES as a distributed gravimeter array, filling the IGETS coverage gap (25 stations globally, ~0 of BSA pilot sites within 500 km).","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-05-25"},{"data_source":"IGETS Superconducting Gravimeter network 25 stations 2009-2026","expected_direction":"negative","feature_set":["m2_amplitude","s2_amplitude","n2_amplitude","k1_amplitude","o1_amplitude","fos_0s2","fos_1s0","fos_0t2"],"git_commit_at_registration":"3d9192321b967174e47f9814583eac0ae41f2bee","hypothesis_id":"20260526_mars_earth_cavendish_newton_g_fr_daa12be5","locked_at":"2026-05-26T05:09:34Z","locked_file_sha256":"d6aa1b03492d4eeee7ca5b1dce4fb5a048b92ce3a01a6be50d0242ed8a027f7e","notes":"MARS-EARTH-CAVENDISH: Using the 25-station IGETS Superconducting Gravimeter network's lunar tidal gravity response (M2 + S2 + N2 + K1 + O1 constituents) combined with free oscillation amplitudes (0S2 + 1S0 + 0T2 from Phase 65aa modules) over the 17-year archive 2009-2026, after Hinderer 1995 atmospheric loading + Path A FES2014b ocean tidal loading + ICET solid-Earth tide correction, measure Newton's G via joint inversion against PREM elastic-property uncertainty. Target: relative uncertainty <= 50 ppm -- a factor of 2.5x improvement over current best Cavendish-balance measurements. Detection if measured G falls within 4 sigma of one BIPM lab group AND outside 4 sigma of another, resolving the long-standing G-measurement discrepancy. Methods-clean null otherwise establishes that planetary-scale G measurements share similar systematics to lab-scale, providing upper bound on G uncertainty pathway. Re-eval 2028-05-25 (730d window).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"fundamental_constant"},"reevaluate_at":"2028-05-25","registered_at":"2026-05-26T05:09:08Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"joint_mcmc_inversion_G_love_numbers","n_permutations":null},"threshold":{"direction":"negative","metric":"relative_uncertainty_ppm","value":50.0},"title":"MARS-EARTH-CAVENDISH Newton G from 25-station IGETS lunar tidal response with free oscillation cross-validation","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2028-05-25"},{"data_source":"GEBCO bathymetry + NOAA WOA seafloor temperature + CTBTO IMS hydrophones + EMAG2 + GRACE-FO + IODP/JOIDES Resolution + national program literature","expected_direction":"positive","feature_set":["bathymetry_500_2000m","seafloor_temp_gradient","hydroacoustic_backscatter_proxy","emag2_magnetic_anomaly","grace_continental_margin_gravity","bsr_signature_proxy","sub_bottom_thickness","methane_volume_estimate","climate_risk_score","national_program_overlap"],"git_commit_at_registration":"3af806b1de4d9dd6f5b9bfccfba2c60382e14f7a","hypothesis_id":"20260526_mars_methane_hydrate_globally_id_16015a37","locked_at":"2026-05-26T05:07:54Z","locked_file_sha256":"39aa289f369d27a74dcecc15bccec2237412469ba23a0a5c7b13edc55a28132b","notes":"Mars-shot MARS-METHANE-HYDRATE.  Largest unexploited fossil reserve (1,500-3,000 Gt methane carbon) + largest climate-risk methane source.  Passive multi-physics scoring on continental margins 500-2000m depth.  Hypothesis split: at least 100 candidate provinces globally + at least 10 supporting either (a) pilot-scale extraction economics OR (b) climate-risk monitoring deployment.  Forward window 730 days.  Pass = both thresholds met.  Honest expectation: detection of catalog candidates is straightforward (literature already documents ~30 known + ~70 candidate provinces); the harder test is whether multi-physics scoring discriminates pilot-class from monitoring-class.  Re-eval 2028-05-25.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2028-05-25","registered_at":"2026-05-26T05:07:37Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"global_continental_margin_multi_physics_atlas","n_permutations":null},"threshold":{"direction":"positive","metric":"n_provinces_high_confidence","value":100.0},"title":"MARS-METHANE-HYDRATE: Globally identify >=100 high-confidence methane hydrate provinces on continental margins via passive multi-physics scoring (hydroacoustic backscatter as BSR proxy + sea floor temperature gradient + bathymetry stability zone + gravity drift signature + EMAG2 magnetic anomaly), at depths 500-2000m on passive + active continental margins. >=10 provinces should support either (a) pilot-scale extraction economics or (b) climate-risk monitoring deployment. Top provinces should overlap with national programs (Japan METI Nankai, US DOE GoM, India ONGC Krishna-Godavari, China CGS South China Sea, Korea KIGAM Ulleung, Norway Equinor Barents).","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2028-05-25"},{"data_source":"usgs_nure_gamma + usgs_mrds + geoscience_australia_gamma + nrcan_gamma + egdi_eu + cprm_brazil + gsi_india","expected_direction":"positive","feature_set":["k_abundance","u_abundance","th_abundance","th_k_ratio","th_u_ratio","mrds_proximity","geological_province","gamma_composite"],"git_commit_at_registration":"3af806b1de4d9dd6f5b9bfccfba2c60382e14f7a","hypothesis_id":"20260526_mars_ree_globally_identify_rare__2a906627","locked_at":"2026-05-26T05:09:00Z","locked_file_sha256":"7a3ebc4d1b5ba822f485e4bf7d2877b3d45323181b768d1748168b37fa60a6e0","notes":"MARS-REE: Globally identify >=500 candidate REE deposits via multi-physics scoring combining gamma spectroscopy (K/U/Th from NURE/cosmic-ray gamma) + USGS MRDS REE occurrence cross-correlation + geological province context + cyclic-pattern Th/K and Th/U ratio anomalies + per-deposit estimated REE concentration. Of these, >=50 sites should score Phase 65w-FS-class composite (>=9.0) and represent realizable commercial discovery candidates at $50-200M/yr extraction economics. Top-10 candidates outside China (US + Australia + Canada + Brazil + India + EU) should support US Critical Minerals Office pilot funding pursuits.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"mineral_discovery"},"reevaluate_at":"2028-05-25","registered_at":"2026-05-26T05:08:28Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"multi_physics_composite_score","n_permutations":null},"threshold":{"direction":"positive","metric":"n_box_elder_class_composite_ge_9","value":50.0},"title":"MARS-REE: Globally identify rare earth element deposits via gamma + MRDS multi-physics scoring","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"GRACE-FO L3 monthly mascons (JPL/CSR/GFZ public) 2009-2026 + IGETS SG archive (24 stations Phase D.0 expanded) + USGS NWIS groundwater levels (US only via NWIS waterservices API) + AQUASTAT national pumping estimates","expected_direction":"positive","feature_set":["grace_recharge_rate_mm_per_yr","igets_gravity_drift_ugal_per_yr","nwis_water_level_drift_m_per_yr","combined_snr","modality_agreement_count","sustainable_yield_m3_per_yr","extraction_cost_usd_per_m3"],"git_commit_at_registration":"407ad73319a5f147349558b3ec7ce1ae48c5e934","hypothesis_id":"20260526_mars_w1_global_1_1_aquifer_recha_c0232de8","locked_at":"2026-05-26T04:58:08Z","locked_file_sha256":"543290caa24e1b3b4376517106ce3100a26dbf3dab3a46d0bbad2518e17f8e40","notes":"Mars-shot codename MARS-W1 of the Mars-shots water-pillar program (CLAUDE.md Phase MARS-SCI). First global 1deg x 1deg aquifer recharge map at planetary scale, constructed from 17-yr GRACE-FO mascon archive + IGETS SG point-precise drift + USGS NWIS groundwater levels. 3-modality cross-validation: GRACE (primary, global coverage, ~150K km^2 resolution), IGETS (secondary, ~25 station global network for verification at km-scale), NWIS (tertiary, US-only validation). Outcome criterion: >=500 high-confidence sustainable-yield basins globally (combined SNR >= 3 across >= 2 of 3 modalities); >=50 at <=USD 0.15/m^3 extractive economics. Honest expectation: most 1deg x 1deg cells will have only GRACE coverage (~150K km^2 resolution natively, downsampled to 1deg = ~110km); NWIS limited to US (~2000-3000 wells in cache); IGETS sparse outside Europe/Japan (same coverage gap as MARS-A3 0/10 plume matches at 500km). Customer-tier: World Bank Water Resources Group, UN-Water, national water ministries, agricultural insurance carriers, desal-plant siting consultants. Forward-window 730d (2yr) for any commercial pilot validation cycle. v0.1 ships planetary atlas now; v0.2 sharpens with NASA GES DISC GLDAS comparison + FAO AQUASTAT cross-validation + global IGRAC well dataset for non-US verification. Methods-clean even if final basin count <500 (publishable as constraint on planet-scale sustainable-yield mapping at current sensor sensitivity). Re-eval 2028-05-25.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"recharge_atlas"},"reevaluate_at":"2028-05-25","registered_at":"2026-05-26T04:57:50Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{"archive_years":[2009,2026],"basin_target_count":500,"bootstrap_n":1000,"detection_p_threshold":0.01,"economics_target_count":50,"economics_threshold_usd_per_m3":0.15,"grace_proximity_cap_km":110,"igets_proximity_cap_km":500,"modality_agreement_threshold":2,"null":"random_grid_label_shuffle","nwis_proximity_cap_km":100,"resolution_degrees":1.0,"snr_threshold":3.0},"method":"cross_modality_snr_agreement_with_bootstrap_label_shuffle","n_permutations":1000},"threshold":{"direction":"positive","metric":"high_confidence_basin_count","value":500.0},"title":"MARS-W1: Global 1°×1° aquifer recharge rate atlas constructed from GRACE-FO monthly mascons + IGETS Superconducting Gravimeter point-gravity drift + USGS NWIS groundwater levels (where available) over the 17-year archive 2009-2026 identifies >=500 high-confidence sustainable-yield candidate basins globally (combined SNR >= 3 across >= 2 of 3 sensor modalities), of which >=50 should support extractive operations at <= 0.15 USD/m^3 economics. Verifies the GRACE-FO where is the water actually being recharged signal against IGETS point-precise gravity drift + NWIS measured levels.","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2028-05-25"},{"data_source":"global_pressurized_atlas_v0p1","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"1c0944c27e7a3423132e1a2e7fb34ac5920c21c1","hypothesis_id":"20260526_mars_w2_globally_pressurized_sub_0a1a0910","locked_at":"2026-05-26T04:45:47Z","locked_file_sha256":"be18fc0ac6e52f4b94bb3c1e03511cb1654cbfca9353e94c75d709af814cadf3","notes":"MARS-W2 (Phase MARS-SCI Water Pillar). Extends 373-site US-Western Master Atlas to global cohort across 4 continents. Pass = >=500 candidate sites surveyed AND >=80 score composite >= 9.0 (Box-Elder-class). Forward window 2 years for commercial pilot conversion. Pre-registered before any data collection on the global cohort. Companion to P069 (USPTO 64/050,287 RPW provisional). Re-eval 2028-05-25.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"global_atlas_expansion"},"reevaluate_at":"2028-05-25","registered_at":"2026-05-26T04:45:31Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"5_axis_p069_screening_with_cyclic_safety","n_permutations":null},"threshold":{"direction":"positive","metric":"n_sites_composite_ge_9p0","value":80.0},"title":"MARS-W2: Globally, pressurized subsurface formations identifiable via passive multi-physics scanning (5-axis P069 methodology + cyclic safety) exist at >=500 candidate sites distributed across South America (Andean basins), Africa (East African Rift + sub-Saharan), Asia (Tarim + Mekong + Indus), and Oceania (Murray-Darling + New Zealand); of these, >=80 should score Phase 65w-FS Box-Elder-class (composite >= 9.0)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2028-05-25"},{"data_source":"submarine_spring_atlas_v2","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"ceed8ee8a368e52e7e1ccc54097061a11981fcb0","hypothesis_id":"20260526_mars_w3_the_submarine_spring_atl_009fa178","locked_at":"2026-05-26T04:53:06Z","locked_file_sha256":"7f5676cd9fe5bcc807e691b69ec1497fe7cdc4841afcd5908fbf2fcf9112de58","notes":"MARS-W3 (Phase MARS-SCI Water Pillar shot 3). Extends Submarine Spring Atlas v1 (122 sites, Phase 65w-VERIFY-CLOSE / Phase 65x) globally to >=250 catalog sites adding Mediterranean basin expansion (Greece/Italy/Cyprus/Crete/Lebanon/Israel/Egypt/Turkey), South Pacific (Fiji/Tonga/Samoa/French Polynesia/Vanuatu/Solomon Islands), Indian Ocean (Maldives/Mauritius/Madagascar/Seychelles/Reunion/Mozambique), Caribbean expansion (Bahamas/Cuba/Dominican Republic/Puerto Rico/Jamaica/Cayman/Antigua), Atlantic islands (Canary/Madeira/Azores/Bermuda/Iceland), North America extension (FL Panhandle/Texas Gulf/PNW/Alaska SE), Mexico continuing (deeper Yucatan/Baja Sur/Gulf of California). Scores each site on v1 5-axis methodology (karst host, hydraulic head, thermal anomaly, salinity reduction, bathymetry geometry). High-confidence cluster = multi-physics agreement across >=3 of 5 axes. Each top-5 new cluster supports a pilot-ready commercial brief at $200K-$500K scale targeting coastal water utilities / national authorities. Companion to v1 atlas + P069/PWI cross-license vision. Re-eval 2028-05-25 (730d forward window).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"global_atlas_expansion"},"reevaluate_at":"2028-05-25","registered_at":"2026-05-26T04:52:23Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"5_axis_multi_physics_sgd_screening","n_permutations":null},"threshold":{"direction":"positive","metric":"n_new_high_confidence_clusters","value":10.0},"title":"MARS-W3: The Submarine Spring Atlas v2 expanded from 122 to >=250 catalog sites identifies >=10 new high-confidence (multi-physics confirmation across >=3 of 5 axes) submarine groundwater discharge clusters with realizable freshwater extraction economics ($0.020/m3 or better, beating reverse osmosis desal by >=30x). Top-5 new clusters should each support pilot proposals to coastal water utilities at $200K-$500K pilot scale.","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2028-05-25"},{"data_source":"Open-Meteo + Sentinel-3 SLSTR + USGS NWIS + IAEA + AQUASTAT + SMU heat flow + Submarine Spring Atlas v1 + NASA SHARAD","expected_direction":"positive","feature_set":["diurnal_temp_amp_C","aquifer_depth_m","geothermal_gradient_C_per_km","salinity_gradient","atm_pressure_variability_kPa","deployability_index","mars_isru_index"],"git_commit_at_registration":"df47ba02ee206fa31e89746b7017910f545e173d","hypothesis_id":"20260526_mars_w6_vacuum_anywhere_atlas_v0_40e9e700","locked_at":"2026-05-26T05:13:28Z","locked_file_sha256":"a7e31e4087aed71f01bb97ef852f812d7666aac1984aa8326bba1914575f78df","notes":"MARS-W6: Globally identify >=500 candidate sites where passive vacuum pumping (Heron of Alexandria extension to 350m depth via combined diurnal thermal cycling + atmospheric pressure differential + subsurface aquifer access) provides freshwater extraction at <=USD 0.10/m3, leveraging diurnal temperature amplitude >=20C + subsurface aquifer presence within 350m + geothermal or salinity gradient supporting passive cycling. Of these, >=50 sites should support pilot-scale demonstration at USD 50-150K per installation. Top-10 sites by aridity x deployability index represent immediate commercial water-access pilots (drought relief). Mars-ISRU dual-use claim: same physics enables future Mars subsurface ice extraction. Pass criterion locked: >=500 CANDIDATE cells globally AND >=50 PILOT-READY cells (composite >= 90) AND >=10 sites with extraction economics <= USD 0.10 per m3. Forward window 730 days for any commercial pilot conversion or NASA-ISRU technology-transfer engagement. Brand DeepMap AI Inc. only. References Heron of Alexandria circa 50 AD vacuum-prime extension extending pumping depth from Torricellian 10.3m to 350m via combined passive vacuum + thermal cycling; cross-references P069 USPTO 64/050/287 Replenishable Pressurized Water Program.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"water_resource_atlas"},"reevaluate_at":"2028-05-25","registered_at":"2026-05-26T05:12:51Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"global_5_axis_composite_screen","n_permutations":null},"threshold":{"direction":"positive","metric":"n_pilot_ready_sites_and_dollars_per_m3","value":500.0},"title":"MARS-W6 vacuum-anywhere atlas v0.1 freshwater extraction via passive vacuum pumping Heron of Alexandria extension to 350m depth","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"data/private/mars_sci/wild3/cache/band_power_50station.db (IRIS MUSTANG noise-PDF + IGETS Phase D.0 expansion + USGS continuous gravimeters + academic gravimeters + Open-Meteo barometric + NOAA SWPC F10.7 + Bos-Scherneck 2013 ocean tidal loading)","expected_direction":"positive","feature_set":["microseism_band_power_0p1_0p5hz","annual_fourier_amplitude_db","stacked_snr_50station","phase_consistency_rayleigh_p_50station","per_station_systematic_correction_residual"],"git_commit_at_registration":"67981e55c63157f3292040f289bcbeb2ac09227b","hypothesis_id":"20260526_mars_wild_3_50_station_global_co_61287beb","locked_at":"2026-05-26T04:54:01Z","locked_file_sha256":"279e82128e0067c5ea9e41b509661cc4894a8b1c4f2ff0bbd68389595dee6966","notes":"Mars-shot codename MARS-WILD-3 (DMAI-SCI-WILD-003 v1.0). Direct sister-shot extension of MARS-A1 (hypothesis 20260526_mars_a1_annual_modulation_residu_b79122e6, currently in 18-month accumulation at 15 quiet stations) to ~50 stations globally for sqrt(50)/sqrt(15) ~ 1.83x additional SNR ceiling boost, or theoretical 7x absolute SNR over single-station baseline. Cohort: 15 quiet Antarctic/island (A1 carryover) + 24 IGETS Phase D.0 expansion (Strasbourg, Wettzell, Moxa, Conrad, Cantley, Metsahovi, Kamioka, Canberra, Sutherland, Concepcion, Djougou, Onsala, Pecny, Walferdange, Larzac, Schiltach, Trappes, Membach, Bad Homburg, Medicina, Hsinchu, Mizusawa, Lhasa, Apache Point) + 8 USGS continuous gravimeters (Boulder CO, Honolulu HI, Anchorage AK, Albuquerque NM, El Paso TX, Reno NV, Bozeman MT, Fairbanks AK) + 3 academic gravimeters (in IGETS-comparable noise floor). NEW REGIME: MeV-scale axion-like particle coupling accessible only at extended-cohort SNR (per-station SNR ~0.3 too low to discriminate from sub-eV bosonic DM regime). Per-station systematic correction: Hinderer 1995 atm loading + ICET solid-Earth tide + Bos-Scherneck 2013 ocean tidal loading (will upgrade to Path A FES2014b via sibling OSO BLQ kickoff) + per-station 5-yr baseline noise floor. Detection: SNR > 3 at 1/365.25 day^-1 across >=14 of 50 stations with phase consistency. Null: methods-clean upper bound on sub-eV bosonic DM AND axion-like coupling at IGETS-class sensitivity floor -- publishable as tightest combined bound to date (PRL/PRD target). COMPLEMENT to A1 not replacement; both run in parallel for Bayesian evidence stacking with independent cohorts. Cross-validation logic: A1 detects + WILD-3 confirms => high-confidence detection; A1 detects + WILD-3 null => systematic candidate (quiet-cohort issue); A1 null + WILD-3 null => cleanest possible bound. HONEST CAVEATS: per-station systematics may NOT cancel perfectly in coherent stack; emergent signal could be common-mode atmospheric or solar-cycle confound -- Apr 25 Phase 5 solar-cycle artifact gate lesson applies (prompt-gravity collapsed +15sigma -> +2.6sigma post-control). 4 quarterly intermediate checkpoints over 18-month window are trajectory-only, NOT go/no-go; final 18-mo read 2027-11-17 is binding. AX41 daily cron at 35 4 * * * (10 min before MARS-A1 45 4 * * * so combined cohort data is ready). Re-eval 2027-11-17. References: Aurel et al. 2025 PRD on sub-eV bosonic DM seismic coupling; Hinderer 1995 atmospheric loading; Bos & Scherneck 2013 ocean tidal loading; Path A Chalmers FES2014b; Phase D.0 IGETS expansion (memory/phase_mars_sci.md); MARS-A1 hypothesis sister-shot 20260526_mars_a1_annual_modulation_residu_b79122e6.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_seasonal_modulation_extended_cohort"},"reevaluate_at":"2027-11-17","registered_at":"2026-05-26T04:53:33Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"accumulation_window_days":540,"band_hi_hz":0.5,"band_lo_hz":0.1,"covariates":["open_meteo_surface_pressure","swpc_f107_daily","hinderer_1995_atm_loading","bos_scherneck_2013_ocean_tidal_loading","icet_solid_earth_tide"],"n_stations_academic":3,"n_stations_igets_d0":24,"n_stations_quiet_a1":15,"n_stations_target":50,"n_stations_usgs_continuous":8,"null":"random_phase_shuffle_per_station","phase_consistency_min_stations":14,"quarterly_checks":["2026-08-26","2026-11-26","2027-02-26","2027-05-26","2027-08-26"],"rayleigh_p_max":0.05,"regime_meV_scale_axion_like":true,"regime_sub_eV_bosonic_dm":true,"snr_boost_expected":7.07},"method":"stacked_annual_fourier_50station_with_per_station_systematic_correction","n_permutations":1000},"threshold":{"direction":"positive","metric":"stacked_snr_50station","value":3.0},"title":"MARS-WILD-3: 50-station global cohort Dark Matter seasonal modulation residual at sqrt(50) ~ 7x SNR boost over MARS-A1, detection or exclusion of sub-eV bosonic DM AND MeV-scale axion-like particle coupling at SNR > 3 over 18-month forward window","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-11-17"},{"data_source":"USGS NURE radiometric (gamma) + EMIT SWIR overtone (su_) + passive-EM SIP (sip_); ground truth = USGS assay backbone (Produced Waters v2.3 + NGDB + WQP)","expected_direction":"positive","feature_set":["pra_nure_k_eu_th_local_anomaly","su_swir_overtone_band_depths_alOH_carbonate_ferric","sip_cole_cole_relaxation"],"git_commit_at_registration":"92ddc99f2e72a0b6a22c39726356ae0c6b0c7cd3","hypothesis_id":"20260616_multi_modal_elemental_resonance__8fdea1d5","locked_at":"2026-06-16T15:05:42Z","locked_file_sha256":"91dc857206adc8a5520a83d867973f957d55bdce9dd550d243956da12c1197ec","notes":"PRE-REGISTERED BEFORE DATA EXISTS (Jun 16 2026). H: a multi-modal elemental/molecular resonance fingerprint (gamma de-excitation lines via NURE K/eU/eTh local anomaly; SWIR molecular-overtone band depths via EMIT Al-OH 2200 / carbonate 2340 / ferric 900; SIP Cole-Cole relaxation) predicts the dominant NON-radioelement assayed commodity (Cu/Au/Li/brine-TDS) at co-located sites, with 5-fold CV macro-AUC beating BOTH the label-shuffle null 95th pct AND the best single channel by >= 0.03 AUC. POSITIVE CONTROL: gamma must predict radioelement (U from eU, K from K pct); if it fails the pipeline is broken, abort. DATA-READINESS WITHDRAW GATE: needs >= 60 sites with an assay within 10 km of >= 2 REAL channels (data_real=1). AT REGISTRATION: assay backbone ABSENT from disk (regeneratable via ingest_geochem_assays scripts); su_ and sip_ return data_real=0 at all Western US test sites (EMIT swath-limited, SIP no co-located passive EM); only gamma broadly real. If after assay rebuild + EMIT densification < 60 qualifying sites, WITHDRAW as underpowered. NULL is publishable. Context: single-channel gamma to commodity is ALREADY proven (NURE K/eU local anomaly r=+0.54 vs federal-lands MMBO, Phase 65y B2). This tests whether multi-modal fusion beats single-channel.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geochemistry"},"reevaluate_at":"2026-09-16","registered_at":"2026-06-16T15:04:41Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"cv_classification_vs_permutation_null_and_best_single_channel","n_permutations":1000},"threshold":{"direction":"positive","metric":"delta_macro_auc_multimodal_minus_best_single_channel","value":0.03},"title":"Multi-modal elemental-resonance fingerprint predicts non-radioelement commodity at assay sites","withdrawn":null,"_ots_anchored":false,"_status_dir":"locked","_re_eval_date":null},{"data_source":"dv/v reference stacks (HVU-PTU + corridor sites); EarthScope EMTF + EMC-CONUS-MT-2019 remote MT; DWMP/artesian-potential scorecards; helium DRI grid","expected_direction":"positive","feature_set":["ogc_score_structural_trap","ogc_score_deep_water_signal","emt_rho_deep_ohmm","ogr_oil_gas_probability_0_1"],"git_commit_at_registration":"a4f64dcfd3921c6c1a20bf357df4fc7207dcf98a","hypothesis_id":"20260618_faulted_trap_deep_confined_fluid_d3eff985","locked_at":"2026-06-18T00:21:59Z","locked_file_sha256":"14bb499c869da406f914ac793e99aa98398e4fb67dc60fb5441f6099f465d756","notes":"DeepMap pre-registration (USU/UGS well logs). PREDICT: each site hosts a confined fluid column at depth within its faulted graben/basin - WATER at Pahvant/Kingston, HYDROCARBON-bearing at CISCO (Greater Cisco field). EVIDENCE: structural-trap axis 79-88, deep-water signal 69-70; real remote MT corroborates - Pahvant/Kingston conductive deep ~1-2 ohm-m (water), CISCO resistive 190 ohm-m backing out PROBABLE_HC (Archie contrast 9.16). WRONG IF: no confined fluid column at depth (dry/tight, no pressure compartment). CAVEAT: CISCO O&G COMPOSITE low because thermal-maturity axis starved (no well-T); HC-contrast Archie DOES fire but is thermal-gated. USU/UGS adjudicates via induction/laterolog logs, DSTs, pressure data.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"groundwater"},"reevaluate_at":"2027-06-17","registered_at":"2026-06-18T00:21:46Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"usu_ugs_dogm_well_log_validation","n_permutations":null},"threshold":{"direction":"positive","metric":"n_sites_with_confined_fluid_column_at_depth","value":2.0},"title":"Faulted trap + deep confined fluid column at Kingston Pahvant CISCO (at least 2 of 3)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"dv/v reference stacks (HVU-PTU + corridor sites); EarthScope EMTF + EMC-CONUS-MT-2019 remote MT; DWMP/artesian-potential scorecards; helium DRI grid","expected_direction":"negative","feature_set":["dvv_deep_band_trend_pct_yr","dvv_deep_band_seasonal_pct","dvv_shallow_band_seasonal_pct"],"git_commit_at_registration":"a4f64dcfd3921c6c1a20bf357df4fc7207dcf98a","hypothesis_id":"20260618_pahvant_deep_water_replenishment_c6033cb7","locked_at":"2026-06-18T00:21:59Z","locked_file_sha256":"f034833362aa4666f996717b4bef68d9c95e2f0bf31817bf0bee351e14f8a86e","notes":"DECISIVE CLAIM. DeepMap pre-registration (USU isotope lab). PREDICT: water chemistry/isotopes show OLD/DEEP signature: tritium dead (<0.8 TU) AND/OR C14 old AND/OR 3He/4He elevated above atmospheric. EVIDENCE: dv/v deep band (0.1-0.5 Hz ~1-3 km) carries -0.199 %/yr trend + 1.16% seasonal swing peaking OCTOBER; shallow band (2-8 Hz) nearly flat (0.10%, peaks April w/ snowmelt) -> deep ~2x trend, ~10x seasonality, out of phase w/ surface snowmelt = lateral/upward pressure transmission, not local infiltration. WRONG IF: tritium >0.8 TU / C14 modern / d18O-d2H matches recent local snowmelt (young meteoric dominates). CAVEAT: dv/v cc moderate 0.66-0.68; mid band no data; robust part is the deep-vs-shallow CONTRAST not absolute magnitude. USU adjudicates via 3He/4He, tritium, d18O/d2H, C14 (helium-survey gas data already exists).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"groundwater"},"reevaluate_at":"2027-06-17","registered_at":"2026-06-18T00:21:46Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"usu_isotope_tritium_he3he4_c14_d18O","n_permutations":null},"threshold":{"direction":"negative","metric":"tritium_TU_of_producible_deep_water","value":0.8},"title":"Pahvant deep-water replenishment is deep (1-3 km) not surface: producible water tritium is dead","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"dv/v reference stacks (HVU-PTU + corridor sites); EarthScope EMTF + EMC-CONUS-MT-2019 remote MT; DWMP/artesian-potential scorecards; helium DRI grid","expected_direction":"positive","feature_set":["lei_data_real","dwmp_deep_water_score","ape_artesian_potential","emt_rho_deep_ohmm"],"git_commit_at_registration":"a4f64dcfd3921c6c1a20bf357df4fc7207dcf98a","hypothesis_id":"20260618_pahvant_hansel_corridor_hosts_de_5a812e3a","locked_at":"2026-06-18T00:21:59Z","locked_file_sha256":"f5875e43ed514eb719c64da84a282c389e7e1f3ddb8e3237f532741dd61e3616","notes":"DeepMap pre-registration (USU/Stephanie Froham + UGS). PREDICT: a borehole/UGS+DOGM deep-well record in the Hansel-Pahvant corridor encounters a confined aquifer with measurable excess hydraulic head (artesian/sub-artesian) at ~300-1500 m. EVIDENCE: helium DRI grid over faulted graben; flowing-artesian + confined-aquifer signals; new real remote MT shows conductive deep column ~1-2 ohm-m (water-bearing). WRONG IF: well data shows no confined aquifer / no excess head (unconfined hydrostatic only). CAVEAT: deep water likely BRACKISH/SALINE (MT ~1-2 ohm-m), i.e. deep+replenishable but mineralized. USU adjudicates via UGS/DOGM well records + new borehole head.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"groundwater"},"reevaluate_at":"2027-06-17","registered_at":"2026-06-18T00:21:46Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"usu_ugs_well_control_validation","n_permutations":null},"threshold":{"direction":"positive","metric":"confined_aquifer_with_excess_head_present","value":1.0},"title":"Pahvant-Hansel corridor hosts deep confined artesian water (excess head at 300-1500 m)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"helium_iot_poc_witness_rssi + meshtastic_mqtt_rssi","expected_direction":"negative","feature_set":["regional_rssi_anomaly_stack","relative_humidity_2m","temperature_2m"],"git_commit_at_registration":"709798eb40dcaae54b1884c773b40b125c72bfd4","hypothesis_id":"20260620_mesh_rf_link_rssi_as_a_temperatu_ddf175e3","locked_at":"2026-06-20T17:04:56Z","locked_file_sha256":"36344b6276cb3e280e290c07f84b9f2bccceac9ee4f22b11df29f2e3517b6cec","notes":"Pre-registered 2026-06-20 before forward data exists. CLAIM: in humid-climate regions the temperature-controlled partial correlation between the coherent regional RSSI-anomaly stack (z-score each fixed pair daily RSSI, average across pairs per day) and daily relative humidity is NEGATIVE (humidity up -> whole-mesh RSSI down). CONFIRMATORY CRITERIA on forward data 2026-06-20..2026-09-20 via weekly multi-city HeliumGeek harvest (>=15 humid + >=6 dry-control cities, >=60 daily stack-days/region): (1) median humid-region partial_r(RSSI,humidity|temp) <= -0.20; (2) >=70pct of humid regions individually negative; (3) dry-control median abs(partial_r) < 0.10; (4) cross-network: same sign on independent Meshtastic network; (5) >=1 rain-event response. Distance-split FLAT so coupling is common-mode (mechanism likely receiver-thermal/antenna/ground-moisture not path-loss). Honest NULL publishable. Exploratory basis NOT confirmatory: NYC partial -0.285 Seattle -0.548 at n=8 days Jun 2026.","ots_proof_path":"hypotheses/locked/20260620_mesh_rf_link_rssi_as_a_temperatu_ddf175e3.json.ots","outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"weather_humidity"},"reevaluate_at":"2026-09-20","registered_at":"2026-06-20T17:04:42Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"regional_coherent_stack_partial_correlation_controlling_temperature","n_permutations":null},"threshold":{"direction":"negative","metric":"median_humid_region_partial_r_rssi_humidity_given_temp","value":-0.2},"title":"Mesh RF link RSSI as a temperature-controlled passive humidity sensor","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-09-20"},{"data_source":"through_globe_stack.db GSN/IRIS/GEOSCOPE antipodal ambient-noise interferometry","expected_direction":"two_sided","feature_set":["dt_bcdf_inner_core","pkikp_lag","pkp_lag"],"git_commit_at_registration":"4b8f437f119df8daa7ec446a818d4d2bedca560e","hypothesis_id":"20260623_ambient_noise_inner_core_differe_868fb110","locked_at":"2026-06-23T02:58:53Z","locked_file_sha256":"13d937cee0eb491c3f9b8bb1fed6a01ca6b7a548f0683efd99095e5a3ea3c7bb","notes":"PRE-REGISTRATION -- independent ambient-noise test of inner-core differential motion.\n\nBACKGROUND. The inner-core differential-rotation debate (Yang & Song 2023; Vidale; Tkalcic) rests almost entirely on repeating-earthquake DOUBLET PKIKP(df)-PKPbc differential traveltimes, criticised for sparse, uneven, source-mislocation-prone sampling. We test the SAME observable by a fully INDEPENDENT method: ambient-noise interferometry. The through-globe stacker cross-correlates continuous LH? records of antipodal pairs (>150 deg) and accumulates the core-traversing Green's function nightly (SNR ~ sqrt(N_days)).\n\nOBSERVABLE. Per pair, inner_core band (0.02-0.05 Hz): dt_bcdf(day) = |PKP_lag| - |PKIKP_lag|. PKIKP (=PKPdf) measured in +/-30 s of its ak135 surface-to-surface arrival; PKP = earliest outer-core branch (bc ~146-155 deg, ab beyond). PKIKP samples the inner core; PKP samples only the fluid outer core, so the DIFFERENTIAL isolates inner-core change and cancels mantle/source common-mode (incl. seasonal noise-source migration to first order).\n\nREALITY GATE (prerequisite). Only pairs whose cumulative core SNR climbs as sqrt(N) (Pearson r(SNR, sqrt(N_days)) >= 0.6, positive slope) are admitted -- excludes converged station/instrument artifacts.\n\nANALYSIS (locked). Per admitted inner_core pair fit slope_pair = d[dt_bcdf]/dt over the window. Ensemble statistic = median per-pair drift (s/yr) with bootstrap-over-pairs 95% CI. CONTROL: also report the raw (uncontrolled) PKIKP-lag drift; the differential must suppress the common-mode seasonal drift relative to raw for the result to count.\n\nCONFIRMATORY CRITERION (detection). Ensemble median BC-DF drift non-zero at >3 sigma (bootstrap), sign-coherent across a majority of admitted pairs, AND exceeding the raw-PKIKP seasonal-drift floor -> independent ambient-noise detection of inner-core differential motion.\n\nNULL / BOUND (also publishable). Drift consistent with zero -> an independent UPPER BOUND on inner-core differential traveltime change over the 2026-2027 epoch, free of doublet sampling biases. The 2020s are argued to sit near an oscillation turning point (low rate), so a near-zero result is a physically legitimate, expected-plausible outcome, NOT a method failure.\n\nWITHDRAW / UNDERPOWERED. If by month 6 fewer than 5 inner_core pairs pass the reality gate, OR per-pair scatter exceeds resolvable drift (no |drift|<0.2 s/yr detectable), declare UNDERPOWERED and re-register with longer window / larger aperture (station count grown 25->53 on 2026-06-22).\n\nDATA/CODE. Frozen method = scripts/through_globe_stack.py at lock-time commit; observable accrues in data/through_globe_stack.db (antipodal_daily.dt_bcdf), populating per-phase daily lags from 2026-06-22 forward. Re-evaluate 2027-06-22.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"inner_core_differential_motion"},"reevaluate_at":"2027-06-22","registered_at":"2026-06-23T02:58:42Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"bcdf_differential_drift_ensemble","n_permutations":null},"threshold":{"direction":"two_sided","metric":"ensemble_bcdf_drift_sigma","value":3.0},"title":"Ambient-noise inner-core differential motion via the BC-DF differential of through-globe Green's functions","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-06-22"},{"data_source":"prediction_ledger.db forecast stream + USGS/EMSC/multi-catalog matched M>=5 events + ETAS + per-region base rates","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"7ed4511d41459fa3d5863ab0b66d7c78ed7ef2d4","hypothesis_id":"20260701_dogma_a_operational_eq_forecasts_13b7ef38","locked_at":"2026-07-01T15:07:40Z","locked_file_sha256":"21b435019cd0c855baaa0a5a02e657cedf2abb8eb896ef835edd5a76b0981a65","notes":"DOGMA A -- the consensus that operational earthquake forecasting adds no skill over a random/ETAS/base-rate model. Pre-registered FORWARD HELD-OUT test: DeepMap's operational EQ forecast stream (convergence_gate + tensor_eq_per_region + tensor_eq_global probabilities in the hash-chained ledger) beats BOTH (1) a per-region base-rate Poisson baseline AND (2) a Reasenberg-Jones ETAS baseline, on M>=5.0 events over 2026-07-01..2027-07-01, scored by Brier Skill Score on (region,day) P(M>=5 in next 168h) cells, at BSS>=0.02 vs base rate AND BSS>0 vs ETAS (bootstrap p<0.05; equivalently dAUC>=0.03 vs better baseline). BASE-RATE SKILL CHECK (the crux, added per the Japan M6.0 case): credit is measured RELATIVE to each region's OWN base rate, so a wide-band HIGH watch on a hyperactive region that merely tracks the base rate scores BSS~0 (no skill). Archetype this test adjudicates: the 2026-07-01 M6.0 off Noda, Japan -- caught by a standing M4.5-8.0 HIGH watch that matched 100+ observations that week; it counts as skill ONLY if the forecast probability/timing beat Japan's base rate. PASS = skill over both baselines (dogma broken). NULL = no skill over base rate (landmark null: catches are base-rate scenery). Magnitude specificity (band width) is a SEPARATE test. Scoring code to be built before re-eval; method fully specified; lock + OTS anchor prove the plan predates the forward data. Brand: DeepMap AI Inc.","ots_proof_path":null,"outcome":{"lookahead_hours":168.0,"min_magnitude":5.0,"type":"earthquake"},"reevaluate_at":"2027-07-01","registered_at":"2026-07-01T15:07:40Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"archetype":"2026-07-01 M6.0 Noda Japan wide-band HIGH watch matched 100+ obs/week","baseline_1_baserate":"per-region historical M>=Mc rate -> Poisson P(event in horizon) [THE base-rate skill check]","baseline_2_etas":"Reasenberg-Jones ETAS a=-1.67 b=0.91 c=0.05 p=1.08 on live catalog","event_set":"M>=5.0 global bounds-first matched","forecast_horizon_hours":168,"forecast_stream":["convergence_gate","tensor_eq_per_region","tensor_eq_global"],"forward_window":["2026-07-01","2027-07-01"],"magnitude_specificity":"scored separately","null_outcome":"no skill over base rate -> landmark null","pass":"BSS>=0.02 vs baserate AND BSS>0 vs ETAS bootstrap p<0.05 (or dAUC>=0.03)","primary_metric":"Brier Skill Score vs baseline_1 on (region,day) cells","secondary_metric":"ROC-AUC delta vs better baseline"},"method":"forward_held_out_skill_vs_etas_and_baserate","n_permutations":null},"threshold":{"direction":"positive","metric":"brier_skill_score_vs_regional_poisson_baserate","value":0.02},"title":"Dogma A: operational EQ forecasts carry skill over ETAS + regional base-rate baselines (12-month held-out forward test)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-07-01"},{"data_source":"microseism_precursor + convergence_gate + CID tensor channels + USGS/EMSC M>=7 events","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"84e279cb7a5aae80118d169b21f7d7ea1b14d6f8","hypothesis_id":"20260701_dogma_c_a_multi_day_precursor_to_1f92590a","locked_at":"2026-07-01T15:33:40Z","locked_file_sha256":"a2fa3ef7b956403077f1de1bfb228dbeb903fda6aaba53f9699b2953b1b1b8ef","notes":"DOGMA C -- the strong-form unpredictability dogma: that NO passive channel carries pre-rupture information at MULTI-DAY lead before a great earthquake. Pre-registered stacked-epoch FORWARD test: across ALL M>=7.0 events in 2026-07-01..2027-07-01, a LOCKED multi-channel tensor feature (stacked-epoch z of microseism secondary-band precursor + convergence-gate multichannel + CID, in the pre-event T-72h..T0 window) shows a >3sigma pre-event excursion vs a time-shuffled null, AND survives the solar-cycle + seasonal controls that collapsed the +15sigma prompt-gravity headline to +2.6sigma (Apr-25). The honest control is the whole point. PASS = >3sigma survives controls -> a multi-day precursor demonstrably exists (dogma broken). NULL = <=3sigma or collapses -> the definitive multi-day-precursor-impossibility null the field needs (landmark, publishable). Withdraw if <8 M7+ events (underpowered). Companion to Dogma A (event skill) + the microseism 24h shadow notice. Brand: DeepMap AI Inc.","ots_proof_path":null,"outcome":{"lookahead_hours":72.0,"min_magnitude":7.0,"type":"earthquake"},"reevaluate_at":"2027-07-01","registered_at":"2026-07-01T15:33:40Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"basis":"Phase-5 microseism +19sigma at -10h + convergence-gate multichannel + CID; solar controls that killed the +15sigma prompt-gravity headline","controls":"solar-cycle (F10.7 quartile match) + seasonal (day-of-year)","event_set":"ALL M>=7.0 global forward, bounds-first","forward_window":["2026-07-01","2027-07-01"],"locked_feature":"stacked-epoch z of {microseism_secondary_band_3s_z, convergence_gate_multichannel, CID} in pre-event T-72h..T0","null":"time-shuffle null on the stacked statistic","null_outcome":"<=3sigma or collapses -> definitive no-multi-day-precursor null","pass":"pre-event stacked z >3sigma AND survives solar+seasonal controls -> multi-day precursor exists (dogma broken)","withdraw":"<8 M7+ events -> underpowered"},"method":"stacked_epoch_M7_multichannel_precursor_forward","n_permutations":null},"threshold":{"direction":"positive","metric":"pre_event_stacked_z_solar_controlled","value":3.0},"title":"Dogma C: a multi-day precursor to a great earthquake exists -- stacked-epoch M7+ forward test with honest solar controls","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-07-01"},{"data_source":"IGETS SG (data/igets_l2.db 16.27M rows / 14 codes / 2008-2025) + GSN broadband (dm_psd_cache 123 stations @ ~20Hz) + public FDSN","expected_direction":"positive","feature_set":["network_coherent_snr","cygnus_moveout_percentile","best_fit_speed_kms","best_fit_direction_offset_deg","null_A_shuffle_snr","null_B_offaxis_snr","degeneracy_frac_differ","transit_rate_95UL_per_yr"],"git_commit_at_registration":"66b3f0a2dbf4b9818fff078a0610eac1f18d85d0","hypothesis_id":"20260701_mars_8_earth_as_dm_detector_cygn_05bbd9ec","locked_at":"2026-07-01T13:32:42Z","locked_file_sha256":"171f68f8027cbb33579cb194b4761090989345d69a6fe2e77c470b143613e8ce","notes":"MARS-SHOT #8 -- DRAFT PRE-REGISTRATION (NOT YET LOCKED; REQUIRES MIKE SIGN-OFF).\n\nHYPOTHESIS: If macroscopic/composite dark-matter objects (nuclearites, strangelets, primordial-black-hole-scale clumps, Q-balls) transit through or near Earth at galactic velocity, the global network of superconducting gravimeters (IGETS SG) + broadband seismometers (GSN) will record coincident transients whose network arrival-time moveout forms a plane wave at ~230 km/s (+- Maxwell-Boltzmann dispersion sigma_v ~ 156 km/s, v_esc ~ 544 km/s) locked to the CYGNUS direction (the DM wind from the Sun's motion through the halo; solar apex ~ RA 18h Dec +38 / Galactic l=90 b=0). No terrestrial (weather, cultural, quarry, calibration) or astrophysical (teleseism, tide, microseism) source produces a Cygnus-directed ~230 km/s plane-wave moveout, so that geometry separates signal from all systematics.\n\nSIGNAL TEMPLATE: per-station Newtonian tidal-tug (or contact/drag) transient parameterized by (clump mass M, transit speed v, geocentric impact parameter b, arrival direction n_hat). Duration tau ~ b/v; antisymmetric pass-by shape. Absolute amplitude calibration (nGal per kg per km) derived in scripts/dm_moveout_search.py + Week-1 report; NOT asserted in the pre-reg to avoid a stale sensitivity claim.\n\nDETECTION STATISTIC: network-coherent SNR = sum_i w_i * mf_i(t_i(n_hat,v)) / sqrt(sum_i w_i^2), where mf_i is the matched-filter output at station i and t_i is the predicted moveout arrival. MAXIMIZED over the galactic speed window and a cone about the Cygnus-away propagation vector.\n\nPARAMETER GRID (LOCKED at lock-time; see test-extra JSON): mass window 1e6..1e15 kg (7 log bins); speed window v in [V_LSR - 2 sigma_v, V_LSR + 2 sigma_v] = [-82..542] clipped to [1, 542] km/s (9 bins); impact parameter b in {inside-Earth chord, 0.5 R_E, 1 R_E, 3 R_E}; direction cone half-angle 30 deg about the per-epoch Cygnus-away vector, 24 trial directions.\n\nDETECTION THRESHOLD (LOCKED): a candidate is a DM-transit CANDIDATE iff (1) network-coherent SNR >= 5.0 at its best-fit (n_hat, v); AND (2) best-fit direction within the Cygnus cone (30 deg) with on_cyg_percentile >= 0.9 vs the off-Cygnus full-sphere distribution; AND (3) best-fit speed within the galactic MB window. A single-station excess is explicitly NOT a candidate.\n\nTWO NULLS (both required): NULL A time-shuffle -- independently circular-shift each station MF stream, rebuild the statistic at Cygnus, repeat N=100+ -> background SNR distribution. NULL B direction-rotate -- with real data, cone-maximize the statistic on off-Cygnus axes over the full sphere (excluding a 30 deg Cygnus cone) and rank the (equally cone-maximized) on-Cygnus statistic against that distribution; a real signal peaks ONLY at Cygnus (percentile ~1), a coherent systematic peaks elsewhere. FAIR COMPARISON: on-Cygnus is cone-maximized identically to off-Cygnus (no look-elsewhere asymmetry).\n\nDEGENERACY GUARD (Apr-28 lesson): before trusting either null, verify frac_shuffles_differ >= 0.5 (the shuffle actually changes the statistic). If the null is degenerate, report a Poisson-only UL and FLAG it; do NOT report a shuffle-derived uncertainty. This directly heeds the April 2026 transit-DM bound audit where 88 percent of snapshots tripped stage-1 coincidence making the shuffle a no-op.\n\nEXCLUSION COMPUTATION: null-subtracted Helene-style 95 percent UL on transit RATE (events/yr), identical math to scripts/compute_null_subtracted_bounds.py: s_hat = n_real_candidates - mean(n_null_candidates); sigma_s = sqrt(n_real + n_null_mean); s95 = max(0, s_hat) + 1.96 sigma_s; rate_95 = s95 / span_yr. The Cygnus-direction discriminant is folded into what counts as a candidate, so this is a DIRECTION-RESOLVED companion bound to the existing PBH <=8.4/yr and strangelet <=10.8/yr bounds.\n\nSENSITIVITY WINDOW (honest): SG + broadband seismometer network probes MACROSCOPIC / COMPOSITE DM with GRAVITATIONAL or CONTACT coupling -- NOT WIMPs (which need underground cryogenic detectors). SG noise floor ~1 nGal = 1e-11 m/s^2 over minutes (slow, large-M end); broadband ~1e-9 m/s^2 over seconds (fast, small-b end). Moveout directional power scales with delay-to-noise ratio: at 230 km/s the max inter-station delay across an Earth-radius aperture is only ~28 s, so directional discrimination is STRONGEST for sharp fast transients (GSN broadband) and weakest for slow SG minute-cadence pulses. This is a scaffold finding to be carried into the mass/coupling reach table.\n\nMOST LIKELY OUTCOME: a world-leading direction-resolved EXCLUSION on the macroscopic-DM transit rate in the probed mass/coupling window; small chance of a Cygnus-locked candidate demanding blind independent follow-up before any claim.\n\nDATA: IGETS SG data/igets_l2.db = 16.27M gravity-residual rows across 14 station-codes, 2008-2025 (CO/ST 17-yr span; ~10 stations 2008-2010 dense; cadence 60-326 s). GSN broadband dm_psd_cache station_meta = 123 stations @ ~20 Hz, extensible via public FDSN. CAVEAT: several IGETS 2-letter L2 codes are ambiguous (e.g. ST = Strasbourg OR Sutherland) and MUST be reconciled against data/igets/<Station>/*.ggp headers before the real scan.\n\nBRAND: DeepMap AI Inc. only. Extends scripts/compute_null_subtracted_bounds.py + scripts/dm_directional_search.py + reports/moonshots_retro/ (does not reinvent the bound math). Real scan gated behind DEEPMAP_DM_MOVEOUT_PREREG_LOCKED=1, set ONLY after this pre-reg is reviewed by Mike, locked, and Bitcoin-anchored.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_transit_exclusion"},"reevaluate_at":"2027-12-23","registered_at":"2026-07-01T04:49:28Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{"bound_formula":"helene_null_subtracted_95_UL","bound_z":1.96,"cone_half_angle_deg":30.0,"cygnus_apex_dec_deg":38.0,"cygnus_apex_ra_deg":270.0,"cygnus_galactic_l_b_deg":[90.0,0.0],"degeneracy_frac_differ_min":0.5,"gsn_psd_cache_db":"data/dm_psd_cache.db","igets_l2_db":"data/igets_l2.db","impact_param_R_earth":[0.0,0.5,1.0,3.0],"mass_kg_window":[1000000.0,1000000000000000.0],"n_cone_directions":24,"n_mass_bins":7,"n_speed_bins":9,"network_coherent_snr_threshold":5.0,"null_A_time_shuffle_n":100,"null_B_exclude_cone_deg":30.0,"null_B_off_axis_n_dir":200,"on_cyg_percentile_threshold":0.9,"real_scan_gate_env":"DEEPMAP_DM_MOVEOUT_PREREG_LOCKED","scaffold_script":"scripts/dm_moveout_search.py","sigma_v_kms":156.0,"speed_kms_window":[1.0,542.0],"v_esc_kms":544.0,"v_lsr_kms":230.0},"method":"network_coherent_cygnus_moveout_matched_filter","n_permutations":100},"threshold":{"direction":"positive","metric":"network_coherent_snr_at_cygnus","value":5.0},"title":"MARS-8 Earth-as-DM-detector: Cygnus-moveout macroscopic-dark-matter transit search (SG + GSN network)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"gas-sample isotope lab (3He/4He, C1-C4 d13C, clumped 13CH3D/12CH2D2) + abiogenic_hc_discovery_monitor screen","expected_direction":"positive","feature_set":["ahc_composite_score","ahc_he3_he4_ra","ahc_score_serpentinization"],"git_commit_at_registration":"794be60d23c99fc4322e9202fafabb7c84de3636","hypothesis_id":"20260703_abiogenic_hc_multi_tracer_conver_97b00a36","locked_at":"2026-07-03T18:47:48Z","locked_file_sha256":"e2ad8f42edcedc073a0e5bb74954a4d6934986b8ec691cbeaab1b7180837a764","notes":"PRE-REGISTERED CANDIDATE SET (ranked by literature 3He/4He R/Ra, the mantle-source tracer), locked BEFORE any new gas sampling so that any future sampling (by us or in the literature) is a blind forward test of the screen: (1) Chimaera Eternal Flames, Turkey R/Ra 5.1 -- STANDOUT: onshore, actively venting CH4, serpentinite-hosted, samplable; (2) Lost City Hydrothermal Field R/Ra 7.8 (offshore type site); (3) Black Rock Volcanic Field, Millard County UT R/Ra 5.8 (local, Basin-and-Range); (4) Salton Sea Geothermal CA R/Ra 6.2; (5) Dixie Valley NV R/Ra 5.4. HONEST FRAMING: R/Ra is already known (necessary, NOT sufficient -- a mantle plume like Yellowstone R/Ra 16 has high He but no HC), so it is not the forward prediction. The genuine forward test is the markers NOT yet measured at these specific sites: PASS = at >=3 of the 5 sites, sampled gas shows >=2 of {d13C isotopic reversal (C1 heavier than C2), clumped-isotope apparent formation T > 150 C, serpentinization H2 co-occurrence with no hopane/sterane biomarkers}. The strongest discriminating variant: HC found where NO biogenic source rock exists. WITHDRAW if no gas-sample isotope data at >=3 sites by re-eval date. Screen: cross_physics/abiogenic_hc_discovery_monitor + abiogenic_replenishment.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"abiogenic_hydrocarbon"},"reevaluate_at":"2028-07-04","registered_at":"2026-07-03T18:47:29Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"multi_tracer_convergence","n_permutations":null},"threshold":{"direction":"positive","metric":"n_sites_with_>=2_abiogenic_markers","value":3.0},"title":"Abiogenic-HC multi-tracer convergence at five high-R/Ra serpentinization-context sites","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"data/dvv_reference.db (28.2M dv/v measurements, 2015-present, 12405 UU pairs) + Utah FORGE public injection schedule (utahforge.com / DOE GDR)","expected_direction":"negative","feature_set":["dvv_pct_milford_stack"],"git_commit_at_registration":"931634c5df80d55d2810ffb3f5e276e6d20435b4","hypothesis_id":"20260712_passive_dv_v_detects_effective_s_7e043492","locked_at":"2026-07-12T21:20:05Z","locked_file_sha256":"27585e75c158abe395d93a655fa29e1d812e4e0fef65f8374faed59ea6df296d","notes":"SPRINGBOARD LINEAGE: first invention through the test-before-file loop. Expired US6977866B2 (ConocoPhillips, Huffman et al) velocity-to-effective-stress transform, surfaced+deep-read by the Invention Springboard Miner 2026-07-12; proposed combination = drive the Bowers/Eaton velocity-effective-stress relation with passive ambient-noise dv/v for 4D pore-pressure monitoring with no active source and no wells. CLAIM: during documented Utah FORGE injection/circulation episodes (2024-04-03..2024-04-27 commercial stimulation + 9h circulation; 2024-07-15..2024-09-15 30-day circulation + extended tests; and any future documented episode), stacked ambient-noise dv/v across Milford-area station pairs DECREASES below the pre-episode baseline (velocity drop = effective-stress decrease as injected fluid raises pore pressure), with injection-window stacked z <= -2.0, while matched same-length control windows outside episodes trigger at <20 pct. PASS: >=1 covered episode meets z<=-2.0 AND control trigger rate <20 pct. FAIL: covered episodes but no z<=-2.0, or controls trigger >=20 pct (signal indistinguishable from background). WITHDRAW GATES (coverage, decided by the runner before any outcome stat is read): (a) an episode is testable only with >=5 Milford-area pairs and >=60 pct of episode+baseline days having a stacked value; (b) if NO documented episode is covered, hypothesis converts to PROSPECTIVE for the next announced FORGE episode instead of being scored. Confounders acknowledged: seasonal/hydrologic dv/v variation (why matched controls are required); dv/v shallow sensitivity vs reservoir depth ~2.5 km is the honest physics risk -- a null is publishable. Runner: scripts/flagship_dvv_forge_pressure_test.py. Test-then-file: a provisional is drafted ONLY if this passes. Re-eval 2026-10-12.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"pore_pressure_change"},"reevaluate_at":"2026-10-12","registered_at":"2026-07-12T21:19:50Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"stacked_dvv_z_vs_baseline_and_matched_controls","n_permutations":null},"threshold":{"direction":"negative","metric":"injection_window_stacked_z","value":-2.0},"title":"Passive dv/v detects effective-stress (pore-pressure) change at Utah FORGE injection episodes (springboard US6977866B2)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-10-12"},{"data_source":"Continuous 3C broadband/short-period waveforms from ≥3 (ideally dozens of) stations over the target (already in FDSN/DeepMap); a 3D velocity model (DeepMap has/derives one); solid-Earth tide + barometric loading time series (computable from platform inputs); regional σ1 azimuth from focal mechanisms/World Stress Map; and for co-inversion, ambient-noise dv/v (already a DeepMap product). All passive","expected_direction":"","feature_set":[],"git_commit_at_registration":"27c31a902e32b849a210a59c52e1bff84b9dcdb7","hypothesis_id":"20260712_springboard_test_us7127353b2_met_90aefeea","locked_at":"2026-07-12T22:59:30Z","locked_file_sha256":"5c395ecd4fd00d46046601f4669b063b385b15a91332d1970cd0948a0f9c9f1f","notes":"SPRINGBOARD LINEAGE (invention loop). Brief-derived hypothesis approved by Mike. CLAIM: In a reservoir/fracture volume under measurable cyclic loading, the SET coherence (semblance) energy of critically-stressed fracture fairways — those oriented within ±45° of the maximum horizontal compressive stress — will rise and fall in phase with the solid-Earth-tide + barometric stress cycle, with a phase lag and amplitude that predict the fairway's hydraulic linkage, such that fairways showing tidally-modulated emission also show the largest co-located dv/v decrease during the same stress phase. TEST: Retrospective, blind-window design on one instrumented target: (1) compute per-voxel SET semblance in 10 s stacks over ≥60 days; (2) independently model the tidal+barometric stress time series and the σ1 azimuth (from focal mechanisms / stress-map database); (3) pre-register which voxel fairways are 'critically stressed' (within ±45° of σ1) BEFORE looking at their emission phase; (4) test whether emission energy of the pre-registered fairways is significantly phase-coherent with the stress cycle (Rayleigh/phase-stacking test) versus non-aligned control fairways; (5) test whether the same voxels show correlated dv/v change. Pass = phase coherence significant only in the critically-stressed set and dv/v correlation present. PASS: Critically-stressed (±45° of σ1) fairway voxels show tidal-phase emission coherence at p<0.01 (Rayleigh) with resultant vector length R>0.2, while σ1-misaligned control voxels do not (their R indistinguishable from noise); AND ≥60% of the tidally-modulated fairway voxels show a co-located dv/v decrease during the peak-stress phase exceeding the 2σ background dv/v fluctuation. Failure = no phase preference, or equal coherence in aligned and misaligned sets. DATA: Continuous 3C broadband/short-period waveforms from ≥3 (ideally dozens of) stations over the target (already in FDSN/DeepMap); a 3D velocity model (DeepMap has/derives one); solid-Earth tide + barometric loading time series (computable from platform inputs); regional σ1 azimuth from focal mechanisms/World Stress Map; and for co-inversion, ambient-noise dv/v (already a DeepMap product). All passively available — no new field deployment for the initial test.  (already collected: YES). Test-before-file: provisional drafted only on pass.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-12","registered_at":"2026-07-12T22:59:30Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"springboard_hypothesis_test","n_permutations":null},"threshold":{"direction":"","metric":"","value":null},"title":"Springboard test: US7127353B2 - Method and apparatus for imaging permeability pathways of geologic fluid reservoirs using seismic emission tomography","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"Already collected: continuous 3C FDSN waveforms, DeepMap 3D velocity model, and live dv/v ambient-noise fields are all in the Earth State Tensor. External-but-free: a published reviewed earthquake catalog for the same region/period for ground truth. Nothing new needs to be deployed.  (already collected: YES)","expected_direction":"","feature_set":[],"git_commit_at_registration":"27c31a902e32b849a210a59c52e1bff84b9dcdb7","hypothesis_id":"20260712_springboard_test_us7391675b2_mic_f603e70e","locked_at":"2026-07-12T22:59:30Z","locked_file_sha256":"924679cb14f827722d66a386817aed092d338d1345ef67c19bbee6ad3232225f","notes":"SPRINGBOARD LINEAGE (invention loop). Brief-derived hypothesis approved by Mike. CLAIM: Adaptive-Tc continuous coalescence migration over an existing sparse regional FDSN array will automatically detect and locate microseismic events at least 0.3 magnitude units below the network's published catalog magnitude-of-completeness, with median epicentral location error ≤ 2 km versus the catalog, with zero human picking. TEST: Pick a region and a 30-day window for which a high-quality reviewed catalog exists (e.g. USGS/regional network). Run the coalescence migration on the raw continuous 3C waveforms for that window blind to the catalog. Compare auto-declared events to the catalog: match by ≤10 s origin-time and ≤5 km distance. Metrics: (a) fraction of catalog events recovered vs magnitude; (b) number of confidently declared events BELOW Mc and their independent corroboration (waveform stacking, template matching); (c) median location error of matched events; (d) false-alarm rate against quiet control periods. Ablation: fixed-Tc vs adaptive-Tc(dv/v) to isolate the novel contribution. PASS: Pass if ≥1 event is confidently declared and independently corroborated at ≥0.3 Mu below Mc AND matched-event median location error ≤2 km AND adaptive-Tc ablation shows ≥15% more sub-Mc detections than fixed-Tc at equal false-alarm rate. Fail otherwise. DATA: Already collected: continuous 3C FDSN waveforms, DeepMap 3D velocity model, and live dv/v ambient-noise fields are all in the Earth State Tensor. External-but-free: a published reviewed earthquake catalog for the same region/period for ground truth. Nothing new needs to be deployed.  (already collected: YES). Test-before-file: provisional drafted only on pass.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-12","registered_at":"2026-07-12T22:59:30Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"springboard_hypothesis_test","n_permutations":null},"threshold":{"direction":"","metric":"","value":null},"title":"Springboard test: US7391675B2 - Microseismic event detection and location by continuous map migration","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"SHADOW record data/imminent_shadow.jsonl (private test phase) scored via scripts/imminent_convergence_predictor.py --evaluate against public USGS FDSN (bbox containment + magnitude gate, 0.5 tolerance)","expected_direction":"positive","feature_set":["n_converged_models","cross_model_corroboration"],"git_commit_at_registration":"6e09b4471e4a5f37cb5a086d7b66b5575dce0d4d","hypothesis_id":"20260712_v2_imminent_convergence_72h_shad_a3f748c7","locked_at":"2026-07-12T00:18:39Z","locked_file_sha256":"a062f8691ca5028cc392079052c61c00f35354ec9576228ad56d72be054a01e9","notes":"v2 of the imminent-convergence 72h forward test (v1 4ec47bcd withdrawn pre-launch, zero data: Mike directed SHADOW-mode testing before public exposure). Trigger rule unchanged: >=4 DISTINCT models holding active in-window HIGH/CRITICAL EQ forecasts on the same region simultaneously -> ONE 72h shadow record per region per episode (data/imminent_shadow.jsonl, private, email-to-Mike only). PASS: --evaluate resolved hit rate >= 70% at n >= 30 expired shadow windows over the first 90 days. Evaluator is reproducible from the shadow record + public USGS by construction; validated both directions Jul 12 (samoa test window recovers the real M5.1; quiet ukraine window scores miss). Backtest basis (60d, dedup region+day): 77.8% hit-within-72h at 1.4 triggers/day; the >=3+CRITICAL alternative measured ANTI-selective (62.1%) and was dropped pre-launch. UNDERPOWERED if n < 30 by re-eval: extend 90 more days rather than judge early. PROMOTION on PASS: flip _SHADOW_MODE=False so the identical trigger logs scored public ledger predictions. Re-eval 2026-10-11.","ots_proof_path":null,"outcome":{"lookahead_hours":72.0,"min_magnitude":null,"type":"earthquake"},"reevaluate_at":"2026-10-10","registered_at":"2026-07-12T00:18:38Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"resolved_hit_rate","n_permutations":null},"threshold":{"direction":"positive","metric":"resolved_hit_rate_pct","value":70.0},"title":"v2 imminent_convergence_72h SHADOW forward test: >=70% hit rate","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-10-11"},{"data_source":"FDSN continuous vertical-component broadband noise; 2 teleseismic dense clusters (60-90 deg), 1-3 yr; earthquake catalog for ground-truth traveltimes","expected_direction":"positive","feature_set":["tgb_pkikp_resid_s","tgb_corephase_resid_s","tgb_slowness_class_accuracy","tgb_bc_df_s"],"git_commit_at_registration":"271cd1261d8889acfb7d9f15cace8f2a2d2f3013","hypothesis_id":"20260714_passive_deep_earth_body_wave_ret_92167244","locked_at":"2026-07-14T20:12:55Z","locked_file_sha256":"f5ec2eec70af05c1975227f929ab4e0f0af6bd444ac543891044a225f73b2cce","notes":"PRE-REGISTRATION (Springboard #384, HAL-tel-01927711v1, ISTerre Campillo/Boue/Roux lineage). CONFIRMATORY, not blind: the continuous noise exists in FDSN archives DeepMap already ingests, but this double-array slowness analysis has never been run before this lock. HYPOTHESIS: using two teleseismically-separated (60-90 deg) dense station clusters already in DeepMap, (a) stable teleseismic P plus >=1 core phase (PcP or PKP) is retrievable from >=12 months of statistics-weighted ambient-noise cross-correlation across >=30 cross-cluster station pairs, recovered within +/-2 s of the ak135-predicted traveltime on >=50% of pairs; AND (b) >=80% of spurious/early arrivals are correctly classified as source-artifacts vs structurally-real phases by their slowness signature in the double-array Radon (tau-p) domain, validated against earthquake-derived P/PcP/PKP traveltimes through the same clusters as ground truth. PASS = BOTH (a) >=50% pairs within +/-2s AND (b) >=80% real-vs-spurious classification accuracy. NULL (a real, publishable constraint) otherwise. WITHDRAW/underpowered = fewer than 30 cross-cluster pairs converge to a stable body-wave arrival, or the body-wave noise fraction is too low to stack over the window. NOVELTY/FTO: pure deep-phase-from-noise is NOT novel (Boue, Poli, Nishida, Lin, Campillo); the claimable core is the automated slowness-based real-vs-artifact classifier + continuous temporal-change monitoring + Earth-State-Tensor fusion + proof-of-when. Run FTO against Poli/Campillo/Nishida before any filing. This test also hardens the existing inner-core rotation pre-registration by supplying the artifact-rejection layer that distinguishes a genuine BC-DF core signal from seasonal microseism-source migration.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2026-10-12","registered_at":"2026-07-14T20:12:55Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"double-array Radon tau-p noise-correlation body-wave extraction + ak135 slowness real-vs-spurious classifier","n_permutations":null},"threshold":{"direction":"positive","metric":"real_vs_spurious_classification_accuracy","value":0.8},"title":"Passive deep-Earth body-wave retrieval with double-array slowness artifact rejection","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"GOES ABI-L2-ACMC cloud (2 new east-fault boxes, built post-lock) + Open-Meteo pressure","expected_direction":"negative","feature_set":["disentangle_excess_p"],"git_commit_at_registration":"5b3020863a1f1c7c87edc8f8815c93744136e630","hypothesis_id":"20260714_pine_valley_barometric_disentang_bed7b5e8","locked_at":"2026-07-14T23:59:51Z","locked_file_sha256":"5b635ef8df5ce645a18f1ad87c941afa462290af5b07a6019e1abd6567825367","notes":"PRE-REGISTRATION -- GENUINELY BLIND: the two east-fault boxes' (Moab 38.57/-109.55, Lisbon Valley 38.15/-109.30) barometric response has NEVER been measured; their GOES cloud series is built AFTER this lock. Follows the #2 barometric PASS whose west/east confound (escalante flat WEST control tied the fault) leaves fault-venting unproven. TEST: are EAST faults significantly more barometrically sensitive than EAST laccolith non-faults (same region, no west/east alias)? PASS (fault-venting CONFIRMED) = excess = mean(east-fault beta) - mean(east-nonfault beta) < 0 AND block-bootstrap p<=0.05. NULL = the #2 barometric PASS is a west/east REGIONAL confound and fault-specific venting is DISCONFIRMED. The west side already shows no fault effect, so a real venting mechanism must make the east faults stand out. Script sha256: bebcc281a57efee367c2097f16c1f8d7ef90d08c39dd0a3cfdf1731bd8bf226c.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-14","registered_at":"2026-07-14T23:59:51Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"east-fault (Moab, Lisbon) barometric beta vs east-laccolith beta; block-bootstrap of the mean difference","n_permutations":null},"threshold":{"direction":"negative","metric":"disentangle_excess_p","value":0.05},"title":"Pine Valley barometric disentangle: east faults vs east non-faults (fault-venting vs regional confound)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"GOES-West ABI-L2-ACMC cloud_index.db 2017-2025 + Open-Meteo/USGS forcing","expected_direction":"negative","feature_set":["barometric_excess_p"],"git_commit_at_registration":"aba6cf27323e8dfd3e1bf1bd60be278881a7611b","hypothesis_id":"20260714_pine_valley_barometric_pumping_c_5d6c8b00","locked_at":"2026-07-14T22:36:04Z","locked_file_sha256":"f8a9f795fa2901dd112a3db0183b1b6a72412059f238dad76a3c4c330aac7b4f","notes":"PRE-REGISTRATION. CONFIRMATORY (data exists in the cloud/forcing/USGS caches; this analysis has run ZERO times on real data before this lock -- only --selftest ran). PASS = excess (fault beta - mean control beta) < 0 AND block-bootstrap p<=0.05. Low pressure eases fault confinement -> extra venting; discriminator is the fault's EXCESS negative pressure sensitivity beyond the shared weather effect. Script sha256: b4659bfa9067b6f7fb7db8a6bac93f6987e00683bd4f94ed5c21af555be1da79. Part of the Pine Valley release-mechanism series following the M2 null (20260714 ...b26f1d93). NULL is a real, publishable constraint.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-14","registered_at":"2026-07-14T22:36:04Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"partial regression of cloud residual on pressure residual per box; block-bootstrap of fault-minus-mean-control excess","n_permutations":null},"threshold":{"direction":"negative","metric":"barometric_excess_p","value":0.05},"title":"Pine Valley barometric-pumping cloud response (fault exceeds controls)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"GOES-West ABI-L2-ACMC cloud_index.db 2017-2025 + Open-Meteo/USGS forcing","expected_direction":"negative","feature_set":["fault_spring_neap_p"],"git_commit_at_registration":"aba6cf27323e8dfd3e1bf1bd60be278881a7611b","hypothesis_id":"20260714_pine_valley_fortnightly_spring_n_b97d3ad8","locked_at":"2026-07-14T22:36:04Z","locked_file_sha256":"a6c515918bcbf5d4e5b5cda847181be92212f257e7cd7a9bb77ce2b7fb7fdc89","notes":"PRE-REGISTRATION. CONFIRMATORY (data exists in the cloud/forcing/USGS caches; this analysis has run ZERO times on real data before this lock -- only --selftest ran). PASS = fault spring-neap p<=0.05 AND fault spring-neap snr > max control snr. Resolves the M2 null: a threshold-triggered fault release responds at the spring-neap ENVELOPE, not the M2 carrier, so a clean M2 null can coexist with a real fortnightly signal. Script sha256: b5531f369393f36898295aef4b159e657b73db39e37b7047f6a8de448eb5a7eb. Part of the Pine Valley release-mechanism series following the M2 null (20260714 ...b26f1d93). NULL is a real, publishable constraint.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-14","registered_at":"2026-07-14T22:36:04Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"LS spring-neap (14.7653d) + Mf (13.6608d) candidate line vs local null grid; fault-exceeds-controls","n_permutations":null},"threshold":{"direction":"negative","metric":"fault_spring_neap_p","value":0.05},"title":"Pine Valley fortnightly/spring-neap cloud modulation (threshold-tidal fault venting)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"GOES-West ABI-L2-ACMC cloud_index.db 2017-2025 + Open-Meteo/USGS forcing","expected_direction":"negative","feature_set":["seasonal_lagged_excess_p"],"git_commit_at_registration":"aba6cf27323e8dfd3e1bf1bd60be278881a7611b","hypothesis_id":"20260714_pine_valley_seasonal_hydrologic__e82c2420","locked_at":"2026-07-14T22:36:04Z","locked_file_sha256":"67db91f0611d86610039730cfcb1ffd6f4fab31474a7bf3c4eb290c8c936024f","notes":"PRE-REGISTRATION. CONFIRMATORY (data exists in the cloud/forcing/USGS caches; this analysis has run ZERO times on real data before this lock -- only --selftest ran). PASS = fault peak-lagcorr > every control AND phase-randomised p<=0.05. Interannual loading anomaly -> lagged venting anomaly, fault-attributable. WEAKEST of the four (snow_depth is a loading proxy, broad lag band) -- a null here is weaker evidence than #1/#2/#4. Script sha256: c5359b30ab170ca229d73615e7eae0e332ea0357da6d24cfb781adc8df42078f. Part of the Pine Valley release-mechanism series following the M2 null (20260714 ...b26f1d93). NULL is a real, publishable constraint.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-14","registered_at":"2026-07-14T22:36:04Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"peak lagged (7-84d) cross-correlation of cloud anomaly vs regional snow-load anomaly per box; phase-randomised null","n_permutations":null},"threshold":{"direction":"negative","metric":"seasonal_lagged_excess_p","value":0.05},"title":"Pine Valley seasonal hydrologic-loading lagged venting (fault exceeds controls)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"GOES-West ABI-L2-ACMC cloud_index.db 2017-2025 + Open-Meteo/USGS forcing","expected_direction":"negative","feature_set":["seismic_postevent_excess_p"],"git_commit_at_registration":"aba6cf27323e8dfd3e1bf1bd60be278881a7611b","hypothesis_id":"20260714_pine_valley_seismic_dynamic_trig_300290ad","locked_at":"2026-07-14T22:36:04Z","locked_file_sha256":"e5fa0a35b8d0d9a0f14cf6ec64adbab274377d366981c72b912adcecc20fc4e5","notes":"PRE-REGISTRATION. CONFIRMATORY (data exists in the cloud/forcing/USGS caches; this analysis has run ZERO times on real data before this lock -- only --selftest ran). PASS = fault post-event (0-3d) minus pre-event (-7..-1d) excess p<=0.05 AND fault excess > max control excess. Earthquakes transiently boost fault permeability -> a venting burst. Script sha256: 106eb239488fda034ef01db70f7d1b4a0dff032dfd5ea04e31a869395048c810. Part of the Pine Valley release-mechanism series following the M2 null (20260714 ...b26f1d93). NULL is a real, publishable constraint.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-14","registered_at":"2026-07-14T22:36:04Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"superposed-epoch post-minus-pre cloud residual around local M>=2.5 + teleseismic M>=6.5 events; event-time-shuffle null","n_permutations":null},"threshold":{"direction":"negative","metric":"seismic_postevent_excess_p","value":0.05},"title":"Pine Valley seismic dynamic-triggering venting burst (superposed-epoch, fault exceeds controls)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"data/fault_energy_state.db stand_downs + USGS FDSN","expected_direction":"negative","feature_set":["fes_energy_state","fes_recent_stand_down","fes_arming_dvv_pct"],"git_commit_at_registration":"20483c9009a0380443b34f3fd88cc3d700a9c755","hypothesis_id":"20260716_fault_energy_state_stand_down_de_25e55634","locked_at":"2026-07-16T17:12:51Z","locked_file_sha256":"02eec6a0b41eed3841903f1e520e115d0f765db56c6dd208117005acffbd8996","notes":"Tracker flags a watched region STOOD_DOWN when its arming dv/v anomaly recovers toward baseline (recover threshold -0.15pct) WITHOUT a qualifying M>=4.5 in-region quake during the armed window. Hypothesis: de-energization is real -> STOOD_DOWN regions have a LOWER M4.5+ rate in the 30d after stand-down than regions that remained HELD (still armed) over a matched 30d window; i.e. rate(stood_down) - rate(held) < 0. PASS: difference < 0, one-sided p<0.05, n>=20 stand-down events. WITHDRAW if <20 stand-downs accrue by re-eval 2026-10-16. This validates the standing-alert downgrade: genuinely de-energized regions should produce fewer subsequent quakes than still-loaded ones.","ots_proof_path":"hypotheses/locked/20260716_fault_energy_state_stand_down_de_25e55634.json.ots","outcome":{"lookahead_hours":720.0,"min_magnitude":4.5,"type":"earthquake"},"reevaluate_at":"2026-10-16","registered_at":"2026-07-16T17:12:30Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"one_sided_rate_comparison","n_permutations":null},"threshold":{"direction":"negative","metric":"stood_down_minus_held_M4.5_rate_30d","value":0.0},"title":"Fault energy-state stand-down: de-energized regions have a lower subsequent EQ rate than held regions","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-10-16"},{"data_source":"cross_physics/rs_etas_caldera_forecaster.py forecast_at() with live GPS/BOTPT (NGL IGS20 + OOI BOTPT); prediction ledger HIGH/CRITICAL rows at the 5 watch-list calderas; USGS + INGV ISIDe catalogs","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"3a88fc3e5db8aeff623876a6d3fb932df9fdc865","hypothesis_id":"20260718_v2_rs_etas_caldera_forecaster_on_4fdfbd38","locked_at":"2026-07-18T16:28:37Z","locked_file_sha256":"5f50b723a0b34168f8cf1a08a632fe6bed1e8fe84e86f0c499e8ed6ed7a66f2d","notes":"Supersedes 20260506_rs_etas_caldera_forecaster_high__eb9944b0, withdrawn as an INVALID test: the NGL IGS14 URL was dead for that entire window so every alert rode a frozen 50 mm/yr literature prior (rs_amp 8.2652 constant > CRITICAL threshold 5.0 by arithmetic). Fixed 2026-07-18 (NGL IGS20 repoint) plus a fail-closed rule that degrades HIGH/CRITICAL to ADVISORY whenever uplift_source is a literature prior, so a frozen input can never again reach the scored ledger. PASS requires ALL of: (1) every scored alert carries a LIVE uplift_source (not literature_prior); (2) a qualifying match is M>=3.0 -- the Jul 2026 audit showed the INGV matcher at minmag=1.5 credits routine bradyseismic M1.5-2.9 unrest, which inflated the old number to ~91 pct while M>=3.0 gave ~30 pct and M>=4.0 ~8 pct; (3) the pooled HIGH/CRITICAL match rate exceeds the COVERAGE NULL -- what a perpetual HIGH watch on the same caldera footprints over the same window would have caught -- by >= 10 percentage points. Beating an always-on watch is the bar because four of five claims refuted in the Jul 18 audit failed exactly there. UNDERPOWERED WITHDRAW: fewer than 10 scored HIGH/CRITICAL alerts across the 5 calderas in the window; with the fail-closed rule live and Campi now reading NOMINAL, a low alert count is a plausible and honest outcome. A null is publishable: it would say deformation-driven RS+ETAS adds nothing over coverage at these calderas.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":3.0,"type":"earthquake"},"reevaluate_at":"2026-10-16","registered_at":"2026-07-18T16:28:27Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"forward_match_rate_vs_coverage_null","n_permutations":null},"threshold":{"direction":"positive","metric":"match_rate_minus_coverage_null_pp","value":10.0},"title":"v2 RS+ETAS caldera forecaster on LIVE deformation beats the coverage null","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"prediction_ledger shadow bands + USGS/multi-catalog for b,Mc + ETAS conditional rate","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"8495b259b1670edee43073f17045f73495530b01","hypothesis_id":"20260723_v2_dynamic_magnitude_evaluator_p_d9bd832d","locked_at":"2026-07-23T15:31:59Z","locked_file_sha256":"87515824b7c5362cc7b4e022e5dc5835a27a2e7476d4e3330f2c3ac8adff9ab7","notes":"DYNAMIC MAGNITUDE EVALUATOR v2 -- shadow test. Supersedes\n20260723_dynamic_magnitude_evaluator_pois_ed6a2456, withdrawn as an INVALID test\n(object mismatch + unit-of-analysis error), not a failed one, before it ran.\n\nWHAT CHANGED FROM v1, AND ONLY THIS\nv1 scored \"fraction of matched OBSERVATIONS inside the band\" while the formula\nP(Mmax < m) is the CDF of the LARGEST event in the window. Measured on 120 days:\n12.7 observations per matched prediction (max 542), only 21% with a single\nobservation, magnitudes spanning M2.5-M8.1 WITHIN one prediction. A\nmax-distribution band scored against every aftershock must under-cover, and\nper-observation scoring is separately a unit-of-analysis error -- 12.7 correlated\naftershocks are not 12.7 independent tests. v2 scores ONE value per prediction:\nthe MAXIMUM observed magnitude. Everything else is carried over verbatim.\n\nWHAT IS BEING TESTED\nThe shipped band is RISK_LEVEL_BUCKETS[tier] -- a lookup keyed on a LABEL --\nplus five corrections (Bath cap, tectonic ceiling, region floor, calibrated\nfloor, log-window nudge). Event RATE never enters it, so a region producing 40\nevents a month and one producing 2 receive the same band from the same tier. The\nwindow nudge (1/b)*log10(T/168h) is identically zero at a 7-day window for ANY b,\nso the b-value cannot act on most predictions at all.\n\nThe candidate uses the closed form for the largest event of a\nPoisson-Gutenberg-Richter process over the forecast window:\n    P(Mmax < m) = exp( -lambda*T * 10^(-b(m-Mc)) )\n    m_q         = Mc + (1/b) * log10( lambda*T / -ln q )\nBand = [m_0.10, m_0.90], capped at the region's Mmax and floored at Mc. This\nsubsumes all five patches: the floor IS Mc, the window enters as lambda*T, the\ntail shape IS b, and b acts on every prediction rather than only on non-7-day\nwindows.\n\nPARAMETERS -- which move is decided by our own prior results, not preference\n  lambda (rate): DYNAMIC. The one component with validated skill -- ETAS beat the\n      long-run baseline on CSEP's own yardstick, replicated across japan\n      (+0.2947) and indonesia (+0.1922), sign-test p=0.0063. Fallback: long-run\n      Poisson rate, recorded as such.\n  b, Mc: LONG-BASELINE (>= 4 years, quarterly refresh), NOT time-varying. Three\n      independent tests say reactive size-distribution parameters are worse:\n      adaptive quantile bands 79.2% -> 75.7% OOS (rejected); recency b 1.34 ->\n      1.49 (noisier, under-predicts large events); tensor-conditioned b\n      pre-registered and FAILED 0 of 3 regions, 0 of 9 configs, indonesia\n      significantly negative (20260719_tensor_adds_magnitude_informatio_3b3bf3bc).\n      b by Aki-Utsu MLE, Mc by max-curvature + 0.2, >= 150 events above Mc, else\n      the global b = 1.0 with the region flagged.\n  T: the forecast's own valid window.\n\nDESIGN\nSHADOW. Both bands are computed and recorded for every new earthquake\nprediction; NEITHER is shipped and the public band is unchanged throughout.\nScoring is PAIRED -- the same predictions carry both bands -- so the comparison\ncannot be confounded by which predictions happened to match.\n\nPRIMARY METRIC\nFor each matched prediction take Mmax_obs = the largest observed magnitude\ncredited to it. Coverage = fraction of matched predictions with\nband_min <= Mmax_obs <= band_max. The band is an 80% band, so the target is 0.80\nand the quantity of interest is |coverage - 0.80|.\n\nPASS  = |coverage_dynamic - 0.80| < |coverage_current - 0.80|\n        AND mean(max-min) for the dynamic band is not more than 10% wider than\n        the current band on the same predictions.\n        (The width guard exists because coverage is trivially bought by widening.)\nFAIL  = anything else.\nUNINFORMATIVE = if |coverage_current - 0.80| <= 0.02 at re-evaluation the\n        incumbent is already calibrated and the test cannot resolve. A power\n        condition fixed in advance, not an outcome-dependent escape.\nWITHDRAW = fewer than 300 matched shadow predictions at re-evaluation.\n\nSECONDARY, reported but never decisive: all-observations coverage (the v1\nmetric, retained for continuity), below-floor and above-ceiling rates,\nper-region coverage, mean width by tier.\n\nEXPECTED BEHAVIOURAL CONSEQUENCE, stated in advance so it is not discovered as\na surprise: the dynamic band takes no tier, so a HIGH and a CRITICAL forecast\nfor the same region and window receive the SAME band. That is defensible --\nrisk_level measured AUC 0.579, barely above chance, so the tier carries almost\nno magnitude information -- but it is a real change and is recorded here as\npredicted, not rationalised afterwards.\n\nHONEST LIMITS, fixed in advance\n1. Coverage is measurable only on predictions that MATCHED an event. This\n   measures band calibration GIVEN an event occurred; it is not forecast skill\n   and must not be reported as such.\n2. Mc is a network-detection property and moves with station coverage, so it is\n   frozen at its locked value for the duration.\n3. The evaluator conforms to the formula and parameters above; this locked\n   criterion governs. Its sha256 is recorded when it lands.\n4. Current shipped-band coverage under the v2 metric was NOT measured before\n   locking, so the bar could not be fitted to it.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":2.5,"type":"earthquake"},"reevaluate_at":"2026-10-21","registered_at":"2026-07-23T15:31:36Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"paired shadow comparison of |coverage-0.80| on max observed magnitude per prediction, with a band-width guard","n_permutations":null},"threshold":{"direction":"positive","metric":"abs_coverage_error_improvement","value":0.0},"title":"v2 Dynamic magnitude evaluator (Poisson-Gutenberg-Richter band) beats the tier-lookup band on out-of-sample calibration of the largest observed event","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"IRISPH5 network 4A (648 nodes, 250 Hz, ~100 km aperture); answer key Wu et al. 2023","expected_direction":"negative","feature_set":["ambient_noise_rayleigh_dispersion_vsz"],"git_commit_at_registration":"8449a3c27a3d57f6956f2387e6dbda403ae9cd85","hypothesis_id":"20260724_yellowstone_4a_answer_keyed_vs_r_20348cda","locked_at":"2026-07-24T19:58:14Z","locked_file_sha256":"1335df77d72ed17fe031ce62f1dd75f38bc3859bda7e8eb442d96532ee3e8dd0","notes":"YELLOWSTONE 4A ANSWER-KEYED RECOVERY + dense-array DECIMATION CURVE. Pre-registered before our passive stack is run on the 4A data, so the recovery cannot be tuned to match the published answer.\n\nPURPOSE. PH5 network 4A (\"Imaging Yellowstone's shallow Magma Reservoir\", 648 three-component nodes, 250 Hz, ~100 km aperture, centroid 44.56N 110.57W, deployed 2020) is a free public DENSE ARRAY over a target with a PUBLISHED structure. This is a METHOD-VALIDATION test (the answer exists; we test whether OUR passive dense-array stack reproduces it), NOT a discovery. Its value is (i) a credibility artifact -- \"our stack recovers the published magma-reservoir structure on public dense data\" -- and (ii) the dense-array/Gen-3 deliverable below.\n\nANSWER KEY (pinned now). Wu et al. 2023 (Yellowstone shallow-magma nodal imaging, surfaced by the Phase-2 PH5 eligibility filter): an exceptionally low shear-velocity zone with >20% Vs reduction at 4-7 km depth (magma-reservoir top / sill-rich zone near ~5 km), with ~20% radial anisotropy (SH faster than SV) consistent with stacked horizontal sills.\n\nMETHOD (our stack). Ambient-noise cross-correlation across 4A station pairs -> Rayleigh-wave group/phase dispersion -> 1D/pseudo-2D Vs(z) inversion beneath the array centroid. The 5-Hz geophone nodes roll off below ~5 Hz, so resolving 4-7 km needs the array's long inter-station paths at the microseism/long-period band; whether our passive stack can reach that depth with these sensors is itself part of what this tests.\n\nPRIMARY TEST (answer-key recovery). Our Vs(z) beneath the 4A footprint recovers a low-velocity zone with >=15% Vs reduction (relative to the overlying crustal Vs) at a depth within [3, 8] km. CONFIRM if recovered; FAIL if not. A FAIL is a real and useful result: it bounds our passive stack's depth reach with 5-Hz nodal data (a Gen-3 sensor/aperture design finding), and must be reported as such, not buried.\n\nSECONDARY (descriptive, not gating). (a) Radial anisotropy sign SH>SV if 3-component dispersion permits its estimation. (b) THE dense-array DECIMATION CURVE: re-run the SAME recovery on random node subsets of size 648, 324, 162, 81, 40, 20 (5 random draws each), and report the smallest node count / largest mean inter-node spacing at which the LVZ is still recovered at >=15% within [3,8] km. This directly answers the RattleSnake / dense-array Gen-3 array-design question (\"how many nodes at what spacing to resolve a target at depth D\") on data with a KNOWN answer, and is the most transferable deliverable of the experiment.\n\nWITHDRAW GATES (coverage/feasibility, decided before any outcome). (a) >=100 4A nodes deliver >=20 continuous days of usable data (else the ambient stack is under-sampled -> report data-limited, do not force a recovery). (b) inter-station paths span >=30 km (needed for the long-period dispersion that senses 4-7 km); the ~100 km aperture satisfies this by construction, confirmed at run start.\n\nRUNNER. The reprocess runs BLIND after this lock is committed/anchored; no inversion parameter is tuned after comparing to the answer key. Both a clean recovery and an honest depth-reach null are valid, reportable outcomes. The decimation curve is computed regardless of the primary verdict.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"structure_recovery"},"reevaluate_at":"2026-10-24","registered_at":"2026-07-24T19:58:14Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"ambient_noise_vsz_recovery_of_published_LVZ + node-decimation curve","n_permutations":null},"threshold":{"direction":"negative","metric":"vs_reduction_pct_in_3_8km","value":15.0},"title":"Yellowstone 4A answer-keyed Vs recovery + dense-array decimation curve (method validation on public dense array)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"telluric_current_monitor geoelectric-field z per region + USGS M>=5.0 events + regional 72h base rates","expected_direction":"positive","feature_set":["telluric_geoelectric_field_anomaly_z"],"git_commit_at_registration":"5b3ee4c426f87a05d382ea7ee0d3d3899b0f31a4","hypothesis_id":"20260804_telluric_geoelectric_anomaly_as__c7b40244","locked_at":"2026-08-04T19:49:00Z","locked_file_sha256":"4dcbed3893558c2a09e9ba32604817b4047a5848bddba76be0e1dc11357d0aa0","notes":"Measurable version of the electricity-shapes-geology thread (Isaac): does a telluric/geoelectric anomaly precede M5.0+ EQs beyond the regional base rate? Legitimate seismo-EM / VAN SES territory, NOT Electric-Universe cosmology. SHADOW forward test on the existing telluric_current_monitor + prediction ledger. TWO mandatory nulls (shuffle signal-removed + base-rate climatology) plus Kp-storm / cultural-EM / solar-cycle confound locks. null > false data: a null is a real seismo-EM contribution, a positive is a genuine contested-field result. Re-eval 2027-02-04.","ots_proof_path":"hypotheses/locked/20260804_telluric_geoelectric_anomaly_as__c7b40244.json.ots","outcome":{"lookahead_hours":72.0,"min_magnitude":5.0,"type":"earthquake"},"reevaluate_at":null,"registered_at":"2026-08-04T19:49:00Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{"baseline":"regional 72h base rate of M>=5.0 (climatology control)","basis":"seismo-electromagnetics / VAN SES + Freund stress-activated currents -- the MEASURABLE version of electricity-shapes/triggers-geology, NOT Electric-Universe cosmology","cascade_rung":"72h telluric SHADOW notice","confound_locks":["exclude Kp>=5 storm epochs (telluric spikes in storms independent of EQs)","exclude cultural-EM-flagged stations","F10.7 solar-cycle-matched control"],"event_set":"M>=5.0 inside the monitored region bbox within 72h of a telluric-z shadow notice","forward_window":["2026-08-04","2027-02-04"],"mode":"SHADOW: advisories only, no user-facing alerts, until calibrated","nulls":["shuffle: identical model with the telluric signal removed","regional base-rate climatology control"],"oos":"threshold theta selected on pre-seal history (date-split), scored on the post-seal forward window","pass_bar":"OOS notice-precision lift over the regional base rate >= 1.3 AND survives the shuffle null (signal-removed same model consistent with lift ~1.0)","reevaluate_at":"2027-02-04"},"method":"sealed_stacked_epoch_shadow_skill","n_permutations":null},"threshold":{"direction":"positive","metric":"notice_precision_lift_over_baserate_shuffle_gated","value":1.3},"title":"Telluric geoelectric anomaly as an M5.0+ earthquake precursor: skill over regional base rate (forward held-out, SHADOW)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2027-02-04"},{"data_source":"gsn_microseism_hourly.db 16-yr GSN hourly secondary-band (0.1-0.4 Hz) envelope archive; stations >= 8 yr coverage; per-station solid-earth-tide and source-region ocean-tide phase","expected_direction":"positive","feature_set":["microseism_band_rms_secondary"],"git_commit_at_registration":"2adfaa780bbbdeff8def1b259fe9e9ae77841022","hypothesis_id":"20260807_crustal_nonlinear_elasticity_via_867f46d6","locked_at":"2026-08-07T02:22:09Z","locked_file_sha256":"3a0e64739a2139829f6c67eba6c520ec82d5bb5977d077fb2eee9c78901583fb","notes":"SEALED PRE-REGISTRATION OF RECORD: strategy/crustal_nonlinearity_intermod_prereg_v1_SEALED.md, SHA-256 3b7900adab036b9c4394b831905a2751e4d21042d4c7a631081b3ffe5101eed8 (drafted + hashed 2026-08-06 in the CoSMoS build wave; Mike gave the word to anchor 2026-08-06; sealed verbatim -- the sealed document is byte-identical to the approved draft, so the approved hash IS the sealed hash). Working copy: _active_today/Crustal_Nonlinearity_Intermod_Prereg_v1.md (identical bytes). BINDING SPEC in the sealed doc: FROZEN criteria (1) M2 line p_LEE < 1e-3 vs circular-shift null >=1000 draws over the frozen 10-14 h period scan window; (2) O1 also present p < 0.01 with m_O1/m_M2 within factor 3 of the theoretical tidal-stress ratio; (3) split-halves: even/odd-longitude AND first/second-half years, all four sub-splits show M2 at p < 0.05; (4) confound gates ALL excluded before a positive: (a) direct tidal ground-motion leakage (survive <0.05 Hz notch before envelope), (b) ocean-source-vs-crust discriminator (receiver-side solid-earth-tide phase clustering must beat source-region ocean-tide alignment, p < 0.05, else report honestly as an ocean-source result), (c) S2/K1 thermally EXCLUDED from all pass criteria (computed for transparency only -- lunar constituents have no thermal counterpart). Default outcome NULL = frozen upper bound on m_M2, the first network-scale in-situ NEWS (nonlinear elastic wave spectroscopy) bound of its kind. SCOPE DISCIPLINE: a MEASUREMENT of crustal nonlinear elasticity, NOT an earthquake prediction product, and will not be framed as one regardless of outcome; any forecasting implication requires its own separate pre-registration. Analysis seed = the OTS block hash of this anchor; no tuning. null > false data.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geophysical_measurement"},"reevaluate_at":"2026-11-06","registered_at":"2026-08-07T02:22:00Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"station_stacked_log_envelope_harmonic_fit_M2_O1_vs_circular_shift_null_with_receiver_vs_source_phase_discriminator","n_permutations":1000},"threshold":{"direction":"positive","metric":"M2_line_p_LEE","value":0.001},"title":"Crustal nonlinear elasticity via tide x microseism intermodulation: M2/O1 envelope sidebands in the 16-yr GSN stack (CoSMoS)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"data/epc_shadow_predictions.db (shadow only; frozen regions hypotheses/epc_reeval_v1_regions.json sha256 8f3b5918d854fb1f2bc9d5ba7060b2ce3dfafb962ec4a0113b6830efc4d5cca8)","expected_direction":"positive","feature_set":["epc_n_active_precursors","epc_b_value","epc_max_expected_magnitude","epc_data_real"],"git_commit_at_registration":"7b89b1479026d52c2cb2fa47b36be68b011eb635","hypothesis_id":"20260807_epc_re_evaluation_v1_the_fixed_e_e3b22cd2","locked_at":"2026-08-07T22:42:36Z","locked_file_sha256":"6c8f6b0f1df211a7c6a6c5e45a88b30e9be4f5c7c908a4355c8eca0ed83471ce","notes":"Prereg doc hypotheses/epc_reevaluation_prereg_v1.md. Denylisted 2026-05-15 on evidence the BSD/HVSR audit Sec 9.4 showed was corrupted (99/117 rows carried an un-provenanced telluric stage; 8 exist only because of it). No retroactive re-crediting; EPC stays denylisted throughout. HONEST CAVEAT: Stage 4b telluric and Stage 5 TEC are both structurally dead on prod (data_real 0.0) so this is a 5-stage chain, not the chain as designed.","ots_proof_path":"hypotheses/locked/20260807_epc_re_evaluation_v1_the_fixed_e_e3b22cd2.json.ots","outcome":{"lookahead_hours":168.0,"min_magnitude":3.5,"type":"earthquake"},"reevaluate_at":"2027-02-03","registered_at":"2026-08-07T22:41:57Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"climatology_b":1.0,"fire_min_active":3,"max_days":180,"max_p_vs_climatology":0.05,"max_p_vs_shuffle":0.05,"min_days":90,"min_hit_rate_pct":15.0,"min_resolved":100,"shuffle_seed":20260807,"void_if_revived_stage_fraction_gt":0.05},"method":"shadow_forward_full_magnitude_vs_poisson_binomial_climatology_and_date_shuffle","n_permutations":10000},"threshold":{"direction":"positive","metric":"full_magnitude_hit_rate_pct","value":15.0},"title":"EPC re-evaluation v1: the fixed earthquake_precursor_chain beats per-region GR climatology AND its own date-shuffle null in shadow mode","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"USGS NWIS continuous water levels + WQP fluid-at-screen labels + well construction; IGETS SG; EIA UGS working-gas inventory; USArray/EMTF MT; PBO/EarthScope borehole strain + pore pressure; NOTA GNSS; FDSN Vp/Vs + event times; published He-isotope + BE/TE compilations","expected_direction":"positive","feature_set":["tidal_wl_admittance_amp_phase_M2_O1_fullband","drained_undrained_plateaus_fc_ratio_phase_slope","barometric_efficiency_BE_f","tidal_efficiency_TE","eta_BE_over_TE_dphi_baro_tide"],"git_commit_at_registration":"59780ea08c7c276965bb6934b668df4adc38fcf0","hypothesis_id":"20260807_tidal_poroelastic_pumping_for_pa_d1fbba3f","locked_at":"2026-08-07T02:38:26Z","locked_file_sha256":"0c33596fcefa8f138e87f4cb737d72d09aa5f6b268b84a433a0684a10a133c74","notes":"SEALED PRE-REGISTRATION OF RECORD: strategy/fluid_discrimination_prereg_v2_1_SEALED.md, SHA-256 fcf6645a51958d6e13c00f422ca72f9b9b756912398464fd28841589ce29e828 (anchored bytes). Working copy: _active_today/Fluid_Discrimination_Prereg_v2.1.md (identical bytes). HONESTY RECORD / HASH SUPERSESSION: the crashed 2026-08-06 session recorded an earlier SHA-256 for this document beginning 8279163e; that hash was NEVER anchored. The bytes changed between that hash and anchoring -- most plausibly when the Phase-0b real-NWIS preliminary PASS section (adjusted AUC 0.745, n=26, within-stratum/covariate-adjusted criterion) was appended to the document AFTER it was hashed. The original 8279163e bytes are not recoverable. Mike reviewed the situation and EXPLICITLY APPROVED anchoring the CURRENT bytes (2026-08-07). Superseded pre-anchor hash: 8279163e (never anchored). Anchored hash: fcf6645a51958d6e13c00f422ca72f9b9b756912398464fd28841589ce29e828. FROZEN CRITERION SUMMARY: tidal-poroelastic fluid discrimination -- Earth+barometric tides as free poroelastic pumps; strain-normalized tidal+barometric admittance vector (<=15 z-scored features, logistic/LDA only) classifies GAS-charged vs LIQUID-saturated ground. H1 passes ONLY if ALL hold on field/aquifer-ID-blocked leave-one-basin-out splits: AUC_fluid >= 0.80 overall; within-stratum/covariate-adjusted AUC >= 0.75 with confinement + screen depth partialled out as CONTINUOUS covariates (residualization + propensity matching) AND exceeding the within-stratum label-shuffle null by >= 0.15 with bootstrap CI excluding that null (NOT a fixed AUC-margin vs a confinement classifier -- Phase 0 proved the raw margin saturates and false-negatives under strong confounding); beats lithology+depth+Vs baseline on same matched splits; AUC_joint - AUC_best_single_channel >= 0.10; construction-audit AUC < 0.60. Any one condition fails -> H1 is a NULL, reported with achieved numbers. PHASE 0 KILL-GATE RESULTS (pre-seal, recorded in the sealed doc): 0a synthetic recovery (2026-08-05) PASS -- gas-vs-liquid AUC 1.000 from WL admittance alone, liquid 3-way typing 1.00 with EM channel, and the corrected within-stratum criterion cleanly separated real signal (AUC 1.00) from confound-only (0.67). 0b REAL-DATA confound pilot (2026-08-06) PASS preliminary/underpowered -- NWIS admittance on 32 wells (SNR up to 4.1), WQP labels on n=26 (15 fresh/7 brackish/4 saline): AUC_fluid 0.655, AUC_depth confound 0.636, within-stratum depth-adjusted AUC 0.745 -> fluid signal beyond the confinement confound, kill-gate PASSED; n=26 underpowered, wide CI -- GREEN LIGHT not the powered anchor; NJ coastal-plain fresh->saline gradient expansion (target >=15 saline + >=15 fresh matched, adjusted-AUC lower-CI > 0.62) is the queued powered confirmation. Re-evaluate 2026-11-06. null > false data.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geophysical_measurement"},"reevaluate_at":"2026-11-06","registered_at":"2026-08-07T02:38:22Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"within_stratum_covariate_adjusted_AUC_vs_within_stratum_label_shuffle_null_on_field_aquifer_blocked_leave_one_basin_out_splits","n_permutations":null},"threshold":{"direction":"positive","metric":"within_stratum_covariate_adjusted_fluid_AUC","value":0.75},"title":"Tidal poroelastic pumping for passive fluid-type discrimination (prereg v2.1): within-stratum covariate-adjusted fluid separation vs within-stratum shuffle null","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-11-06"},{"data_source":"IGETS Level-2 superconducting-gravimeter residuals (tides, ocean loading, atmosphere, polar motion, drift removed by the standard pipeline); stations >= sealed coverage; band 1e-8..0.1 Hz","expected_direction":"positive","feature_set":["igets_sg_residual_spectra"],"git_commit_at_registration":"2bbc6237f71342dcbd0c08b1fafd1569f3866bf4","hypothesis_id":"20260807_ultralight_scalar_dm_oscillating_4d41c614","locked_at":"2026-08-07T02:30:52Z","locked_file_sha256":"9f8a348dedb7cc62d1de3c391401d80585e1ea5a9bd6a14220802d8c2172175f","notes":"SEALED PRE-REGISTRATION OF RECORD: strategy/dm_oscillating_g_prereg_v1_SEALED.md, SHA-256 6affe3071d05e9bad7e0fe06fd3f2c24ef0d551d17a9a61b7eae0470c313ab86 (drafted + hashed 2026-08-06; hash re-verified this session against the recorded 6affe307 prefix per Mike's standing verify-then-proceed intent; sealed verbatim -- the sealed document is byte-identical to the approved draft, so the approved hash IS the sealed hash). Working copy: _active_today/DM_Oscillating_G_Prereg_v1.md (identical bytes). BINDING SPEC in the sealed doc: FROZEN candidate criteria, ALL THREE required: (1) global look-elsewhere-corrected narrow-line significance p_LEE < 1e-3 over the 1e-8..0.1 Hz band (time-scramble / station-phase-shuffle null, >= 1e4 draws); (2) cross-station phase coherence z > sealed value matching the DM plane-wave model, vs the shuffled-inter-station-phase coherence null; (3) DM velocity-modulation fingerprint: sideband triplet f, f +/- 1/yr (annual orbital) and f +/- 1/sidereal-day (rotational) at exactly known spacings, matched in amplitude AND phase (p < 0.05), with demodulation collapse yielding coherent SNR gain (CoSMoS house modulation-fingerprint method); a bare monochromatic systematic line does not collapse. Multi-axis coherent stack over station x sideband x annual-cycle x splitting component. VERIFICATION GAUNTLET mandatory before ANY detection claim: systematic audit + fingerprint match + independent-instrument corroboration + blind replication on held-out stations AND held-out epochs; a candidate surviving all four is the only claimable detection; the sealed prereg + candidate timestamp establish priority, the claim waits. Default outcome NULL = per-frequency upper-limit curve on the DM-matter coupling across 1e-8..0.1 Hz, a real limits result extending the certified DM-gravimeter v2 null (strain A_UL ~3e-13 @ 5.82 mHz). Phase 1 is data-only (zero new hardware); Phase 3 resonant-coupler moonshot gated on a Phase-1 candidate. Analysis thresholds FIXED here and do not move post hoc. null > false data; at a positive, the gauntlet before any claim.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geophysical_measurement"},"reevaluate_at":"2026-11-06","registered_at":"2026-08-07T02:30:49Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"coherent_network_stack_plane_wave_phase_LEE_band_scan_with_cross_station_coherence_and_sideband_triplet_fingerprint_nulls","n_permutations":10000},"threshold":{"direction":"positive","metric":"global_line_p_LEE","value":0.001},"title":"Ultralight scalar-DM oscillating-coupling search in the SG network: coherent global stack + velocity-modulation fingerprint (CoSMoS)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"data/prediction_ledger.db reactive-model predictions; C4 via scripts/matched_null_reactive_models.py; RJ-1989 + Wells-Coppersmith 1994 parameters FIXED not fitted","expected_direction":"negative","feature_set":["observed_full_mag_rate","c4_etas_bracket_lo","c4_etas_bracket_hi","count_calibration_ratio"],"git_commit_at_registration":"7ab946cf024e213677a9aaa101854606cc5efb10","hypothesis_id":"20260808_etas_underperformance_v1_do_our__0448fa93","locked_at":"2026-08-08T22:19:02Z","locked_file_sha256":"1e931362e22e1bc4084e669e53c799d5426a74dbeb1642de7b62dc79ea2adc39","notes":"Forward-only. CALIBRATION GATE RUNS FIRST: if the reference count ratio is outside [0.5,2.0] for BOTH bracket bounds the verdict is REFERENCE_MISCALIBRATED and NO claim about the module is made -- the test can void itself by design. Otherwise: below the bracket LOWER bound at p<=0.01 = MODULE_UNDERPERFORMS (actionable engineering target); inside = INDISTINGUISHABLE_FROM_ETAS; above UPPER at p<=0.01 = BEATS_ETAS, which needs its own replication before any public claim. n>=200 per model else UNDERPOWERED, a fail not an extension.","ots_proof_path":"hypotheses/locked/20260808_etas_underperformance_v1_do_our__0448fa93.json.ots","outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"earthquake"},"reevaluate_at":"2026-12-06","registered_at":"2026-08-08T22:18:37Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{"bracket_lower":"causal_gardner_knopoff_parents","bracket_upper":"aggregate_rj_productivity","calibration_gate_ratio":[0.5,2.0],"etas_a":-1.67,"etas_b":0.91,"etas_c":0.05,"etas_p":1.08,"hard_stop_days":365,"min_n":200,"null_min_mainshock_mag":2.0,"p_max":0.01,"prereg_doc":"hypotheses/etas_underperformance_prereg_v1.md","trigger_buffer_km":300,"trigger_catalog_days":180},"method":"poisson_binomial_vs_etas_bracket","n_permutations":2000},"threshold":{"direction":"negative","metric":"observed_minus_bracket_lower_pct","value":0.0},"title":"ETAS underperformance v1: do our aftershock modules score below their own reference ETAS?","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"data/prediction_ledger.db predictions.risk_score + observations; frozen regions hypotheses/risk_delta_v1_regions.json sha256 ba96199974f4ee1b6a6af4af32dbacbd2a34144fdb65a373aad9bdec80841466 (935 regions, 21 never-moving + whole-earth excluded); contaminated predictions (critical/warning provenance notes) excluded from both arms","expected_direction":"positive","feature_set":["risk_score_delta_72h","risk_score_level"],"git_commit_at_registration":"9fc12b1ad876579903c8cdf5bd73f0465ef4e4ee","hypothesis_id":"20260808_risk_score_delta_carries_earthqu_95c82071","locked_at":"2026-08-08T02:59:20Z","locked_file_sha256":"d1e2c001aa9254102f8f1d75e5cd110d71828770f39feb351af0cc0fa09b382a","notes":"Forward-only: events AFTER the lock timestamp. The 180d exploratory window that motivated this is excluded and cannot contribute. PASS requires ALL FOUR: n>=300 isolated events, diff>=+0.50 points, p<=0.01 vs C1 AND C3, not worse than C2 (ETAS). 3 of 4 is a FAIL. Threshold 0.01 not 0.05 because the in-sample result sat on the Bonferroni line across three exploratory tests -- a replication bar should be stricter than the search that found it. A fail is published exactly as a pass would be.","ots_proof_path":"hypotheses/locked/20260808_risk_score_delta_carries_earthqu_95c82071.json.ots","outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"earthquake"},"reevaluate_at":"2026-12-06","registered_at":"2026-08-08T02:59:05Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{"controls":["C1_matched_random_times","C2_etas_operational_alternative","C3_date_shuffle"],"controls_per_event":3,"decluster_h":168.0,"hard_stop_days":365,"impl":"scripts/does_the_score_move.py --isolated-only","isolation":"no in-region event within 72h","lookback_h":72.0,"min_n_events":300,"not_worse_than":"C2_etas","p_max_vs_c1":0.01,"p_max_vs_c3":0.01,"prereg_doc":"hypotheses/risk_delta_timing_prereg_v1.md","seed":20260807,"underpowered_is_fail":true},"method":"matched_control_permutation_delta","n_permutations":5000},"threshold":{"direction":"positive","metric":"delta_diff_score_points","value":0.5},"title":"Risk-score DELTA carries earthquake timing information (v1)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"earth_state_tensor.db state_snapshots + feature_manifests; prediction_ledger.db tensor_eq_per_region","expected_direction":"positive","feature_set":["sc_* (26)","regional_microseism_* (9)","ctd_* (9)"],"git_commit_at_registration":"94ee350186a1d49b89b1b1e19f4a70eb50e1ec16","hypothesis_id":"20260811_v2_forward_array_stacking_tensor_f65b6d4d","locked_at":"2026-08-11T17:36:31Z","locked_file_sha256":"6ca42a5e4a8d6dd54332992b3d52bc738da319623ca0a288d0a0337fbaf32f8c","notes":"FORWARD TEST v2, re-issued 2026-08-11 after the v1 lock (2026-06-01) proved UNTESTABLE -- not disconfirmed. v1 could not be graded because the tensor archive carries no layout manifest before 2026-08-08, so the feature vectors at the prediction epochs were unreadable, and because no prediction inside the manifest window had expired.\n\nEVALUATION IS FIXED IN ADVANCE:\n  harness   scripts/eq_tensor_ablation.py at commit 8853cc7c (or a later commit\n            whose --self-test still passes unchanged)\n  eligible  ONLY predictions whose tensor_epoch matches, within 3600 s, a\n            snapshot carrying a layout_sha resolvable in feature_manifests.\n            Unlabelled snapshots are REFUSED, never aligned by width: 230\n            distinct widths exist and two different name lists can share a\n            length.\n  labels    ONLY predictions whose valid_until has passed. An unexpired,\n            unmatched prediction is UNDECIDED and is skipped, never scored 0.\n  model     GradientBoostingClassifier(n_estimators=150, max_depth=5,\n            subsample=0.8), per region, min 30 eligible predictions/region.\n  split     TEMPORAL, first 70% train / last 30% test. Never random: intra-\n            region predictions are dense and autocorrelated, so a random split\n            leaks future into past and inflates both arms.\n  arms      identical split, identical seed, identical rows; ONLY the block\n            columns differ.\n  seeds     averaged over 7 seeds (SEED=20260601+k). Mandatory, because\n            subsample=0.8 makes each fit stochastic and dropping columns shifts\n            the RNG stream: a single-fit ablation moves held-out AUC by about\n            +-0.018 even when the removed block is IDENTICALLY ZERO, which is\n            roughly four times the threshold below.\n\nDECISION RULE, three outcomes pre-committed:\n  PASS        mean delta AUC across scored regions >= 0.005 AND the standard\n              error of that mean < 0.005.\n  FAIL        mean delta AUC < 0.005 AND the standard error < 0.005.\n  UNRESOLVED  standard error >= 0.005, or |mean delta| < 2 SE. Recorded as\n              UNRESOLVED, never as FAIL: reporting \"did not reach 0.005\" when\n              the instrument cannot see 0.005 records a measurement that was\n              not made. UNRESOLVED is an expected and publishable outcome.\n\nRe-eval 2026-11-10, giving ~94 days of manifest-covered snapshots with matured\noutcomes. Withdrawn v1 preserved at hypotheses/withdrawn/ -- the record of an\nuntestable pre-registration is itself the finding.\n\nBLOCK UNDER TEST: the 44 array-stacking tensor features -- sc_* per-array coherence index (26), regional_microseism_* region-local coherent sub-stack (9), ctd_* coherent-transient/slow-slip (9). CARRIED-FORWARD CAVEAT from v1: at the 8 continental-US anchor sites regional_microseism_* and ctd_* read 0, so any signal concentrates where they actually fire; per-region nonzero cell counts are reported alongside every delta.","ots_proof_path":"hypotheses/locked/20260811_v2_forward_array_stacking_tensor_f65b6d4d.json.ots","outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"earthquake"},"reevaluate_at":"2026-11-10","registered_at":"2026-08-11T17:36:19Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"ablation_holdout_auc_multiseed","n_permutations":null},"threshold":{"direction":"positive","metric":"delta_auc","value":0.005},"title":"v2 forward: array-stacking tensor block (44 features) raises per-region EQ held-out AUC, manifest-backed","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-11-10"},{"data_source":"earth_state_tensor.db state_snapshots + feature_manifests; prediction_ledger.db tensor_eq_per_region","expected_direction":"positive","feature_set":["lts_* (8)","swm_* (4)","lnm_* (7)","emit_* (8)","eco_* (6)","swot_* (6)"],"git_commit_at_registration":"94ee350186a1d49b89b1b1e19f4a70eb50e1ec16","hypothesis_id":"20260811_v2_forward_satellite_tensor_bloc_f5941117","locked_at":"2026-08-11T17:36:52Z","locked_file_sha256":"7786166cab1eaeabcb3a148e91ac9df88827d5ec8b7929c83ad12de03425c001","notes":"FORWARD TEST v2, re-issued 2026-08-11 after the v1 lock (2026-06-01) proved UNTESTABLE -- not disconfirmed. v1 could not be graded because the tensor archive carries no layout manifest before 2026-08-08, so the feature vectors at the prediction epochs were unreadable, and because no prediction inside the manifest window had expired.\n\nEVALUATION IS FIXED IN ADVANCE:\n  harness   scripts/eq_tensor_ablation.py at commit 8853cc7c (or a later commit\n            whose --self-test still passes unchanged)\n  eligible  ONLY predictions whose tensor_epoch matches, within 3600 s, a\n            snapshot carrying a layout_sha resolvable in feature_manifests.\n            Unlabelled snapshots are REFUSED, never aligned by width: 230\n            distinct widths exist and two different name lists can share a\n            length.\n  labels    ONLY predictions whose valid_until has passed. An unexpired,\n            unmatched prediction is UNDECIDED and is skipped, never scored 0.\n  model     GradientBoostingClassifier(n_estimators=150, max_depth=5,\n            subsample=0.8), per region, min 30 eligible predictions/region.\n  split     TEMPORAL, first 70% train / last 30% test. Never random: intra-\n            region predictions are dense and autocorrelated, so a random split\n            leaks future into past and inflates both arms.\n  arms      identical split, identical seed, identical rows; ONLY the block\n            columns differ.\n  seeds     averaged over 7 seeds (SEED=20260601+k). Mandatory, because\n            subsample=0.8 makes each fit stochastic and dropping columns shifts\n            the RNG stream: a single-fit ablation moves held-out AUC by about\n            +-0.018 even when the removed block is IDENTICALLY ZERO, which is\n            roughly four times the threshold below.\n\nDECISION RULE, three outcomes pre-committed:\n  PASS        mean delta AUC across scored regions >= 0.005 AND the standard\n              error of that mean < 0.005.\n  FAIL        mean delta AUC < 0.005 AND the standard error < 0.005.\n  UNRESOLVED  standard error >= 0.005, or |mean delta| < 2 SE. Recorded as\n              UNRESOLVED, never as FAIL: reporting \"did not reach 0.005\" when\n              the instrument cannot see 0.005 records a measurement that was\n              not made. UNRESOLVED is an expected and publishable outcome.\n\nRe-eval 2026-11-10, giving ~94 days of manifest-covered snapshots with matured\noutcomes. Withdrawn v1 preserved at hypotheses/withdrawn/ -- the record of an\nuntestable pre-registration is itself the finding.\n\nBLOCK UNDER TEST: the 39 satellite tensor features across 6 channels -- Landsat thermal/SWIR lts_ (8), ESA WorldCover swm_ (4), Cop-DEM lineament fabric lnm_ (7), EMIT mineral band-depths + methane emit_ (8), ECOSTRESS thermal anomaly eco_ (6), SWOT water level swot_ (6).","ots_proof_path":"hypotheses/locked/20260811_v2_forward_satellite_tensor_bloc_f5941117.json.ots","outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"earthquake"},"reevaluate_at":"2026-11-10","registered_at":"2026-08-11T17:36:19Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"locked","test":{"extra":{},"method":"ablation_holdout_auc_multiseed","n_permutations":null},"threshold":{"direction":"positive","metric":"delta_auc","value":0.005},"title":"v2 forward: satellite tensor block (39 features) raises per-region EQ held-out AUC, manifest-backed","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":"2026-11-10"},{"data_source":"prediction_ledger.db predictions+observations; swarm-box catalogue prior to predicted_at","expected_direction":"positive","feature_set":["swarm_detector_reactive expected_mag_min/expected_mag_max"],"git_commit_at_registration":"2bb5e6c98ce49c9eed48339bc45aa8320d06b842","hypothesis_id":"20260812_swarm_continuation_band_beats_an_faaeea4e","locked_at":"2026-08-12T03:27:53Z","locked_file_sha256":"d71fe78fee2fb210e12829f5b83d99d2840c76953295f5036fe41796d30dc76d","notes":"Forward-only from 2026-08-12. Pinnacles CA M3.9 (2026-08-11) motivated this and is EXCLUDED. Equal-width baseline: a wider band trivially wins, so the baseline gets our width and optimal placement. Secondary claim graded separately: upper bound biased LOW (exceedance above hi > shortfall below lo) -- the two shortest-lead Pinnacles calls capped hi at 3.76 against an M3.9. n<100 at grade date -> UNRESOLVED, never folded into either side. Full document: hypotheses/swarm_continuation_vs_etas_prereg_v1.md","ots_proof_path":"hypotheses/locked/20260812_swarm_continuation_band_beats_an_faaeea4e.json.ots","outcome":{"lookahead_hours":null,"min_magnitude":0.0,"type":"paired in-band rate difference vs best equal-width empirical G-R band"},"reevaluate_at":"2026-11-12","registered_at":"2026-08-12T03:27:29Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"paired binomial, two-sided","n_permutations":null},"threshold":{"direction":"positive","metric":"paired in-band rate difference (ours minus baseline)","value":0.1},"title":"Swarm-continuation band beats an equal-width G-R baseline","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"Teleseismic deep-event (M6+, depth>=100km, 88-125deg) BH?/HH? waveforms via FDSN, stacked per station; SWS-DB (Liu/EarthScope SWS-DB-MST) published per-station fast-azimuth + delay as reference truth, multi-layer-flagged stations excluded a priori","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"a359860279391fa25917987cf74ecbb671fe552e","hypothesis_id":"20260814_vq45_v2_virtual_anisotropy_field_78cf499d","locked_at":"2026-08-14T16:07:54Z","locked_file_sha256":"b0c46e751b23169faea3a74db1bf798c1331156616578b3ba23431ad437c99a2","notes":"VQ#45 v2 Virtual Anisotropy Field, SKS Silver-Chan splitting. FORWARD RE-ISSUE after v1 (ffb62804) was graded DISCONFIRMING: 43% of 23 stations within 25 deg (bar 0.50), median delay error 0.83 s (bar 0.70). This is a NEW entry, never a re-grade of ffb62804 -- the method changed, so the old seal must stand as recorded. DIAGNOSIS CARRIED FORWARD: v1 compared AUTOMATED SINGLE-EVENT picks against SWS-DB station averages that are curated over many events and backazimuths, and many reference stations are two-layer. The instrument was not wrong; the comparison was unequal. WHAT CHANGES IN v2, all fixed before running: (1) STACK before picking -- Silver-Chan transverse-energy minimisation is applied to the STACKED energy surface over all qualifying events at a station, not per-event with the angles averaged afterwards. Averaging axial angles across events is itself a trap (0 deg and 179 deg average to 90 deg), so the stack happens in energy space. (2) MINIMUM 5 qualifying SKS events per station, else the station is not scored. (3) BACKAZIMUTH BINNING -- where a station has events in 2 or more backazimuth bins, per-bin phi and dt are reported alongside the stacked value; a station whose per-bin values disagree by more than 30 deg is REPORTED as two-layer-suspect but still scored, so the exclusion below cannot be quietly widened. (4) TWO-LAYER EXCLUSION BY PUBLISHED METADATA ONLY -- stations flagged multi-layer in SWS-DB are dropped BEFORE running. Never dropped on our own fit quality or residual, which would be selecting on the metric being tested. (5) n >= 30 scored stations required, up from 23. Two sibling VQs (32, 36) were withdrawn for exactly this and the power must be stated in advance. BARS ARE UNCHANGED FROM v1 ON PURPOSE: frac_dphi_within_25 >= 0.50 AND median |ddt| <= 0.70 s. Raising or lowering a bar after seeing the v1 number would make a pass uninterpretable; keeping it means a pass is unambiguous method improvement. NEW REQUIREMENT -- A LIVE CONTROL, which v1 did not have: fast-axis is AXIAL, so a random azimuth lands within 25 deg with probability 2*25/180 = 0.278. v1's 43% was therefore ABOVE chance and BELOW bar -- informative but insufficient. v2 must ship a shuffled station-to-truth pairing whose frac_dphi_within_25 falls below 0.35. If the shuffled control does NOT collapse toward 0.278, the pass is an artefact of the comparison and is void regardless of the primary metric. OUTCOME VOCABULARY: PASS requires primary AND secondary AND a live control. If the standard error on the primary exceeds the margin to 0.50, report UNRESOLVED rather than FAIL -- reporting a failure the instrument could not have resolved records a measurement that was never made.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"reproducibility"},"reevaluate_at":null,"registered_at":"2026-08-14T16:07:44Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"backazimuth_bins":true,"chance_level":0.278,"control":"shuffled_station_truth_pairing","control_max_frac":0.35,"min_events_per_station":5,"min_scored_stations":30},"method":"silver_chan_transverse_energy_minimization_stacked","n_permutations":null},"threshold":{"direction":"positive","metric":"frac_dphi_within_25","value":0.5},"title":"VQ45 v2 virtual anisotropy field SKS multi-event stacked splitting reproducibility","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"NIED KiK-net PS-logging velocity profiles (672 stations, all starting at 0 m) + co-located KiK-net surface accelerometer records; frozen subset = 61 stations with a >=2.0x Vs contrast at 30-300 m. BVDA (SB network) held out as an out-of-corpus check.","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"2149c4bf41e2727cb1b5d97b1aad35526d470f73","hypothesis_id":"20260814_vq46_bar_s1b_virtual_sonic_log_h_97c36f9b","locked_at":"2026-08-14T19:11:19Z","locked_file_sha256":"7fa67ebc2bdf831ef5715b28d5a6db8b75395b5209322bd6784c09b4daea6131","notes":"VQ#46 Virtual Sonic Log, BAR-S1b. FORWARD RE-ISSUE after FORGE BAR-S1 FAILED TWICE. The failed bars stand as recorded; this is a new entry on a new corpus, never a re-grade. WHY FORGE WAS ILL-POSED: the DSI log in well 58-32 is BLIND ABOVE 657 m, so its 'strongest contrast at 1050 m' was strongest only WITHIN THE LOGGED INTERVAL. HVSR by construction reports the SHALLOWEST strong contrast and found ~170 m -- real, and unloggable by that well. The method found structure the truth data could not adjudicate. THE NEW CORPUS: NIED KiK-net PS-logging profiles, pulled and characterised 2026-08-14. 672 stations parsed, 0 unparseable; ALL 672 START AT 0 m; 671 carry both Vp and Vs on every layer; median 5 layers; median logged depth 80 m. Each station is a surface + downhole pair recording continuously, so truth and observation are co-located BY CONSTRUCTION. This is the property FORGE lacked. FROZEN POPULATION, fixed before any run: the stations whose LARGEST consecutive Vs jump is >= 2.0x AND whose contrast depth lies in 30-300 m. That is 61 stations on the corpus as characterised. 556 stations have a >=2.0x contrast but MOST SIT AT 1-6 m -- the soil-rock boundary -- which tests nothing about a passive depth method; quoting 556 would be quoting a number that cannot falsify. The 30-300 m band is chosen because it is where HVSR operates and where the FORGE argument was actually had. 61 also clears the n>=30 that withdrew VQ32 and VQ36 as underpowered. PRIMARY METRIC IS A FREQUENCY, NOT A DEPTH, AND THAT IS DELIBERATE. Converting a measured f0 into a depth needs a Vs, and the only Vs available is THE LOGGED PROFILE ITSELF -- using truth to predict truth. So the test runs FORWARD instead: predict f0 from the logged profile (quarter-wavelength, f0 = Vs_avg / 4H over the layers above the contrast) and compare against f0 measured by HVSR on the SURFACE sensor alone. Nothing from the log enters the measurement. BAR: frac_f0_within_30pct >= 0.50 across the frozen 61. SECONDARY (reported, not gating): median |log10(f0_meas/f0_pred)|, and the depth-domain comparison using a REGIONAL Vs independent of each station's own log. CONTROL, which FORGE BAR-S1 did not have: shuffled station-to-truth pairing must fall BELOW the primary metric by a stated margin. The chance level is NOT assumed -- it is measured from the corpus depth distribution and reported. If the shuffled control does not collapse, a pass is an artefact of the comparison and is VOID regardless of the primary. HELD OUT ENTIRELY: Borrego Valley Downhole Array (BVDA, SB network, published surface-to-230 m profile, Vs 300 -> 750 m/s then a jump to ~2500 at the granite interface). It is a US site with an independent provenance and is NOT scored; it is the out-of-corpus check run after the fact. OUTCOME VOCABULARY: PASS requires primary AND a live control. UNRESOLVED, not FAIL, if the standard error on the primary exceeds the margin to 0.50 -- reporting a failure the instrument could not resolve records a measurement never made. SCOPE STATEMENT that must ship with any pass: this corpus is SHALLOW (median logged depth 80 m, p90 186 m). A pass licenses claims about shallow interface recovery and says NOTHING about basement depth at a kilometre -- which is the exact mismatch that made FORGE ill-posed, in the other direction.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"reproducibility"},"reevaluate_at":null,"registered_at":"2026-08-14T19:11:06Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"chance_level":"measured_not_assumed","contrast_min_ratio":2.0,"control":"shuffled_station_truth_pairing","corpus_median_logged_depth_m":80,"depth_band_m":[30,300],"frozen_n":61,"held_out":"BVDA","primary":"frac_f0_within_30pct"},"method":"hvsr_f0_forward_prediction_vs_ps_logging","n_permutations":null},"threshold":{"direction":"positive","metric":"frac_f0_within_30pct","value":0.5},"title":"VQ46 BAR-S1b virtual sonic log HVSR interface recovery vs KiK-net PS logging","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"IGETS 13-station frozen set (Canberra,Cantley,Conrad,Larzac,Metsahovi,Moxa,Onsala,Pecny,Schiltach,Strasbourg,Trappes,Walferdange,Wettzell)","expected_direction":"positive","feature_set":["R_raw","R_geo","delta_z"],"git_commit_at_registration":"e877e907712da58a6dd703ca259789cb9b07b72b","hypothesis_id":"20260815_c5_wind_projection_geometry_disc_e98eed7d","locked_at":"2026-08-15T17:21:12Z","locked_file_sha256":"4cc191deda0470508a9162b7b8eb70872eb31dbfeb9c1e04333904c1aec604f7","notes":"c5 statistic Delta=R_geo-R_raw, operating z_min=1.03 (0.7 x geometric ceiling z_max=1.48), delta_null mu=+0.001 sd=1.353 over 60000 draws, dec=48.0. SHADOW criterion: does not alter any c1-c4 verdict. Degenerate alternative (c1-clearing bins show Delta<0) REJECTED and recorded: pure noise selected at R_raw>p99.99 already gives median Delta=-1.789 vs -2 for a real line. Mandatory pre-condition: per-station polarity confirmed against theoretical M2 tide phase; >2 of 13 failing = UNRESOLVED not FAIL. Full prereg hypotheses/c5_wind_geometry_prereg_v1.md, code frozen at e877e907.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_candidate_classification"},"reevaluate_at":"2026-09-30","registered_at":"2026-08-15T17:21:03Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"blind end-to-end injection challenge, 200 injections (100 wind-fixed / 100 common-mode) added to raw g BEFORE residualisation and whitening; seed = first 8 bytes SHA-256 of the first Bitcoin block hash after 2026-08-22 00:00 UTC","n_permutations":null},"threshold":{"direction":"positive","metric":"balanced_accuracy","value":0.9},"title":"c5 wind-projection geometry discriminates wind-fixed signals from shared environmental lines","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"data/igets/* (24 stations, 2009-2026 EOST L2 .ggp; path corrected from v1 stale igets_sg_archive)","expected_direction":"positive","feature_set":["peak_amplitude_uhz","peak_p_value","n_stations_passing","phase_coherence_score","coherent_stack_snr"],"git_commit_at_registration":"6b3e448b6a28da9c8c9d7c9ca7b2fe12a4dfd3f1","hypothesis_id":"20260822_v2_mars_a2_0s0_fundamental_whole_e3fdca22","locked_at":"2026-08-22T18:35:14Z","locked_file_sha256":"222e1dbc92c60a1e4a0268288dda2e15cde92dbe5197ab3a8db3cf4d3ca184dc","notes":"Replaces withdrawn v1 20260526_mars_a2_0s0_fundamental_whole_ea_1b0bbb26 (withdrawn 2026-08-22: untestable as registered -- the stack+bootstrap instrument was never built; lock-time-readability lesson applied here as an explicit instrument-readiness gate in test.extra). Same scientific question: does the canonical 0S0 breathing mode (PREM ~813.92 uHz, period 20.5 min; Park-Gilbert 1986 first observational confirmation after large events) appear as a persistent AMBIENT narrowband line in the IGETS SG archive. PRELIMINARY CONSTRAINT carried from v1 grading attempt (2026-08-22, real data, NOT the registered test): scripts/mars_a2_0s0_confirmation_v0p1.py on the most-recent 60-day window reads 24/24 stations with data, peak SNR max 1.26 at 813.5-814.2 uHz, 0 stations >= SNR 3 -- consistent with 0S0 amplitude below the single-60-day-window sensitivity floor, saying nothing yet about the 17-year stack. The registered instrument is the D.1-SPRINT engineering (ICET + Hinderer 1995 + M2/S2/N2 harmonic subtraction + multi-window complex coherent stack + per-station 50k block bootstrap); a methods-clean NULL from it is publishable as a global constraint on ambient 0S0 amplitude/Q at the IGETS sensitivity floor.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"free_oscillation_detection"},"reevaluate_at":"2027-02-28","registered_at":"2026-08-22T18:35:01Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"bootstrap_method":"block_bootstrap_n50000","frequency_tolerance_uhz":1.0,"instrument_readiness_gate":"grading REQUIRES the registered stack+bootstrap instrument (D.1-SPRINT deliverable); if absent at re-eval, WITHDRAW again -- never grade with a proxy","mode":"0S0","period_minutes":20.5,"preprocessing":"ICET_corrected_plus_M2_S2_N2_harmonic_subtraction_plus_Hinderer_1995_atmospheric_loading_correction","stack_method":"complex_coherent_stack_across_stations_multi_window_2009_2026","target_frequency_uhz":813.92},"method":"hanning_window_fft_coherent_stack","n_permutations":50000},"threshold":{"direction":"positive","metric":"n_stations_passing_p_lt_0p001","value":5.0},"title":"v2: MARS-A2 0S0 fundamental whole-Earth radial mode at ~813.92 uHz detected at p<0.001 (50k block bootstrap) at >=5 of 24 IGETS SG stations via 2009-2026 multi-window coherent stack","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"data/prediction_ledger.db + cross_physics/paired_megafault_predictor.py (hourly cron) + ComCat count endpoint (measured trailing-365d M3+ baselines per bbox)","expected_direction":"positive","feature_set":["pmf_joint_score","pmf_posterior_alert_prob_7d","pmf_n_pairs_joint_elevated","pmf_strongest_pair_a_count_7d","pmf_strongest_pair_b_count_7d"],"git_commit_at_registration":"0444c7e21ea86dbaf73577ec80db9ae9562e02e5","hypothesis_id":"20260822_v3_paired_megafault_paleoseismic_86677bbd","locked_at":"2026-08-22T19:07:34Z","locked_file_sha256":"97f62f42c5e18f74f928b3309537dec34de6b9fd29ac374353a37d654d4cf2bf","notes":"Replaces withdrawn v2 20260521_v2_paired_megafault_paleoseismic_70095bae. V2 DIAGNOSIS CORRECTED BY MEASUREMENT (91-day autopsy, daily rolling 7d M3+ counts vs the monitor own ComCat feed): the hardcoded historical_rate_per_month constants were 7-9x too high (cascadia median 1.0/7d vs assumed 7.0; ucerf3 3.0 vs 28.0), so the v2 joint gate demanded ~13x the true rate in both regions at once -- unreachable by arithmetic; closest joint approach in the whole window was 12% of threshold. The v2 null was therefore BRANCH 2 (gate miscalibration, the declared-constant-never-tested class), not quiet faults. V3 gate (commit 0444c7e2): per-region elevation = Poisson tail P(X>=count | lambda) < 0.05 with lambda = MEASURED trailing-365d rate from the same count endpoint, refreshed each scan; gating requires fully-real inputs (live counts real AND baseline fetched) -- synthetic fallbacks can never fire it. Chosen over rescaled multipliers because small-lambda regions need a tail probability: 1.5x a median of 1 fires on 18%-likely fluctuations, which the Poisson form correctly rejects. Expected chance-level joint fires over the 184-day window ~0.46 days if regions are independent; the Goldfinger coupling hypothesis predicts excess co-elevation. Same scientific test as v1/v2: filter model_name = paired_megafault, pass = paired_match_rate / unpaired_match_rate >= 2.0 AND paired_n >= 5. Live smoke at registration: cascadia count 1 vs lambda 2.57, ucerf3 2 vs 2.95, both correctly not elevated.","ots_proof_path":null,"outcome":{"lookahead_hours":168.0,"min_magnitude":4.5,"type":"earthquake"},"reevaluate_at":"2027-02-22","registered_at":"2026-08-22T19:07:34Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"elevation_gate":"poisson_tail_p_lt_0.05_vs_measured_trailing_365d_rate","gate_commit":"0444c7e2","joint_null_rate_per_day_if_independent":0.0025,"withdraw_if":"paired_n < 5 at re-eval -- and with the gate now CALIBRATED, a zero is a real statement that no joint co-elevation occurred, not arithmetic"},"method":"match_rate_ratio_paired_vs_unpaired","n_permutations":null},"threshold":{"direction":"positive","metric":"match_rate_ratio","value":2.0},"title":"v3: paired-megafault paleoseismic prior boost (Poisson gate on measured baselines) -- paired predictions match rate >=2x unpaired baseline","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"data/igets_cleaned_v2r (manifest sha256 29c2a4f7c6e3be3ada244393128d3396ca342c2505f58a285951ba05c098e81d, 20 piers / 303.3 station-years) via scripts/igets/slichter_search_v5.py (sha16 78c060cb026034c0, self-test 110/110)","expected_direction":"positive","feature_set":["slichter_v5_a95_q500","slichter_v5_a95_q1000","slichter_v5_a95_q2000","slichter_v5_a95_q5000","slichter_v5_neff_inband","slichter_v5_live_piers_per_onset"],"git_commit_at_registration":"226d8e964ba4ea42dc73c70735b77921f9d9aedb","hypothesis_id":"20260823_slichter_1s1_search_v5_damped_on_841f9f76","locked_at":"2026-08-23T23:25:32Z","locked_file_sha256":"2b8a01914efc18e0f6a8162a913e202dbb90f3c01852fa10f2c2b7957a7ebd43","notes":"v4's damped tier was UNRESOLVED in all 15 cells because block-median whitening eats a damped mode's line width (HWHM 0.372 whitening blocks at Q=500) and a persistent single-bin filter mismatches a Lorentzian at a penalty scaling ~Q instead of ~sqrt(Q). BOTH fixes were validated on synthetic injections BEFORE this seal: v5 sits at 0.91-0.93 of sqrt(2 pi f T / Q) flat in Q while v4 sits at 0.87-0.89 of its own ~Q form, so the two laws separate cleanly rather than being fitted apart; and at Q=500 geometry v5's gain is flat to 2% over a 1000x sweep where the v4 path collapses x122, reproducing v4's sealed x103 as a negative control. Trials came out LOWER than v4 (1.52e4 vs 1.40e5) because a matched filter is a smoother -- a real property, not a relaxed bar. Limitations declared in the doc: sensitivity validated on pure noise, array coverage rather than the filter sets detectability at a fixed onset (5 of 20 piers live at onset 0 vs 19 of 20 mid-record), the v5 null tail is heavier so it must be fitted with 9.4 extrapolation decades printed, adjacent-Q templates overlap 0.90-0.94, Smylie exclusion language stays forbidden, and Ding & Chao remain tighter on the persistent quantity.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"solid_earth_normal_mode"},"reevaluate_at":"2026-12-31","registered_at":"2026-08-23T23:24:50Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"Q_declared_not_extendable":[500,1000,2000,5000],"must_beat_v4":"a finite A95 for at least Q=5000 and Q=2000 where v4 had none; if v5 is also UNRESOLVED the build did not deliver","never_compare":"S_thr(v5) vs S_thr(v4) -- different scales; cross-version statements in A95 only","onset_grid":"12 uniform onsets over the record INCLUDING poorly-covered ones; reported per onset, never best-onset","outcome_classes":["NULL_A95_per_cell","DETECTION-CANDIDATE","UNRESOLVED"],"predecessors":["20260822_slichter_1s1_triplet_search_v2_o_9af67bec","20260823_slichter_1s1_triplet_search_v3_i_8a6809e7","20260823_slichter_1s1_search_v4_the_first_84e85c3b"],"predicted_a95_nm_s2":{"1000":[1.39,1.46,1.4],"2000":[0.99,1.05,1.0],"500":[2.0,2.11,2.01],"5000":[0.63,0.67,0.64]},"prereg_doc":"hypotheses/slichter_v5_damped_prereg_v1.md","prereg_doc_sha256":"7f93e4fcc5d5d5ef121a432c2394b9d20d0c56791442d69d527c0c5207605e96","red_flag":"materially BETTER than the predicted table is a RED FLAG -- it is a pure-noise projection and the real band carries the M3 satellite forest, so the measurement should be WORSE","scope":"DAMPED ONLY -- the persistent bound stays graded at v4 and is not re-scored"},"method":"sealed_matched_filter_search","n_permutations":null},"threshold":{"direction":"positive","metric":"trials_corrected_p_fa","value":0.001},"title":"Slichter 1S1 search v5 (DAMPED-ONLY): matched Lorentzian filter + Q-dependent whitening + onset scan -- the first Q-aware bound","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"NGL tenv3 IGS20 24hr GNSS; USGS Slab2 geometry; Okada 1985 GF","expected_direction":"positive","feature_set":["gnss_gf_matched_filter_projection","gnss_slip_inversion_where"],"git_commit_at_registration":"03a7cecf5184cda70d94caed442eb2985b938f85","hypothesis_id":"20260824_gnss_green_s_function_matched_fi_4da88a9d","locked_at":"2026-08-24T20:35:09Z","locked_file_sha256":"2045ef7ac0665c9ed52605ed8276b727c990d6c65b0d15d91b6b428ccfb3f165","notes":"ADOPTED method (trackAS). Frozen g = unit thrust slip on Slab2 locked/updip zone via Okada 1985 (validated 2.7e-4 vs screw-dislocation analytic), computed BEFORE pre-event data. MF = sum(g.d/sigma^2)/sqrt(sum|g|^2/sigma^2) ~ N(0,1) under noise. Gates: must recover Cascadia SSE + return chance on Ridgecrest dense-net control. Subduction updip-dense margins ONLY (N Chile, Cascadia, Nankai/Boso, dense Alaska); never standalone; window-sweep + foreshock exclusion MANDATORY (does not remove coseismic leakage). Forward-only, per-event, NO pooling. Frozen spec + code: scripts/gnss_precursor/ (trackAS_report.md, trackAS_mf.py, trackAS_okada.py, trackAS_geom.py, trackAS_slab2.py).","ots_proof_path":null,"outcome":{"lookahead_hours":504.0,"min_magnitude":6.5,"type":"earthquake"},"reevaluate_at":null,"registered_at":"2026-08-24T20:34:54Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{},"method":"GF matched-filter projection onto frozen locked-zone slip pattern; event-time shuffle null (>=300), window-endpoint sweep, foreshock-window exclusion, far-annulus common-mode filter","n_permutations":300},"threshold":{"direction":"positive","metric":"MF_projection_percentile","value":95.0},"title":"GNSS Green's-function matched filter detects distributed pre-event slip on subduction updip-dense margins (forward)","withdrawn":null,"_ots_anchored":true,"_status_dir":"locked","_re_eval_date":null},{"data_source":"NGL IGS20 tenv3 (geodesy.unr.edu); 4 subduction updip margins; frozen station set sha256 19b59408","expected_direction":"positive","feature_set":["coherence_pctile","trench_proj_pctile","azimuth_concentration_R","angle_err_vs_trench_deg"],"git_commit_at_registration":"3a3a8e3dbbd4a848a923d43d0e4952abd4b939fb","hypothesis_id":"20260826_gnss_distributed_trenchward_rate_378219d7","locked_at":"2026-08-26T21:02:32Z","locked_file_sha256":"900ed786d602b65ae78ffcaaaa456469a546d0c47f7d97102064635687241fd5","notes":"The one precursor channel with a passing positive control (trackAQ): coherent trenchward network GNSS transient over a subduction updip zone before a great EQ. Records only (no ledger, no alarm). Full frozen spec + provenance in the draft. Positive controls (Iquique 2014, Cascadia ETS) pass; Ridgecrest negative control passes.","ots_proof_path":null,"outcome":{"lookahead_hours":504.0,"min_magnitude":7.5,"type":"earthquake"},"reevaluate_at":"2027-08-26","registered_at":"2026-08-26T21:02:15Z","registered_by":"Mljessop@Mike","result":null,"status":"locked","test":{"extra":{"cmf":"far-annulus spatial-median common-mode removed from near stations","draft":"hypotheses/drafts/gnss_trenchward_precursor_prereg_DRAFT.md","exclusions":["event_day","prior_M6_foreshock_window","prior_earthquake_postseismic_afterslip_radius_max(150km,4L)_tau_to_1yr"],"grading":"forward-only from lock date, per-event, NO pooling","margins_trench_az_deg":{"aleutian":190,"cascadia":260,"n_chile":270,"nankai":135},"null":"300 event-time shuffles, day-of-year matched, station geometry fixed","positive_criterion":"coherence_pctile>=97.5 AND angle_err_vs_trench<=25deg AND azimuth_concentration_R>=0.6","pre_window_days":[3,21],"recorder":"scripts/gnss_trenchward_precursor.py --record","stations_frozen":"data/gnss_trenchward_frozen_stations.json","stations_sha256":"19b59408f9967d34fe977a440f2ff629b7e7759289c6fce23c3dbeb814fd35d0","statistic":"CMF-corrected 21d network coherence + trench-normal projection (Statistic B)"},"method":"cmf_network_coherence_trenchward_vs_event_time_shuffle","n_permutations":300},"threshold":{"direction":"positive","metric":"coherence_pctile","value":97.5},"title":"GNSS distributed trenchward-rate precursor (trackAQ forward test)","withdrawn":null,"_ots_anchored":false,"_status_dir":"locked","_re_eval_date":null},{"data_source":"Mars InSight XB.ELYSE.02.BHU/BHV/BHW -> Z (2020-01-01..2021-12-31 core) via IRIS/IPGP/EarthScope FDSN; Earth GSN {IU.ANMO,II.BFO,IU.TSUM,IU.CCM}.00.BHZ calibrated to ACC","expected_direction":"NULL_BOUND","feature_set":["gamma2_cross_coherence","S_x_cross_power"],"git_commit_at_registration":"fbfde76eba111acff90ea4c11ddd3890a19537c8","hypothesis_id":"20260902_earth_x_mars_cross_planetary_ult_1084ce55","locked_at":"2026-09-02T03:03:59Z","locked_file_sha256":"f3ad33a143c843bea334b6fc774af6ba831b0c73aeb21cc1f90d3d8bdb54b3ce","notes":"Earth x Mars cross-planetary ultralight-DM Compton-line COHERENCE search (v0). Frozen code scripts/dm_earthmars_coherence.py sha256 4ab0f4e4c68ef042acdbf0f74bf7c49ddaeeb5698dfad450ad95e5b737800723 (git b6d2321c31d5dc08dd151048880ace051f35a2bd); prereg _active_today/dm_earthmars_coherence_prereg_v0_DRAFT.md (git fbfde76eba111acff90ea4c11ddd3890a19537c8, sha256 a8e5aebd73b4b2ae8cdfe5a6255ba6880334d43c31e68f3d16b13ebf4e687012). Projection Z-only both bodies; Mars XB.ELYSE.02.BHU/V/W->Z over the 2020-01-01..2021-12-31 core; Earth GSN {IU.ANMO,II.BFO,IU.TSUM,IU.CCM}.00.BHZ (pre-declared duty-cycle drop rule; common-mode single estimate, NOT sqrt(N)). Statistic gamma^2=|sum A_E conj(A_M)|^2/(sum|A_E|^2 sum|A_M|^2) on a Delta_f/f=1e-6 log grid over 0.01-1 Hz, N_bins=4.605e6. expected_direction=NULL_BOUND. Discovery = analytic Sidak z>=5.593 (per-bin alpha 1.114e-8) AND single-body HARD coincidence bar AND in-run injection positive-control gate per-arm [0.75,1.33] at fiducial R_E/R_M x xi (xi=1). Nulls: per-body phase-scramble(>=500) + time-shift/block-permutation + dummy-frequency-grid(>=150); NO circular-shift. Mars night/low-wind gate = MEDIAN (50th-percentile) keep rule (wind_keep_percentile=50); threshold RMS value data-derived forward-blind at run. Seed = sha256(confirmed OTS Bitcoin block hash of the recorded anchor), forward-blind, set AT SEAL; NO real science bin scored before the anchor confirms. The RUN (--stage run --sealed-seed <seed> --seal-anchor <registry_id>) is a later Mike-fired step AFTER Bitcoin confirmation.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_line_search"},"reevaluate_at":null,"registered_at":"2026-09-02T03:03:59Z","registered_by":"Mljessop@deepmap-compute","result":null,"status":"locked","test":{"extra":{"code_git_commit":"b6d2321c31d5dc08dd151048880ace051f35a2bd","code_path":"scripts/dm_earthmars_coherence.py","code_sha256":"4ab0f4e4c68ef042acdbf0f74bf7c49ddaeeb5698dfad450ad95e5b737800723","discovery_rule":"candidate iff analytic Sidak z >= 5.593 over N_bins=4.605e6 (Carter/Nuttall MSC null) AND single-body HARD coincidence bar (single-body injection yields NO coincidence) AND injection positive-control per-arm recovered/injected in [0.75,1.33] at the fiducial Earth:Mars ratio R_E/R_M x xi (xi=1); else NULL.","expected_direction":"NULL_BOUND","expected_outcome":"FIRST cross-planet DM line search; a NULL reported as a 95% exclusion on the observable per-body line amplitude (m/s^2) + the cross-coherence bound + an unfakeable Earth-Mars coincidence veto; coupling (cm^2/epsilon) DEFERRED to a theorist (retraction discipline).","frozen_config":{"band_hz":[0.01,1.0],"bar_sidak_z":5.593,"df_over_f":1e-6,"earth_duty_floor_rule":"drop a station BEFORE scoring if its 2020-2021 duty < the pre-declared floor (frozen rule, not post-hoc)","earth_stations":["IU.ANMO.00.BHZ","II.BFO.00.BHZ","IU.TSUM.00.BHZ","IU.CCM.00.BHZ"],"global_alpha":0.05,"injection_tol":[0.75,1.33],"mars":{"chans":["BHU","BHV","BHW"],"loc":"02","net":"XB","sta":"ELYSE"},"mars_baseline":["2020-01-01T00:00:00","2021-12-31T23:59:59"],"n_bins":4605170.185988092,"null_battery":{"circular_shift":"EXCLUDED","dummy_floor":150,"phase_scramble":500,"time_shift":500},"projection":"Z","ratio_E_over_M":1.8796282637557162,"seal_anchor_id":null,"seed_rule":"science RNG seed = sha256(confirmed OTS Bitcoin block hash of the recorded anchor); forward-blind, set AT SEAL","single_body_rejection":"HARD","wind_keep_percentile":50.0,"xi_elastic":1.0},"integrity_gate":"real-data scoring is refused without the recorded seed bound to code sha256 4ab0f4e4 (SEAL_ANCHOR_ID None -> --seal-anchor supplied at run time).","prereg_git_commit":"fbfde76eba111acff90ea4c11ddd3890a19537c8","prereg_path":"_active_today/dm_earthmars_coherence_prereg_v0_DRAFT.md","prereg_sha256":"a8e5aebd73b4b2ae8cdfe5a6255ba6880334d43c31e68f3d16b13ebf4e687012","run_command_after_btc_confirm":"--stage run --sealed-seed <sha256(bitcoin_block_hash)> --seal-anchor <registry_id>  per f_C chunk over 0.01-1 Hz","seed_rule":"science RNG seed = sha256(confirmed OTS Bitcoin block hash of the recorded anchor); forward-blind, set AT SEAL; NO real science bin scored before the anchor confirms.","wind_gate":"MEDIAN / 50th-percentile Mars night/low-wind keep rule (wind_keep_percentile=50); threshold RMS value is data-derived forward-blind at run over the 2020-2021 core."},"method":"CoSMoS phase-demod Earth x Mars magnitude-squared cross-coherence gamma^2 per f_C over 0.01-1 Hz; discovery = analytic Sidak z>=5.593 (N_bins=4.605e6) + single-body HARD coincidence bar + in-run injection positive-control gate [0.75,1.33] (xi=1); empirical nulls: per-body phase-scramble(>=500) + time-shift/block-permutation + dummy-frequency-grid(>=150), NO circular-shift","n_permutations":500},"threshold":{"direction":"","metric":"sidak_z","value":5.593},"title":"Earth x Mars cross-planetary ultralight-DM Compton-line coherence search v0","withdrawn":null,"_ots_anchored":false,"_status_dir":"locked","_re_eval_date":null},{"hypothesis_id":"20260902_earth_x_mars_cross_planetary_ult_1084ce55","bitcoin_block_hash":"00000000000000000001d6a7334dd6624a03397ed241c9a276b13b7c9db27bf4","bitcoin_block_height":965117,"bitcoin_block_merkle_root":"f55900516975bd538ce069768cc44d5660ce7a9f1778369cf67a4fac73582a23","bitcoin_block_time_utc":"2026-09-02T03:10:35Z","attesting_blocks_in_ots":[965117,965121,965152],"rule":"EARLIEST Bitcoin block among all BitcoinBlockHeaderAttestation entries in the upgraded .ots proof -- the first moment the seal became irrevocable. Deterministic from the proof alone (anyone re-deriving from the .ots reaches the same block); no operator choice.","block_hash_sources":["blockstream.info/api/block-height","mempool.space/api/block-height (agree)"],"code_sha256":"4ab0f4e4c68ef042acdbf0f74bf7c49ddaeeb5698dfad450ad95e5b737800723","seed_rule":"seed = sha256(bitcoin_block_hash hex) (derive_seed_from_block)","seed":"ae93093e44519da7818a876465098eff7cea9488466849a0eeefa8f13bcf7d79","decided_utc":"2026-09-02T15:15:11Z","decided_by":"Claude Fable 5.1 session 2026-09-02 per pickup_sep1 action 1 (Mike delegated the run sequence)","note":"SIDECAR ONLY -- the canonical locked .json is untouched (editing it would break the OTS anchor).","_ots_anchored":false,"_status_dir":"locked","_re_eval_date":null},{"data_source":"Mars InSight XB.ELYSE.02.BHU/BHV/BHW -> Z (2020-01-01..2021-12-31 core) via IRIS/IPGP/EarthScope FDSN; Earth GSN {IU.ANMO,II.BFO,IU.TSUM,IU.CCM}.00.BHZ calibrated to ACC; prepared store (--stage prepare)","expected_direction":"","feature_set":["gamma2_cross_coherence","S_x_cross_power"],"git_commit_at_registration":"e4b78c624b6795bf32a1b9256eb519d89ae0d659","hypothesis_id":"20260902_earth_x_mars_cross_planetary_ult_813fc905","locked_at":"2026-09-02T22:02:35Z","locked_file_sha256":"e2c8fc579b8218d5ebc0cd691905ee3b54f36a50963bff422b780e1be047370a","notes":"Earth x Mars cross-planetary ultralight-DM Compton-line COHERENCE search v1.1 (scale amendment; supersedes v1 75974252 withdrawn before any run and v0 1084ce55 halted at its gate). Frozen code scripts/dm_earthmars_coherence.py sha256 ee81b4b9d39a9b6bc3a0ddbfd4c9cbb70aba32db2925a17bf9011e2206ac522f (git e4b78c62); prereg _active_today/dm_earthmars_coherence_prereg_v1_AMENDMENT.md sha256 522c5ff179da8f6cd6af3f3c76dce3f70d04c5d814fbc6f7a5c16441ab5a3d41. Changes vs v0 (all stated before any bin is scored): scored population = prepared store of the frozen 2020-2021 core; FFT evaluation of the same lock-in amplitude (pad 32, chunked) with an in-run equivalence gate (40 patches x 600 bins, tol 1e-3, measured 5e-5); LOCAL injection floor; injection amplitude = the search's claimed sensitivity (expected gamma2 = 2 x bar at the run's K); single-body HARD rule operationalised (z < bar AND median p over 5 draws > 0.05); empirical null on candidates + 10k-bin calibration subsample. UNCHANGED: band 0.01-1 Hz, df/f 1e-6, N_bins 4.605e6, statistic gamma2, Sidak z 5.593, INJ_TOL, xi=1, R=1.8796, projection Z, wind percentile 50, no circular-shift null, seed = sha256(EARLIEST attesting Bitcoin block hash of THIS anchor), no real bin scored before confirmation. expected_direction=NULL_BOUND.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_line_search"},"reevaluate_at":null,"registered_at":"2026-09-02T22:02:24Z","registered_by":"Mljessop@deepmap-compute","result":null,"status":"locked","test":{"extra":{"code_git_commit":"e4b78c624b6795bf32a1b9256eb519d89ae0d659","code_path":"scripts/dm_earthmars_coherence.py","code_sha256":"ee81b4b9d39a9b6bc3a0ddbfd4c9cbb70aba32db2925a17bf9011e2206ac522f","discovery_rule":"candidate iff analytic Sidak z >= 5.593 over N_bins=4.605e6 (Carter/Nuttall MSC null) on FFT-evaluated lock-in amplitudes (in-run equivalence gate on 40 seeded patches x 600 bins: <= 1e-3 rel amplitude error vs patch_amplitude, else exit 6) AND empirical phase-scramble p <= 0.05 with a live null AND single-body HARD rule (one-body analytic z < bar AND median scramble p over 5 one-body draws > 0.05) AND in-run positive control: per-arm recovered/injected in [0.75,1.33] against a LOCAL floor (60 log-uniform dummies within +/-20% of f_inj=0.05 Hz) at the injection amplitude whose expected gamma2 = 2 x gamma2*(K), joint z >= 5.593; else NULL.","empirical_null_policy":"phase-scramble >=500 draws on every candidate + 10,000-bin calibration subsample x 200 draws (aliveness + FAR reported, never promoting)","evaluation":"fine spectrum per patch (pad 32, complex64) + cubic interpolation in 200,000-bin chunks; same arithmetic as patch_amplitude","expected_direction":"NULL_BOUND","expected_outcome":"FIRST cross-planet DM line search; a NULL reported as a 95% exclusion on the observable per-body line amplitude (m/s^2) = the injected amplitude scaled to gamma2*(K), + the cross-coherence bound + the Earth-Mars coincidence veto; coupling (cm^2/epsilon) DEFERRED to a theorist.","frozen_config":{"amendment_v1":{"empirical_null_policy":"phase-scramble on every candidate (>= 500 draws) + 10000-bin calibration subsample (200 draws) reported, never promoting","equivalence_bins":600,"equivalence_patches":40,"evaluation":"FFT path (patch_fine_spectrum + cubic interpolation) reproducing patch_amplitude; gated by --stage equivalence in-run","fft_equiv_tol":0.001,"fft_pad":32,"injection_amplitude":"amp_E such that expected gamma2 = 2.0 x gamma2*(K); amp_M = amp_E / ratio","injection_floor":"LOCAL: 60 log-uniform dummy freqs within +/-20% of f_inj","scan_chunk_bins":200000,"scored_population":"prepared store of the frozen core (MARS_BASELINE_START..END), daily-chunked, wind percentile applied over the whole baseline","store_version":"v1","unchanged":["band","df_over_f","n_bins","statistic","bar_sidak_z","single_body_hard","inj_tol","xi","ratio","projection","wind_keep_percentile","seed_rule"]},"band_hz":[0.01,1.0],"bar_sidak_z":5.593,"df_over_f":1e-6,"earth_duty_floor_rule":"drop a station BEFORE scoring if its 2020-2021 duty < the pre-declared floor (frozen rule, not post-hoc)","earth_stations":["IU.ANMO.00.BHZ","II.BFO.00.BHZ","IU.TSUM.00.BHZ","IU.CCM.00.BHZ"],"global_alpha":0.05,"injection_tol":[0.75,1.33],"mars":{"chans":["BHU","BHV","BHW"],"loc":"02","net":"XB","sta":"ELYSE"},"mars_baseline":["2020-01-01T00:00:00","2021-12-31T23:59:59"],"n_bins":4605170.185988092,"null_battery":{"circular_shift":"EXCLUDED","dummy_floor":150,"phase_scramble":500,"time_shift":500},"projection":"Z","ratio_E_over_M":1.8796282637557162,"seal_anchor_id":null,"seed_rule":"science RNG seed = sha256(confirmed OTS Bitcoin block hash of the recorded anchor); forward-blind, set AT SEAL","single_body_rejection":"HARD","wind_keep_percentile":50.0,"xi_elastic":1.0},"pre_seal_disclosure":"pre-seal equivalence measurements formed gamma2 on 2 x 300 sampled real grid bins of a 3-day K=6 subset (both paths agreed on the candidate set; max z not recorded); pre-seal --stage equivalence is amplitude-only thereafter; no other real bin looked at","prereg_path":"_active_today/dm_earthmars_coherence_prereg_v1_AMENDMENT.md","prereg_sha256":"522c5ff179da8f6cd6af3f3c76dce3f70d04c5d814fbc6f7a5c16441ab5a3d41","scored_population":"prepared store of the frozen 2020-01-01..2021-12-31 core (--stage prepare, daily-chunked); wind percentile 50 over the whole baseline; manifest sha256 + counts recorded by the run","supersedes":["20260902_earth_x_mars_cross_planetary_ult_1084ce55 (v0, HALTED at its in-run gate, zero bins scored)","20260902_earth_x_mars_cross_planetary_ult_75974252 (v1, WITHDRAWN before any run: scale defects -- K x 4.6e6 matrix, 60-h equivalence)"],"synthetic_validation":{"fft_equivalence_real_patches":"pad 32: 5.2e-5 (E) / 7.8e-5 (M) rel amp err, |dgamma2| 1.2e-5","phase0_original":"27/27 + freeze-scan 6/6","phase0fft_full_grid_K250":{"calibration":{"draws_per_bin":200,"empirical_far_at_bar":0.07,"expected_far_at_bar":0.05,"median_gamma2":0.002711184685767127,"n_bins":400,"null_alive_fraction":1.0},"checks":{"calibration_far_near_bar":true,"calibration_null_alive":true,"fft_equivalence_synthetic":true,"gate_passes_coincident":true,"line_candidates_null_alive":true,"line_found_at_bar":true,"no_line_no_candidates":true,"one_body_no_candidate":true},"line":{"K_median":250.0,"bar_sidak_z":5.593,"chunk_bins":200000,"f_at_z_max":0.07310305063135607,"gamma2_max":0.1524848803039705,"n_bins":4605173,"n_candidate_clusters":1,"n_candidates":492,"scan_seconds":267.7494411468506,"z_max":8.729019722266509},"no_line":{"K_median":250.0,"bar_sidak_z":5.593,"chunk_bins":200000,"f_at_z_max":0.6175611124025623,"gamma2_max":0.047754879582223506,"n_bins":4605173,"n_candidate_clusters":0,"n_candidates":0,"scan_seconds":276.7593641281128,"z_max":4.412456051280824}},"phase0fft_local":"8/8"}},"method":"fft_cross_coherence_sidak_scan_v1_1","n_permutations":500},"threshold":{"direction":"","metric":"sidak_z","value":5.593},"title":"Earth x Mars cross-planetary ultralight-DM Compton-line coherence search v1.1 (scale amendment; supersedes v1 75974252 and v0 1084ce55)","withdrawn":null,"_ots_anchored":false,"_status_dir":"locked","_re_eval_date":null},{"hypothesis_id":"20260902_earth_x_mars_cross_planetary_ult_813fc905","bitcoin_block_hash":"00000000000000000000a6f1eef2d5ba58725d4b8795fa1772e910fa360d2cd1","bitcoin_block_height":965229,"rule":"EARLIEST Bitcoin block among all BitcoinBlockHeaderAttestation entries in the upgraded .ots (D1)","block_hash_sources":["blockstream.info","mempool.space (agree)"],"code_sha256":"ee81b4b9d39a9b6bc3a0ddbfd4c9cbb70aba32db2925a17bf9011e2206ac522f","seed":"b37d1401909f0b61de51b5bb72fa099befd305e5686009d4e1ab4a5c8ac23dc4","decided_utc":"2026-09-03T00:09:38Z","note":"SIDECAR ONLY -- canonical locked .json untouched","_ots_anchored":false,"_status_dir":"locked","_re_eval_date":null},{"anchor":"20260903_dm_magnetic_k1_adjudication_v2_4_cf01e1fb","block":965345,"block_hash":"00000000000000000001b97f2d606db533cd1bb6b344bee7500f125409ce1547","k1_seed_campaign":1003605714,"btc_confirmed":true,"_ots_anchored":false,"_status_dir":"locked","_re_eval_date":null},{"data_source":"Mars InSight XB.ELYSE.02.BHU/BHV/BHW -> Z (2020-01-01..2021-12-31 core) via IRIS/IPGP/EarthScope FDSN; Earth GSN {IU.ANMO,II.BFO,IU.TSUM,IU.CCM}.00.BHZ calibrated to ACC; prepared store REUSED from v1.1 (manifest 23b964f7b9d0)","expected_direction":"","feature_set":["gamma2_cross_coherence","S_x_cross_power"],"git_commit_at_registration":"a68508b71df885bcd7906b57c2f3414a9fa9df5f","hypothesis_id":"20260903_earth_x_mars_cross_planetary_ult_01fea7ac","locked_at":"2026-09-03T02:16:10Z","locked_file_sha256":"56d2f866b70c4fcb938708f1a4845cdbe1417dac335841f2a6e438bc1e819e0e","notes":"Earth x Mars cross-planetary ultralight-DM Compton-line COHERENCE search v1.2 (differential positive control; supersedes v1.1 813fc905 which was sealed, Bitcoin-anchored at block 965229, RAN and HALTED at its own injection positive control with ZERO real bins scored -- UNTESTABLE-AS-FROZEN, NOT a null). Frozen code scripts/dm_earthmars_coherence_v12.py sha256 c1d04a58f3d2fb47f5c87f409a5e082e45870bf6c83824cec068bdaa0afe69b8 (git a68508b7); prereg _active_today/dm_earthmars_coherence_prereg_v1_2_AMENDMENT.md sha256 db56bee072ceae5845c2e0bdedc095addc18f971febf67ab2db9c68312e33754. THE ONLY CHANGE vs v1.1: the injection positive control recovers by DIFFERENTIAL power on the same patches, phase-averaged over 4 injection phases, instead of subtracting an ESTIMATED local floor -- because the estimated floor explained only 73.7% of the Earth arm's power and would have had to be accurate to 2.3% of a quantity 38x the signal. Validated BEFORE sealing: differential recovers 1.0000 in both the Earth-like and Mars-like regimes while the estimated floor returns 2.3828 in the Earth-like one. UNCHANGED: band 0.01-1 Hz, df/f 1e-6, N_bins 4.605e6, statistic gamma2, Sidak z 5.593, INJ_TOL (0.75,1.33), xi=1, R=1.8796, projection Z, wind percentile 50, no circular-shift null, seed = sha256(EARLIEST attesting Bitcoin block hash of THIS anchor), no real bin scored before confirmation. expected_direction=NULL_BOUND.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_line_search"},"reevaluate_at":null,"registered_at":"2026-09-03T02:15:56Z","registered_by":"Mljessop@deepmap-compute","result":null,"status":"locked","test":{"extra":{"amendment_v1_2":{"change":"the injection POSITIVE CONTROL recovers the line by DIFFERENTIAL power on the same patches, phase-averaged over N_INJ_PHASES=4, instead of subtracting an ESTIMATED local floor. rec = sqrt(max(mean_j <|A(inject at psi+2*pi*j/N)|^2> - <|A(un-injected)|^2>, 0)).","cross_term":"the power difference carries 2*Re(A_inj conj(A_amb)), zero ONLY in phase average; measured 1 phase -> 1.1585, 4 phases -> 1.0000","estimated_floor_still_reported":true,"inj_tol_unchanged_reason":"widening a positive-control tolerance after failing it is goalpost movement and is refused","n_inj_phases":4,"unchanged":["band","df_over_f","n_bins","statistic","bar_sidak_z","xi","ratio","projection","wind_keep_percentile","seed_rule","fft_path_and_equivalence_gate","empirical_null_policy","single_body_hard_rule","coincidence_test","INJ_TOL"],"why":"v1.1 halted with per-arm recovered/injected E=3.708 (M=1.000) vs INJ_TOL (0.75,1.33). Measured: Earth injected power 6.349e-17, estimated local floor 2.447e-15 (38.5x the injection), total mean power 3.320e-15 (52.3x) -- the floor explained only 73.7% of the arm's power. For rE to land in tolerance the ESTIMATED floor must be accurate to 2.3% of the total power, which a 60-dummy neighbour-frequency median cannot deliver when the quantity estimated is 38x the signal. Mars passes trivially at 835x ABOVE its floor -- a ~4.6-order Earth/Mars SNR asymmetry. Structural (INJ_GAMMA2_MARGIN=2 puts the injection at 2x the discovery bar, deep inside Earth's ambient), not a coding error, and not spectral (0.05 Hz sits near the seismic noise notch)."},"code_git_commit":"a68508b71df885bcd7906b57c2f3414a9fa9df5f","code_path":"scripts/dm_earthmars_coherence_v12.py","code_sha256":"c1d04a58f3d2fb47f5c87f409a5e082e45870bf6c83824cec068bdaa0afe69b8","discovery_rule":"candidate iff analytic Sidak z >= 5.593 over N_bins=4.605e6 (Carter/Nuttall MSC null) on FFT-evaluated lock-in amplitudes (in-run equivalence gate on 40 seeded patches x 600 bins: <= 1e-3 rel amplitude error vs patch_amplitude, else exit 6) AND empirical phase-scramble p <= 0.05 with a live null AND single-body HARD rule (one-body analytic z < bar AND median scramble p over 5 one-body draws > 0.05) AND in-run positive control: per-arm recovered/injected in [0.75,1.33] using DIFFERENTIAL recovery (v1.2: exact ambient cancellation on the same patches, averaged over 4 injection phases) at the injection amplitude whose expected gamma2 = 2 x gamma2*(K), joint z >= 5.593; else NULL.","empirical_null_policy":"phase-scramble >=500 draws on every candidate + 10,000-bin calibration subsample x 200 draws (aliveness + FAR reported, never promoting)","evaluation":"fine spectrum per patch (pad 32, complex64) + cubic interpolation in 200,000-bin chunks","expected_direction":"NULL_BOUND","expected_outcome":"FIRST cross-planet DM line search; a NULL reported as a 95% exclusion on the observable per-body line amplitude (m/s^2) = the injected amplitude scaled to gamma2*(K), + the cross-coherence bound + the Earth-Mars coincidence veto; coupling (cm^2/epsilon) DEFERRED to a theorist.","frozen_config":{"amendment_v1":{"empirical_null_policy":"phase-scramble on every candidate (>= 500 draws) + 10000-bin calibration subsample (200 draws) reported, never promoting","equivalence_bins":600,"equivalence_patches":40,"evaluation":"FFT path (patch_fine_spectrum + cubic interpolation) reproducing patch_amplitude; gated by --stage equivalence in-run","fft_equiv_tol":0.001,"fft_pad":32,"injection_amplitude":"amp_E such that expected gamma2 = 2.0 x gamma2*(K); amp_M = amp_E / ratio","injection_floor":"LOCAL: 60 log-uniform dummy freqs within +/-20% of f_inj","scan_chunk_bins":200000,"scored_population":"prepared store of the frozen core (MARS_BASELINE_START..END), daily-chunked, wind percentile applied over the whole baseline","store_version":"v1","unchanged":["band","df_over_f","n_bins","statistic","bar_sidak_z","single_body_hard","inj_tol","xi","ratio","projection","wind_keep_percentile","seed_rule"]},"band_hz":[0.01,1.0],"bar_sidak_z":5.593,"df_over_f":1e-6,"earth_duty_floor_rule":"drop a station BEFORE scoring if its 2020-2021 duty < the pre-declared floor (frozen rule, not post-hoc)","earth_stations":["IU.ANMO.00.BHZ","II.BFO.00.BHZ","IU.TSUM.00.BHZ","IU.CCM.00.BHZ"],"global_alpha":0.05,"injection_tol":[0.75,1.33],"mars":{"chans":["BHU","BHV","BHW"],"loc":"02","net":"XB","sta":"ELYSE"},"mars_baseline":["2020-01-01T00:00:00","2021-12-31T23:59:59"],"n_bins":4605170.185988092,"null_battery":{"circular_shift":"EXCLUDED","dummy_floor":150,"phase_scramble":500,"time_shift":500},"projection":"Z","ratio_E_over_M":1.8796282637557162,"seal_anchor_id":null,"seed_rule":"science RNG seed = sha256(confirmed OTS Bitcoin block hash of the recorded anchor); forward-blind, set AT SEAL","single_body_rejection":"HARD","wind_keep_percentile":50.0,"xi_elastic":1.0},"pre_run_validation":{"bars_required_and_met":["failure regime reproduced (>10x ambient, floor <95%)","differential ~1.0 in BOTH regimes","estimated floor FAILS the same case (fix is doing real work)"],"disclosure":"the FIRST version of this test used WHITE ambient, where a neighbour-frequency floor is unbiased; v1.1's method passed it (1.0020) and the test correctly returned NOT PROVEN. The rewritten test uses red noise + a local bump and ASSERTS the regime was reproduced before believing any verdict.","earth_like_coloured_ambient":{"ambient_over_injection":11.7,"floor_explains_pct":51.6,"v11_estimated_floor":2.3828,"v12_differential":1.0},"exit_code":0,"mars_like":{"v11_estimated_floor":0.9999,"v12_differential":1.0},"method":"both recovery methods driven through the v1.2 module's own functions (real call path)","one_phase_control":{"v12_differential":1.1585},"script":"scripts/test_dm_earthmars_v12_recovery.py"},"prereg_path":"_active_today/dm_earthmars_coherence_prereg_v1_2_AMENDMENT.md","prereg_sha256":"db56bee072ceae5845c2e0bdedc095addc18f971febf67ab2db9c68312e33754","scored_population":"the SAME prepared store v1.1 would have scored, REUSED with no re-prepare: frozen 2020-01-01..2021-12-31 core, 730 day files, 723 days with data, 2884 candidates, 1442 kept (wind p50), manifest sha256 23b964f7b9d0. NO real bin has been scored by any version: v0 and v1.1 both halted at their positive control, v1 never ran.","supersedes":["20260902_earth_x_mars_cross_planetary_ult_1084ce55 (v0, HALTED at its in-run gate, zero bins scored)","20260902_earth_x_mars_cross_planetary_ult_75974252 (v1, WITHDRAWN before any run: scale defects)","20260902_earth_x_mars_cross_planetary_ult_813fc905 (v1.1, SEALED + BTC-anchored block 965229 + RAN, HALTED at its injection positive control, ZERO real bins scored, UNTESTABLE-AS-FROZEN not a null)"]},"method":"fft_cross_coherence_sidak_scan_v1_2","n_permutations":500},"threshold":{"direction":"","metric":"sidak_z","value":5.593},"title":"Earth x Mars cross-planetary ultralight-DM Compton-line coherence search v1.2 (differential positive control; supersedes v1.1 813fc905, v1 75974252, v0 1084ce55)","withdrawn":null,"_ots_anchored":false,"_status_dir":"locked","_re_eval_date":null},{"hypothesis_id":"20260903_earth_x_mars_cross_planetary_ult_01fea7ac","bitcoin_block_hash":"00000000000000000001d1e995331925ad215b193702fb4bbfed10fa582f673c","bitcoin_block_height":965248,"rule":"EARLIEST Bitcoin block among all BitcoinBlockHeaderAttestation entries in the upgraded .ots (D1)","block_hash_sources":["blockstream.info","mempool.space (agree)"],"code_sha256":"c1d04a58f3d2fb47f5c87f409a5e082e45870bf6c83824cec068bdaa0afe69b8","seed":"3d397dffca3142f055e131587dd20019567041e1d2d94825a3850ec26fbc1345","decided_utc":"2026-09-03T04:18:16Z","note":"SIDECAR ONLY -- canonical locked .json untouched","_ots_anchored":false,"_status_dir":"locked","_re_eval_date":null},{"hypothesis_id":"20260904_earth_x_mars_cross_planetary_ult_2a4273ed","bitcoin_block_hash":"00000000000000000000d7ba6917def943b0f07f8c029c62c1574f1da3a8c04c","bitcoin_block_height":965400,"rule":"EARLIEST Bitcoin block among all BitcoinBlockHeaderAttestation entries in the upgraded .ots (D1)","block_hash_sources":["orchestrate_v13.sh at confirmation (blockstream.info / mempool.space)"],"code_sha256":"4958f6560c8ed12fa9c79b034ffaab4ab8d792c69ef36946bac79d61656b0d93","seed":"29999e41d3e5cbef45640f56b3fabdf512e37c1cb33cdfb615221e0e90c207d7","decided_utc":"2026-09-04T01:04:47Z","note":"SIDECAR ONLY -- canonical locked .json untouched. Key is bitcoin_block_hash (the scan's _confirmed_block_hash reads THIS key); the first-write sidecar used block_hash and the scan refused. Seed = sha256(bitcoin_block_hash), forward-blind.","_ots_anchored":false,"_status_dir":"locked","_re_eval_date":null}],"tested":[{"data_source":"data/prediction_ledger.db,common/magnitude_utils.py","expected_direction":"positive","feature_set":["expected_mag_min","REGION_FLOORS={iceland:2.5","azores:2.5","newberry:2.5","ukraine:2.5}"],"git_commit_at_registration":"3285f15dcbbf70849990986aec6770ef2b289859","hypothesis_id":"20260504_lowering_expected_mag_min_to_2_5_5039cbab","locked_at":"2026-05-04T20:05:22Z","locked_file_sha256":"696823c64ad6d98ac6c68d811d1db90fda5aae21dcf310f8943cbaf0c60cf73b","notes":"INTERVENTION hypothesis. Diagnosis 2026-05-04: 4 regions (iceland/azores/newberry/ukraine) have 0% hit rate over 30d on n=23 predictions across all models. All 4 are low-magnitude zones (volcanic-rift or stable-interior) where typical events are M2-4; our default expected_mag_min derived from risk_level is 3.0-4.5. Many real M2.5-3.5 events go uncredited because they fall below the prediction floor and get tagged magnitude_shortfall. Intervention: add REGION_FLOORS dict to common/magnitude_utils.py mapping these 4 regions to 2.5, applied in compose_expected_range as min(bucket_mn, region_floor). Forward test: re-evaluate on 2026-08-04 by querying predictions in these regions filed after 2026-05-04 deployment, target >=30% hit rate AND global EQ hit rate within +-2pp of pre-intervention 54.3%. If both: confirming. If hit rate climbs but global degrades by >2pp: partial-confirming (regional gain at unacceptable cost). If regional hit rate stays <30%: disconfirming (the magnitude floor is not the binding constraint).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"earthquake"},"registered_at":"2026-05-04T20:02:15Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"76c275a3812c6853dc7c47d0144972596398ac8c","interpretation":"confirming","metric_value":0.77,"n_observed":187,"notes":"CONFIRMING on the primary criterion, graded 2026-08-11 (7d past the 2026-08-04 due date; the registry had no due-date alert until today).\n\nRESULT: regional hit rate 7.6% (7/92, pre 2026-05-04) -> 77.0% (144/187, post). Target was >=30%. Per region: ukraine 90.6% (29/32), azores 78.3% (54/69), iceland 77.2% (61/79), newberry 0.0% (0/7) -- newberry is a clear NON-RESPONDER and the intervention did nothing there.\n\nMECHANISM CHECK (the criterion alone is nearly tautological -- lowering a magnitude floor mechanically makes matches easier, so a hit-rate rise is close to guaranteed). The discriminating test is whether the newly-credited events are the low-magnitude ones the 2026-05-04 diagnosis predicted. They largely are: 137 of 404 post-intervention match rows are events below the old M3.0 floor. The proposed mechanism is doing real work.\n\nGLOBAL GUARD UNINTERPRETABLE: criterion required global EQ hit rate within +/-2pp of the pre-intervention 54.3%. Actual is 70.5%, i.e. +16.2pp -- outside the band but in the FAVOURABLE direction. Over a 3-month window too much else changed for this to serve as a clean control. The guard was written to catch degradation; there is none. Recorded as neither satisfied nor violated.\n\nMATERIAL CAVEAT: hit rate is COVERAGE, not skill. Measured the same day (scripts/eq_forecast_skill.py, reports/eq_forecast_skill_measured.md): tau=85.6%, Area Skill Score 0.4877 vs 0.5 for no skill, pooled BSS -0.0629. So this result confirms the intervention did what it was designed to do -- credit real low-magnitude events that were being tagged magnitude_shortfall -- and is NOT evidence that forecasting improved. Do not quote 77.0% as accuracy.","ots_proof_path":null,"passed":true,"tested_at":"2026-08-11T17:17:12Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"before_after_30d_window_per_region","n_permutations":null},"threshold":{"direction":"positive","metric":"regional_hit_rate_lift","value":0.3},"title":"Lowering expected_mag_min to 2.5 for iceland/azores/newberry/ukraine raises hit rate in those 4 regions from 0% to >30% over 90d without degrading global rate","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-08-04"},{"data_source":"data/retro_cache/iris_mustang.db","expected_direction":"positive","feature_set":["microseism_secondary_band_3s"],"git_commit_at_registration":"d16b85a03ec88f79033f669e7c0228dc12ad3dec","hypothesis_id":"20260504_microseism_universal_precursor_c_99f4c2c7","locked_at":"2026-05-04T19:36:20Z","locked_file_sha256":"9283187e030940a076c96e14e368915b907b7660203372b16742e630d8a43782","notes":"FORWARD replication of Phase 5 (Apr 25 2026; +19.12sigma post-control on N=7,120 M>=5.5 EQs). This entry pre-registers the hypothesis that the same effect will persist on EVENTS THAT OCCUR BETWEEN 2026-05-04 AND 2026-11-04, evaluated against the live microseism feed. Result will be recorded after Nov 4 2026 using events that did not exist at lock time. This is a literature-replication forward test, not a self-disconfirmation candidate.","ots_proof_path":null,"outcome":{"lookahead_hours":168.0,"min_magnitude":5.0,"type":"earthquake"},"registered_at":"2026-05-04T19:35:17Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"673dff0ac5e4841da96c6dd84964911ac8411165","interpretation":"disconfirming","metric_value":0.08,"n_observed":82,"notes":"DISCONFIRMED by independent verification wave (trackAR, 2026-08-24). The feature microseism_secondary_band_3s is MUSTANG pct_above_nhnm -- a BROADBAND whole-day metric that any earthquake's own energy inflates. The z is earthquake self-contamination: the event-day time profile spikes at k=0 with SYMMETRIC foreshock/aftershock shoulders (not a pre>>post precursor), and on a daily feature '-10h' lands on the same UTC day as the event 58% of the time. On the un-biased band_rms (0.14-0.5 Hz, hourly) metric with a clean baseline + declustering, -10h z = +0.08/-0.23/-0.08 across magnitude bins with NO scaling -- decisively excluded (a real precursor at n=82 M7.5+ would give t~13); positive control event-hour z=+30.3. Reproduces the sealed positive control exactly (19.02 sigma), so the effect is real AS COMPUTED but is the earthquake, not an ocean-microseism precursor. Reports: ~/trackAR/trackAR_report.md, ~/trackAN/.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-24T20:27:57Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"stacked_epoch_z","n_permutations":10000},"threshold":{"direction":"positive","metric":"z_score","value":5.0},"title":"Microseism universal precursor continues to fire at z>=5 for M5+ EQ at -10h, May-Nov 2026","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-11-04"},{"data_source":"Utah DOGM WIS API + cross_physics.og_convergence_monitor","expected_direction":"positive","feature_set":["ogc_composite_score"],"git_commit_at_registration":"e182c4a2d0e4b9476bfeedd18746df4af9db0c8a","hypothesis_id":"20260511_non_dry_historical_sevier_desert_865aa86c","locked_at":"2026-05-11T02:30:33Z","locked_file_sha256":"d0131f86634ec20a014d2a712c3426d153a231661ed1c84cd04ee4591e11f7a0","notes":"Predictive Geology Phase 3 of 3 (companion to papers/predictive_geology_validation_design.md §4 H3 -- directional check). PRE-REGISTRATION REQUIRED before any DOGM ingest. Among the ~3000 historical Sevier Desert / Pavant / Black Rock / Beaver County wells, those classified OIL or GAS by DOGM (non-dry; expected count ~20-50) have higher OGC composite_score than dry wells at p<0.05 (Mann-Whitney). This is a SMALL-N DIRECTIONAL check, not a primary hypothesis -- with non-dry sample size likely 20-50, statistical power is limited. Reports effect-size estimate + CI regardless of p-value. Failure interpretation: small N means absence of significant difference may reflect insufficient power rather than null effect. Re-eval target: 2026-08-31.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"oil_gas_well_production_class"},"registered_at":"2026-05-11T02:30:24Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"8ebb5217e6d0eba124fb857b6c8f4f98851d7ccb","interpretation":"disconfirming","metric_value":-0.4088,"n_observed":80,"notes":"GRADED 2026-09-14. Sealed test run as specified: Mann-Whitney U one-sided (non-dry > dry) on ogc_composite_score, metric rank_biserial_correlation_dry_lt_nondry, threshold >=0.30 with p<0.05. RESULT: r = -0.4088, 95% bootstrap CI [-0.6366,-0.1587] (5000 resamples), U=449.0, p=0.9989. DRY n=49 median 64.37; NON-DRY n=31 median 59.48. FAIL, and the effect is REVERSED: producing wells score LOWER than dry holes and the CI excludes zero, so the OGC composite is anti-correlated with productivity in the Sevier Desert. The seal declared its power limited in advance and pre-committed to reporting effect size + CI regardless of p; non-dry n=31 is inside its own stated 20-50 window, so this is a measured reversal rather than an underpowered null. HANDLING: 52 of 132 wells in the sealed bbox returned ogc_data_real=0 and were EXCLUDED, never scored 0 -- counting nulls as zeros would have pushed the DRY group down and manufactured the predicted effect. DEVIATION (protocol): graded with the HEAD scorer, NOT the lockfile's git_commit_at_registration (e182c4a2), because that sealed commit returns ogc_data_real=0 with 0/5 axes at every point in the target area and can produce no measurement. HEAD also carries the 2026-09-14 fault-catalogue repair (USGS Qfaults restored in place of a 40-entry inline fallback that was wrong by ~30x at Sevier), verified present in this run by fault_km median 14.45 km.","ots_proof_path":null,"passed":false,"tested_at":"2026-09-14T16:47:36Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"mann_whitney_u_test_dry_vs_nondry","n_permutations":null},"threshold":{"direction":"positive","metric":"rank_biserial_correlation_dry_lt_nondry","value":0.3},"title":"Non-dry historical Sevier Desert wells have higher OGC composite_score than dry wells (Predictive Geology H3 directional)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-08-31"},{"data_source":"cross_physics.og_convergence_monitor + USGS/UGS petroleum geology maps","expected_direction":"positive","feature_set":["ogc_composite_score","ogc_nearest_fault_km"],"git_commit_at_registration":"e182c4a2d0e4b9476bfeedd18746df4af9db0c8a","hypothesis_id":"20260511_top_decile_ogc_cells_in_sevier_d_9c8de680","locked_at":"2026-05-11T02:30:33Z","locked_file_sha256":"8837ca2394016a3c0fe09381632501625f85c29438df96448a62b5ddfb74dac5","notes":"Predictive Geology Phase 2 of 3 (companion to papers/predictive_geology_validation_design.md §4 H2). PRE-REGISTRATION REQUIRED before any USGS trap-map digitization. Top 10% of cells by OGC composite_score across Sevier Desert basin (cells inside bbox 38.0-40.0 N, -113.0 to -111.5 W) cluster within 10 km of mapped basin-margin frontal-fault structures (Pavant frontal, House Range frontal, Mineral Mountains frontal). Random baseline expects ~25% of cells within 10 km of any frontal fault; top-decile expectation is >=60%. One-sided binomial test, p<0.001 pre-specified. 10% of ~10000 corridor cells = ~1000 top-decile cells. Effect size 60% vs 25% baseline trivially exceeds power at p<0.001. Failure interpretation: OGC top scores spread uniformly across basin = no fairway discrimination. Re-eval target: 2026-08-31.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"oil_gas_structural_trap"},"registered_at":"2026-05-11T02:30:13Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"4627deb1a6df77ee8f011c55beba4431efb66e30","interpretation":"disconfirming","metric_value":0.08,"n_observed":25,"notes":"GRADED 2026-09-14 on a converged uniform grid. Sealed metric: fraction_of_top_decile_cells_within_10km_of_frontal_fault, threshold >=0.60, one-sided binomial p<0.001, assumed baseline ~25%, frontal structures named as Pavant / House Range / Mineral Mountains. RESULT on the seal's OWN named structures (PRIMARY set): 2/25 = 0.080 -- FAIL. Measured baseline over all scored cells is 0.096, so top-decile cells are if anything slightly LESS likely to sit near the named frontal faults than average cells; binomial vs the sealed 0.25 gives p=0.993. GRID: 500-cell uniform lattice over the sealed bbox 38.0-40.0N/-113.0--111.5W, 250 scored with real data (50%), 250 nulled by the LEI sediment gate and EXCLUDED rather than scored 0; composite over scored cells min 12.7 / median 61.4 / max 75.7; top decile = 25 cells. MINERAL MOUNTAINS IS ABSENT from the USGS Qfaults Utah names entirely, so PRIMARY is 2 of 3 named structures and can only UNDERSTATE -- but it cannot change the verdict: reaching 0.60 needs 15/25, and a 12x larger superset of named basin-margin normal faults (RANGE_FRONT, 1502 segments vs 118) still reaches only 9/25 = 0.360. The missing third structure cannot bridge that gap, so this is a measured negative with a bound, not an untestable seal. WHY THE FAULT SETS WERE DECLARED BEFORE SCORING: on the ANY-Quaternary-fault definition -- 3000 segments, and the field the pipeline conveniently exposes as ogc_nearest_fault_km -- the same cells give 21/25 = 0.840 with p=9.7e-10, a headline PASS. The seal does not ask about any fault; it names three basin-margin frontal structures. Grading on the convenient field would have answered a different question in the direction of PASS. DEVIATIONS: (1) graded with the HEAD scorer, not git_commit_at_registration e182c4a2, which returns ogc_data_real=0 with 0/5 axes at every point in the target area; (2) 500 uniform cells instead of the planned ~10,000 -- a uniform subsample of the same population, so its top decile is an unbiased estimate, unlike the pre-existing 190-cell HiRes/ELITE file which is selected on the very quantity being ranked; (3) HEAD carries the 2026-09-14 fault-catalogue repair (USGS Qfaults restored in place of a 40-entry inline fallback wrong by ~30x at Sevier). Artifacts: data/private/predictive_geology_2026_09_14/.","ots_proof_path":null,"passed":false,"tested_at":"2026-09-14T19:05:36Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"binomial_test_proportion_within_10km_of_frontal_fault","n_permutations":null},"threshold":{"direction":"positive","metric":"fraction_of_top_decile_cells_within_10km_of_frontal_fault","value":0.6},"title":"Top-decile OGC cells in Sevier Desert corridor cluster within 10 km of mapped basin-margin frontal-fault structures (Predictive Geology H2)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-08-31"},{"data_source":"data/retro_tensor.db (2005-2018, N=122712 hourly snapshots) + data/pbh_known_events_cache.db (M5.5+ events, N=7120)","expected_direction":"positive","feature_set":["microseism_primary_band_6s","microseism_secondary_band_3s"],"git_commit_at_registration":"67881aee2e934e69d2254e5921f9abffadbdb53b","hypothesis_id":"20260514_phase_5_microseism_universal_pre_dad7e9e7","locked_at":"2026-05-14T06:29:33Z","locked_file_sha256":"544ebcc1f96d5acc10216cf9c08de6d4d0ae65627969c7ae6b186cbfc8b353bf","notes":"CORRECTED ADDENDUM to parent hypothesis 20260504_microseism_universal_precursor_c_99f4c2c7. Supersedes withdrawn hypothesis 20260514_phase_5_microseism_universal_pre_41a196bf (which contained a coordinate-mapping error in interpretation -- see that withdrawal note for full disclosure). Two independent 2026-05-13/14 tests re-confirm the original Phase 5 narrative WITHOUT modification. (1) Forward-control stacked-epoch on retro tensor, N=7,120 M5.5+ events: peak z = +34.52 at lag -10h, z@+10h = +2.95 -> 11.7x pre-event asymmetry. Magnitude-stratified runs: M4.5+ N=102,632 peak z=+38.84 at lag -10h; M6.5+ N=690 peak z=+30.25 at lag -10h; M7.5+ N=82 peak z=+19.79 at lag -10h. Phase 5 finding is universal across all magnitude bands. (2) Time-D entropy classifier on same retro tensor: microseism_primary_band_6s is THE single most-irreversible feature in the entire 112-dim tensor (R^2 = -0.216 -- post-event tensor state predicts pre-event state WORSE than chance), and microseism_secondary_band_3s is the strongest non-solar irreversible feature. Two independent methodologies, same result: microseism is the dominant pre-rupture entropy-producing signal. PASS CRITERION: forward test on M5.5+ events 2026-05-14 to 2026-11-14 (6 months), median z at lag -10h >= 5 across N>=50 events AND directional asymmetry ratio z@-10h / z@+10h >= 3 across same N. Methodology files: scripts/eq_precursor_stacked_epoch_time_reversed.py (with --forward-control flag), scripts/time_reversed_entropy_classifier.py. Result JSONs: reports/science/eq_precursor_epoch_forward_control_20260514T035034Z.json, reports/science/eq_precursor_epoch_time_reversed_20260514T0611{59,300,309}Z.json (M4.5/M6.5/M7.5 strata), reports/science/time_reversed_entropy_20260514T033934Z.json.","ots_proof_path":null,"outcome":{"lookahead_hours":168.0,"min_magnitude":5.5,"type":"earthquake"},"registered_at":"2026-05-14T06:29:26Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"673dff0ac5e4841da96c6dd84964911ac8411165","interpretation":"disconfirming","metric_value":0.08,"n_observed":82,"notes":"DISCONFIRMED by independent verification wave (trackAR, 2026-08-24). The feature microseism_secondary_band_3s is MUSTANG pct_above_nhnm -- a BROADBAND whole-day metric that any earthquake's own energy inflates. The z is earthquake self-contamination: the event-day time profile spikes at k=0 with SYMMETRIC foreshock/aftershock shoulders (not a pre>>post precursor), and on a daily feature '-10h' lands on the same UTC day as the event 58% of the time. On the un-biased band_rms (0.14-0.5 Hz, hourly) metric with a clean baseline + declustering, -10h z = +0.08/-0.23/-0.08 across magnitude bins with NO scaling -- decisively excluded (a real precursor at n=82 M7.5+ would give t~13); positive control event-hour z=+30.3. Reproduces the sealed positive control exactly (19.02 sigma), so the effect is real AS COMPUTED but is the earthquake, not an ocean-microseism precursor. Reports: ~/trackAR/trackAR_report.md, ~/trackAN/.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-24T20:27:57Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"cross_validation":"time_d_entropy_R2_at_microseism_primary_band_6s_lt_zero","directional_asymmetry_test":"z_at_minus_10h_over_z_at_plus_10h","min_asymmetry_ratio":3.0,"min_peak_z_at_lag_minus_10h":5.0,"window_h":[-168,24]},"method":"stacked_epoch_z_forward_control","n_permutations":10000},"threshold":{"direction":"positive","metric":"z_at_lag_minus_10h","value":5.0},"title":"Phase 5 microseism universal precursor reproduces at z=+34.52 at lag -10h with 11.7x pre>>post asymmetry; Time-D entropy classifier independently confirms microseism is most-irreversible feature in 112-dim retro tensor","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-11-14"},{"data_source":"data/private/submarine_springs/ancient_mariner_catalog.json + data/private/submarine_springs/control_sample_catalog.json (53 historical + 53 control, deterministic seed 20260517)","expected_direction":"positive","feature_set":["composite_score","composite_minus_axis4","axis_mt_contrast","axis_tidal_em","axis_mcf_aquifer","axis_karst_proximity","axis_hydroacoustic"],"git_commit_at_registration":"","hypothesis_id":"20260518_coastal_sites_cited_in_pre_1900__873c2e65","locked_at":"2026-05-18T01:23:36Z","locked_file_sha256":"ae1bd2aaf11f96a1a185a9eca224bd871c731d45688efd6ea14491900e939432","notes":"Coastal sites cited in pre-1900 historical records as 'good water' landings (Manila galleon resupply logs, Alta California + Baja California Jesuit/Franciscan mission siting decisions, Seri/Yaqui/Cucapa indigenous oral tradition as documented in Felger & Moser 1985 and Castetter & Bell 1951, Greco-Roman classical texts including Strabo + Pliny + Pausanias, Dilmun cuneiform tablets ~2500-2000 BCE) score significantly higher on the 5-axis SGD composite (cross_physics.submarine_freshwater_monitor.score_sgd_5axis) than randomly-sampled coastal control sites within the same geographic province, when evaluated at the pooled level with Welch's t-test p < 0.05 AND with Bonferroni correction across the 6 regional sub-cohorts (alpha_per_region = 0.05/6 = 0.0083). HYPOTHESIS: pre-1900 mariners encoded survivorship-filtered ground truth about coastal freshwater availability in their navigational records; modern multi-physics SGD scoring should recover what mariners and indigenous peoples learned by direct experience.\n\nROBUSTNESS REQUIREMENT: because score_sgd_5axis axis 4 (karst_proximity) computes proximity to KNOWN_SGD_SITES in submarine_freshwater_monitor.py, and 4 historical catalog sites (Chekka, Vrulja, Mortola, Port Miou) appear in or near that atlas, the karst-proximity axis carries a self-fulfilling bias toward the historical cohort. The pre-registered analysis therefore tests BOTH composite_score (primary) AND composite_minus_axis4 (robustness, re-weighted to 1.0 over the 4 remaining axes). The test PASSES only if BOTH metrics show p < 0.05 with Bonferroni correction.\n\nPASS CRITERION: pooled Welch's t-test p < 0.05 AND >= 3 of 6 regional sub-cohorts pass the per-region Bonferroni-corrected threshold (p < 0.0083) AND the composite_minus_axis4 robustness Welch t p < 0.05.\n\nFAIL CRITERION: either pooled p >= 0.05, or fewer than 3 regional sub-cohorts pass, or composite_minus_axis4 robustness fails. Failure means physics-based scoring does NOT recover historical attestation -- either our 5-axis scoring is insensitive to coastal freshwater signals, or historical attestation is noisier than we assumed, or both. Either is a publishable result.\n\nWITHDRAW RULE: if score_sgd_5axis returns n_axes_real_data < 2 for >50pct of either cohort (live-data starvation, e.g. IRIS or Sentinel-2 outage), declare underpowered and withdraw; re-register after data path is restored.\n\nNULL DIRECTION: if pooled historical mean is LOWER than control mean (negative t-stat), the result is informative-negative (mariners may have been water-blind for the wrong physical reason, OR our scoring is anti-correlated with their criteria); we report and interpret rather than re-tune.\n\nFORWARD-LOOKING: positive result strengthens the case that Phase 65w-VERIFY SGD scoring should be deployed for new prospecting (Sonoran coast Mexico, Hermosillo/Puerto Penasco area, Persian Gulf). Negative result triggers axis-by-axis diagnostic to identify which sensing modality fails on historically-attested coastal freshwater.\n\nCULTURAL NOTE: indigenous knowledge entries cite anthropological literature (Felger & Moser 1985 'People of the Desert and Sea', Castetter & Bell 1951 'Yuman Indian Agriculture', Spicer 1980 'The Yaquis', Beals 1945 'Cahita Indians') as the authoritative voice for water-source knowledge. The hypothesis is NOT that DeepMap physics 'discovered' these sources; the hypothesis is that DeepMap physics recovers what those communities learned and recorded by direct experience.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"submarine_groundwater_discharge"},"registered_at":"2026-05-18T00:00:00Z","registered_by":"Mljessop@DeepMap","result":{"git_commit_at_test":"ffec46b5aaad6d2be24d7d74c7a7dc0a47de7c16","interpretation":"disconfirming","metric_value":0.00002,"n_observed":106,"notes":"Pooled Welch t p=0.00002 SIGNIFICANT but 3-condition pre-registered gate FAILS: only 1/6 regions pass Bonferroni (baja_california p=0.0074); robustness metric composite_minus_axis4 NULL (p=0.77). Effect attributable to axis-4 self-similarity bias. Baja California sub-signal genuine. Publishable as methodological null per Section 7 nulls paper.","ots_proof_path":null,"passed":false,"tested_at":"2026-05-18T02:11:36Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{"alternative_hypothesis_direction":"historical > control","bonferroni_alpha_per_region":0.0083,"deterministic_control_seed":20260517,"n_control":53,"n_historical":53,"n_regional_sub_cohorts":6,"primary_metric":"composite_score","robustness_metric":"composite_minus_axis4","scoring_function":"cross_physics.submarine_freshwater_monitor.score_sgd_5axis","scoring_version":"v1_phase65w_verify","stratification_variable":"region (manila_galleon_route, alta_california, baja_california, sonora_seri_yaqui, mediterranean_classical, persian_gulf_dilmun)"},"method":"welch_t_two_sample_with_mann_whitney_u_backup","n_permutations":null},"threshold":{"direction":"positive","metric":"welch_t_pooled_p_value","value":0.05},"title":"Coastal sites cited in pre-1900 historical records as 'good water' landings score higher on the 5-axis SGD composite than geographically-matched random control sites","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"IGETS L2 (EOST + ICET when available) 2009-2026 + D.1-SPRINT correction pipeline (Hinderer 1995 atmospheric loading + FG5 K-scale registry + joint LS 12-constituent tidal subtraction)","expected_direction":"negative","feature_set":["slm_v4_matched_filter_L","slm_v4_n_stations_in_stack","slm_v4_bootstrap_p001_threshold","slm_v4_canonical_period_4_8h_L","slm_v4_canonical_period_5_4h_L","slm_v4_canonical_period_6_0h_L"],"git_commit_at_registration":"45e5ad7b9476ffcaf80327bf9d31eaa16fd7c253","hypothesis_id":"20260520_slichter_mode_v4_methods_clean_n_5af2ca52","locked_at":"2026-05-20T01:55:54Z","locked_file_sha256":"c389a846ef246604a54790682052aa40e9ba3e3cc5896d02ebbb2daf8aab68fe","notes":"D.1-SPRINT v4 forward test: pre-registered BEFORE running v4 pipeline.  re-eval YYYY-MM-DD = 2027-05-19.  Two outcomes:  (1) DETECTION at L > p001 in >=2 canonical periods + triplet shape match -- Nature paper; (2) PUBLISHABLE NULL with sigma constrained <= 0.1 microGal at all 3 periods (5x tighter than Rosat 2005) -- JGR Solid Earth / GJI / EPSL paper as methods-clean update to the 65-year Slichter null.  Prior v3 result on 6 stations was NULL with detection_score=3.0 (MARGINAL).  v4 expects ~6.4x raw sensitivity gain from 24-station + matched-filter combined.  Pre-eval 2026-11-19 (6 months) recommended to check if any new data shifts the picture; full re-eval 2027-05-19.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"matched-filter-likelihood threshold test"},"registered_at":"2026-05-20T01:55:41Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"5eec07b1f1a274ae26d3a1966228e032f87d907d","interpretation":"disconfirming","metric_value":0.18033,"n_observed":15,"notes":"Cross-seed 50k bootstrap (seeds 42/43/44/45) gives L_stacked=0.18033 > p<0.001 threshold 0.17296-0.17353 at all 4 seeds (+3.9-4.3% margin), so the matched-filter statistic alone classifies as DETECTION. However triplet-splitting consistency check FAILS: 0 of 3 canonical periods pass loose p<0.05. Per-canonical L: 4.8h=0.0094 / 5.4h=0.0087 / 6.0h=0.0135 (all 5-7% of stacked 0.18033); L_stacked/max(per-canonical) = 13.3x (real triplet would be ~1x). Verdict TRIPLET_INCONSISTENT. Most plausibly broadband band-power detection not triplet-resolved Slichter modes. Likely physical explanation: residual long-period atmospheric/oceanic loading not fully removed by D.1-SPRINT Hinderer+FG5+joint LS pipeline. 15 of 24 IGETS stations contributed; top-3 Strasbourg(0.371)/Walferdange(0.351)/Onsala(0.345) carry 39.5% but broad coverage (stations 4-7 each >10%); not outlier-driven. The pre-reg required triplet-splitting consistency for DETECTION; without it the hypothesis FAILS the discovery criterion. Publishable as methods-clean update to the 65-year Slichter null + DETECTION CANDIDATE paper (JGR Solid Earth / GJI / EPSL), NOT as discovery. See reports/slichter_v4_post_detection_verification_2026_05_20.md for full disposition. Verification commit: 39dab27b.","ots_proof_path":null,"passed":false,"tested_at":"2026-05-20T14:30:54Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"bootstrap_method":"Welch random-phase surrogate (Theiler 1992)","bootstrap_n_perm":50000,"calibration_pipeline":"D.1-SPRINT v3.1+3.2 (Hinderer 1995 + FG5 + joint LS + 24-station + matched filter)","canonical_periods_hr":[4.8,5.4,6.0],"detection_statistic":"matched_filter_L_likelihood","integration_years":5,"matched_filter_basis":"slichter_triplet_template (slichter_matched_filter.py)","predecessor":"slichter_mode_search_monitor v3 (commit b3e6bde3) -- NULL on 6-station 5-year stack","stations":["strasbourg","wettzell","moxa","conrad","cantley","apache_point","metsahovi","kamioka","canberra","sutherland","onsala","pecny","walferdange","larzac","schiltach","trappes","membach","bad_homburg","medicina","concepcion","hsinchu","mizusawa","lhasa","djougou"]},"method":"Slichter v4 matched-filter L statistic at all 3 canonical PREM periods (4.8h / 5.4h / 6.0h) is below the 50,000-permutation Welch random-phase surrogate bootstrap p<0.001 threshold across the 24-station IGETS coherent stack","n_permutations":50000},"threshold":{"direction":"negative","metric":"max_period_L_above_p001_threshold","value":0.0},"title":"Slichter mode v4 — methods-clean null with 24-station coherent stack + matched-filter detection","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-11-19"},{"data_source":"reports/phase_d1_a6_v4_l4_admixture_*.json","expected_direction":"negative","feature_set":["fit_epsilon_4_pct","bootstrap_2sigma_low_pct","bootstrap_2sigma_high_pct","cos2_normalized_spread_pct","n_stations"],"git_commit_at_registration":"62c8379b498813143f2c6e418150b8eeded0c996","hypothesis_id":"20260521_wahr_1981_l_4_admixture_v4_forwa_a6ae7dc5","locked_at":"2026-05-21T02:55:05Z","locked_file_sha256":"77f54d3c2391bd921c90bbf1918f6139a01b7bbb352ce75f103929996492d9e1","notes":"FORWARD test of the Wahr 1981 +0.700% l=4 admixture prediction under a tightened v4 calibration pipeline. v3j (commit 80c9617b, OTS-anchored at 15e80117) returned epsilon_4=-1.216% with 2sigma CI [-2.504%, -0.131%], placing Wahr's +0.7% prediction +0.831% above the CI upper edge (~1.4sigma excluded). The v3 pipeline has a documented ~1-2% calibration systematic from Path B published-range OTL midpoints + Boy 2002 K-drift fallback for 11 of 23 stations + empirical-fit Hinderer beta where 7 of 23 stations show >25% v1-v2 divergence. The v4 forward test substitutes 2 systematic-reducing engineering changes: (1) OSO Chalmers Path A FES2014b per-station BLQ tables for ocean tidal loading (replacing Path B midpoints); (2) Hinderer v2 literature-calibrated atmospheric admittance beta per station (replacing v1 empirical fit at the 7 outlier stations). DISPOSITION RULES: PASS if CI width tightens below 2.0% AND CI upper bound remains below Wahr +0.700%; AMBIGUOUS if CI widens or moves toward zero; NULL-CONFIRMED if CI tightens below 2.0% AND now includes Wahr +0.700% (interpretation b of the v3j .md was correct, the v3j NULL was systematic-dominated). Pre-conditions: Mike sends OSO Chalmers BLQ email (Mike-action, send-ready memo at reports/phase_d1_a6_item4_oso_chalmers_send_ready_memo.md), parses tables via scripts/parse_oso_blq_to_otl_dict.py, switches script to use cross_physics/_otl_path_a_blq.OTL_M2_NGAL_PATH_A_PER_STATION + cross_physics/hinderer_atmospheric_loading_correction.apply_hinderer_correction. Re-eval window: 2026-08-15 (90 days from lock, allowing for OSO BLQ ~24h turnaround + parser ship + script update + v4 production run + bootstrap analysis). Companion paper: reports/papers/wahr_1981_l4_admixture/draft_v0_1.md (Wahr 1981 GJI letter v0.1, commit 62c8379b). The v4 forward test result will discriminate interpretation (a) PREM wrong from interpretation (b) v3 systematic in the v3j .md.","ots_proof_path":null,"outcome":{"lookahead_hours":2160.0,"min_magnitude":0.0,"type":"earth_tide_admixture"},"registered_at":"2026-05-21T02:54:53Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"612219f4a05c9ffb62de574bc803058942a30815","interpretation":"disconfirming","metric_value":0.103,"n_observed":23,"notes":"AMBIGUOUS disposition per pre-registered rules: CI WIDENED (v3j 2.373% -> v4 2.579%) AND moved toward zero (v3j upper -0.131% -> v4 upper +0.103%, now includes 0). v4 CORRECTED Path A FES2014b OSO BLQ + Hinderer v2 result: epsilon_4 central -1.195%, 2sigma CI [-2.476%, +0.103%], cos2-normalized spread 60.00% (slight improvement from v3j 62.43%). Wahr 1981 PREM +0.700% prediction STILL EXCLUDED (CI upper +0.103% < +0.7%) but by narrower ~0.9sigma margin (was 1.4sigma in v3j). Pure-elastic epsilon_4=0 is now INSIDE the v4 CI (was outside in v3j). Strict PASS branch not met (PASS required CI width <2.0%; v4 width 2.579% > 2.0%). NULL-CONFIRMED branch not met (would have required CI to include Wahr +0.7%; v4 CI upper +0.103% still excludes it). Scientific interpretation: v4 CONFIRMS the v3j NULL with slightly-better per-station calibration; Wahr 1981 prediction not corroborated under either Path A or Path B OTL; but v4 CI is still systematic-dominated enough that pure-elastic epsilon_4=0 is also viable. Publishable as methods-clean null on 23-station SG-network coherent bootstrap; stronger 'PREM wrong' claim NOT yet supported. v5 path: NLOADF + GLDAS atmospheric loading + ETERNA/BAYTAP-G per-station drift removal to push systematic below 1% would let v5 CI width tighten below 2.0% threshold.","ots_proof_path":null,"passed":false,"tested_at":"2026-05-26T18:36:15Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"basis":"Y_4^2_associated","calibration_pipeline":"v4 (OSO Path A FES2014b BLQ + Hinderer v2 literature beta)","fallback_v3":"if Path A BLQ unavailable, Path B published-range midpoint; if Hinderer v2 alias miss, v1 empirical fit"},"method":"bootstrap_2sigma_ci_disposition","n_permutations":10000},"threshold":{"direction":"negative","metric":"ci_upper_bound_pct","value":0.7},"title":"Wahr 1981 l=4 admixture v4 forward test: OSO Path A BLQ + Hinderer v2 literature beta tightens CI while excluding +0.7%","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-08-15"},{"data_source":"Sentinel-2 L2A via Microsoft Planetary Computer STAC + IRIS FDSN 3-component + IRIS FDSN microseism","expected_direction":"positive","feature_set":["phc_kaolinite_score","phc_vegetation_chlorosis_score","phc_iron_oxide_score","phc_pol_em_deflection_score","phc_vertical_splitting_score","phc_composite_score_0_1"],"git_commit_at_registration":"e49ebcf6ac8a979ec96471a0554f1476562adacc","hypothesis_id":"20260522_phc_v0_3_with_250km_play_scale_r_6a8cd857","locked_at":"2026-05-22T00:07:00Z","locked_file_sha256":"7a73faf3e28f25a8e757ee4a18788e1a5b46f6bd6bb329e23bdadbc7a981a714","notes":"PRE-REGISTERED FORWARD TEST.  Re-run scripts/backtest_passive_hc_composite.py --mode regional_anomaly --cohort-radius 250 on AX41 in 90 days (2026-08-21).  Pass criteria (BOTH required): (1) recall at threshold 0.15 >= 60% on the 3 documented Hingeline producers (Covenant Field, Providence Field, Anschutz Ranch East -- per Moulton & Pinnell 2005 OGJ + Chidsey et al. 2007 UGA Pub 36 + Floyd_Moulton-2.pdf), AND (2) false-positive rate at threshold 0.15 <= 20% on the 5 non-producer controls (Bonneville Salt Flats, GSL center, SLC urban, Sevier Lake basin axis, Park City Cu/Au porphyry).  TODAY'S v0.3 BASELINE (2026-05-21T23:57Z, calibration data committed in data/backtest_ax41_phc_v0_3_regional_anomaly_250km.json): threshold=0.15 gives recall 67% (Providence 0.231 + Anschutz 0.198 fire, Covenant 0.005 suppressed) AND FP 0% (all 5 controls < 0.15).  Methodological note: v0.3 is the FIRST PHC iteration to achieve clean Moulton-style discrimination at a useful threshold without false alarms, vindicating the 50km->250km play-scale cohort-radius expansion as the right fix for the v0.2 finding that the 50km cohort was inside the Hingeline trend itself.  Known v0.3 limitation: Covenant signal lost (0.005) because Covenant's neighborhood ALSO has Moulton-anomaly signatures across multiple counties; v0.4 work would explore play-margin-aware cohort selection.  Parallel issue: mss (spectral_splitting_monitor) channel still returns data_real=0 at all sites due to hvsr_basin_mapper assertion failure on master/AX41 (hvsr fix committed in this PR; mss should start contributing on next AX41 backtest run after PR merges + cron pulls).  References: Moulton & Pinnell 2005 OGJ; Chidsey et al. 2007 UGA Pub 36 'Petroleum geology of Covenant oil field'; Schumacher 1996 AAPG Memoir 66 'Hydrocarbon-Induced Alteration of Soils and Sediments'; Cassidy 2007; Pride 1994; PICKUP_UTAH_OG_MAY21.md (dense-array synthesis).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"oil_gas_surface_signature"},"registered_at":"2026-05-22T00:06:36Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"bf4209b4a5d96ecd6c4a012963e9c780ddbbf1f7","interpretation":"confirming","metric_value":0.667,"n_observed":8,"notes":"PASS, graded 2026-08-22 (due 2026-08-21). Forward 90-day held-out re-run of scripts/backtest_passive_hc_composite.py --mode regional_anomaly --cohort-radius 250 on AX41, executed from a git worktree pinned at fd671f6d -- the SEALED-ERA instrument: the satellite_hyperspectral scale/CEM repair (commit 731015a8) landed 2026-08-22, AFTER the window closed, so grading at HEAD would have graded a different instrument (shadow rule). Criterion 1 recall at threshold 0.15: 2/3 producers fire (Providence 0.334, Anschutz Ranch East 0.156; Covenant 0.002 suppressed) = 66.7% >= 60% PASS. Criterion 2 FP: 0/5 controls fire (Bonneville 0.001, GSL-center 0.063, SLC-urban 0.003, Sevier-axis 0.004, Park City 0.004) = 0% <= 20% PASS. Reproduces the 2026-05-21 baseline (67%/0%) with the same per-site pattern. n_real 4-5 channels per site. dense-array forward seeds (no ground truth): Baker Hot Springs 0.004, Seeker Field 0.009. Artifact: /tmp/wt_phc_seal/data/backtest_passive_hc_composite_regional_anomaly_250km_20260822T183836Z.json (copy to repo data pending). A HEAD-instrument shadow re-run will be reported separately and is NOT part of this grade.","ots_proof_path":null,"passed":true,"tested_at":"2026-08-22T18:43:33Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"10_site_hingeline_backtest_v0_3_250km","n_permutations":null},"threshold":{"direction":"positive","metric":"recall_at_threshold_0_15","value":0.6},"title":"PHC v0.3 with 250km play-scale regional anomaly normalization reproduces threshold=0.15 clean discrimination (recall >= 60% AND FP <= 20%) on Hingeline backtest","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-08-21"},{"data_source":"nasa_omni_1h.db + sse_catalog_expanded.csv","expected_direction":"two_sided","feature_set":["ae_index_1h","al_index_1h","au_index_1h","imf_Bz_nT","solar_wind_speed_km_s"],"git_commit_at_registration":"a19d249f8acbbe45ac3a4ed55504cd063890d8f5","hypothesis_id":"20260522_sse_piezo_electromagnetic_precur_682c9ac6","locked_at":"2026-05-22T17:39:00Z","locked_file_sha256":"8961935d72a6fe3a081ddab3da291131de5a48537e804e0e8642265711e1388d","notes":"RETROSPECTIVE test of Tributsch 1978 / Freund 2002 / Varotsos VAN-method piezoelectric current generation hypothesis at slow-slip faults. Phase 5d (Apr 25 2026) found microseism_primary_band_6s = -7.22 sigma at lag -336h before N=78 documented SSE events (Cascadia ETS + Japan Bungo/Tokai/Boso + NZ Hikurangi, 2005-2018). The piezoelectric hypothesis predicts the same -336h window should ALSO show a magnetic-activity signature if pre-SSE stress in quartz-bearing host rock generates measurable telluric/electrojet currents. This test runs stacked-epoch z-score analysis of NASA OMNI 1-hour global magnetic activity indices (AE/AL/AU electrojet + IMF Bz + solar wind speed for confounds) against the expanded N=79 SSE catalog (data/retro_cache/sse_catalog_expanded.csv), with F10.7 solar-cycle quartile-matched control (per Phase 5 protocol that collapsed +15sigma prompt-gravity headline to +2.6sigma after control). Decision rule: post-control |z| >= 3.0 at any lag in [-336h, 0h] across at least one channel, Bonferroni-corrected over 5 channels x 8 lag bins = 40 tests => threshold p<0.00125. CAVEAT: global magnetic indices are coarse for localized fault piezo currents; a null result is therefore an upper bound on globally-detectable signature, not a falsification of the localized mechanism. Either outcome is publishable: positive corroborates Freund P-hole mechanism (Nature paper); null establishes upper bound on global-magnetic-index detection of SSE piezo coupling (GRL paper).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2026-05-30","registered_at":"2026-05-22T17:38:48Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"a19d249f8acbbe45ac3a4ed55504cd063890d8f5","interpretation":"disconfirming","metric_value":-4.2,"n_observed":null,"notes":"MARGINAL/NULL at the strict pre-registered threshold. Bonferroni FWER 0.001 (|z|>=4.215): 0 of 40 tests survive. Bonferroni FWER 0.05 (|z|>=3.227): 3 of 40 tests survive -- imf_Bz_nT at lag -147h z=-4.20 (peak), au_index_1h at -315h z=+3.90, imf_Bz_nT at -273h z=-3.87. INTERPRETATION: Marginal survivors are dominated by IMF Bz at lags -147h, -273h. IMF Bz southward turnings trigger magnetospheric storms. F10.7 controls 11-year solar cycle but NOT sub-cycle (~days) CME / solar-storm correlation with SSE timing. Most plausible explanation is sub-cycle solar-storm confound NOT removed by F10.7, NOT piezo coupling. HARNESS CAVEAT: microseism cross-check did NOT reproduce Phase 5d (-7.22 sigma at -336h on N=78). Got -3.37 sigma at -105h on N=5 due to retro_microseism archive being ~11 rows/day not hourly. Phase 5d code-path probably used different aggregation. VERDICT: methods-clean upper bound on globally-detectable pre-SSE magnetic-activity signature via OMNI 1-hour indices, NOT a discovery of piezo coupling. PUBLISHABLE as null result. Result JSON at reports/sse_piezo_em_result.json.","ots_proof_path":null,"passed":false,"tested_at":"2026-05-22T17:48:23Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"stacked_epoch_z_with_solar_cycle_control","n_permutations":10000},"threshold":{"direction":"two_sided","metric":"z_score_post_F10p7_control","value":3.0},"title":"SSE piezo-electromagnetic precursor: AE/AL/AU/Dst/IMF-Bz signature in -336h pre-SSE window, N=79 catalog","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"IGETS SG network, 10yr, D.1-SPRINT corrected","expected_direction":"positive","feature_set":["epoch_fold_triplet_5.40h"],"git_commit_at_registration":"768b9231edb4525fe96d17ec70086d4097b93cdd","hypothesis_id":"20260607_slichter_5_40h_triplet_replicate_38bc5632","locked_at":"2026-06-07T02:23:49Z","locked_file_sha256":"7f2e146e5596ea3360a4ddee631591e7c44c52862fb170942b5f53c3da673235","notes":"CONSISTENCY test on existing data (full-data v2 already seen: 5.40h triplet, 0.83% split, p_lookelsewhere=0.04). Criterion locked BEFORE the split: REPLICATES iff BOTH the even/odd-station AND first/second-half-year splits show the 5.40h triplet at p_lookelsewhere<0.10 per half with best-split-frac consistent within +/-50%. PASS(both)=detection-grade Slichter triplet; PARTIAL(one)=promising; FAIL(neither)=v2 marginal was stack-specific coherent noise -> definitive null.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"free_oscillation"},"reevaluate_at":"2026-06-07","registered_at":"2026-06-07T02:23:30Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"768b9231edb4525fe96d17ec70086d4097b93cdd","interpretation":"disconfirming","metric_value":1.0,"n_observed":null,"notes":"PARTIAL: 1 of 2 splits replicated (pre-reg required 2). TIME split (first/second-half years) REPLICATES -- both halves p_lookelsewhere<0.05 at the SAME 0.40% splitting. STATION split (even/odd) FAILS -- both halves show a triplet (p 0.014/0.052) but at INCONSISTENT splittings (0.69% vs 2.12%); a single global rotational mode would give one splitting. Central 5.40h line is robust across all splits; the triplet SIDEBAND structure does not reproduce across station subsets. => NO confirmed Slichter detection. Robust central line + time-only sideband replication motivate the D.0 station expansion as the final word.","ots_proof_path":null,"passed":false,"tested_at":"2026-06-07T02:27:11Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"triplet look-elsewhere p in independent halves (even/odd stations + first/second-half years)","n_permutations":500},"threshold":{"direction":"positive","metric":"n_splits_replicating","value":2.0},"title":"Slichter 5.40h triplet replicates across independent data splits","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"USGS FDSN M>=4.5 catalog 2005-2018; earthquake.predictor REGIONS bboxes","expected_direction":"positive","feature_set":["schuster_R_tidal_sensitivity"],"git_commit_at_registration":"b72820063f794e027fc496058efeb6a6e7b1bb37","hypothesis_id":"20260607_tidal_fault_criticality_predicts_5e1c6d12","locked_at":"2026-06-07T00:35:44Z","locked_file_sha256":"fbe3c0f946f77e7989dd1b230afe71bf5d8c852a111a8e29fd56841e98c799b5","notes":"CONFIRMATORY held-out (data exists, not blind forward). Predictor=per-region Schuster R of M>=4.5 tidal phases 2005-2014 (>=50 events, 46 regions). Outcome=2015-2018 M>=5.5 count minus baseline (n_train_M5.5*0.4). PASS = rho>=0.25 AND p_perm<0.05; else NULL (publishable). Integrity: criterion+method+split locked BEFORE the test correlation was computed; train predictor inspected, test NOT. Cochran 2004 Schuster.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":5.5,"type":"earthquake"},"reevaluate_at":"2026-06-07","registered_at":"2026-06-07T00:35:23Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"b72820063f794e027fc496058efeb6a6e7b1bb37","interpretation":"disconfirming","metric_value":-0.0074,"n_observed":null,"notes":"HELD-OUT NULL. n_regions=46, spearman_rho=-0.0074, p_perm=0.955 (criterion rho>=0.25 AND p<0.05 NOT met). Train tidal-sensitivity map IS real (Japan -log10p=105.9, Bonin 44.9, Kuril 38.8 robust) but does NOT predict 2015-2018 failure at regional scale. Likely: criticality is fault-segment-specific not region-wide; 4yr test window short; M5.5 failures sparse per region. Methods-clean null -- publishable.","ots_proof_path":null,"passed":false,"tested_at":"2026-06-07T00:36:22Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"spearman_rho(train_schuster_R, test_excess_failure) + phase-shuffle permutation null","n_permutations":5000},"threshold":{"direction":"positive","metric":"spearman_rho","value":0.25},"title":"Tidal fault-criticality predicts failure (held-out 2005-14 to 2015-18)","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"retro tensor data/retro_tensor.db (122712 hourly snapshots 2005-2018) via retro_tensor_builder.feature_names(); F10.7 from data/solar_cycle_archive.db; common.deconfound annual+semiannual+F10.7","expected_direction":"negative","feature_set":["microseism_secondary_band_3s","cosmic_ray_thule_counts","qgr_temporal_gravity_change_rate","cosmic_ray_oulu_counts","dst_index_1h","ae_index_daily_max","f10p7_flux"],"git_commit_at_registration":"b2547b9df3b7307c7b80ddc7c1406a0cd4dfd069","hypothesis_id":"20260609_de_confounding_preserves_microse_72295a03","locked_at":"2026-06-09T17:02:53Z","locked_file_sha256":"0e857f92522dec47e9cc0f62b6b404a257991339537feb986a4e6def03f43501","notes":"De-confounding TRUST gate before wiring into classifier-training/mining. De-confounding (common.deconfound: annual+semiannual+real F10.7) must reproduce the INDEPENDENT Apr-25 solar-cycle artifact-gate classification. PASS = ALL THREE: (1) PRESERVE microseism_secondary_band_3s (the +19sigma precursor, gate=REAL) confound_r2 < 0.30; (2) KILL median confound_r2 of {qgr_temporal_gravity_change_rate, cosmic_ray_oulu_counts, dst_index_1h, ae_index_daily_max} (gate=ARTIFACT) > 0.50; (3) RANK artifact-group median > REAL-group median. f10p7_flux reported as positive control (expect ~1.0, basis contains F10.7) but EXCLUDED from the KILL median to avoid triviality. Mechanism: an event-locked -10h EQ precursor is aperiodic so a periodic annual/solar basis cannot capture it -> microseism survives by construction; prompt-gravity was spurious BECAUSE it tracked the solar cycle -> removed. PASS authorizes wiring de-confounding into the retro classifier/mining inputs (gated, opt-in). FAIL (esp. if microseism r2 >= 0.30) means de-confounding is too aggressive and must not be wired. Confirmatory on existing 14-yr static data; locked before first run. Eval: scripts/deconfound_preserve_kill_eval.py","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"methodology"},"reevaluate_at":"2026-06-09","registered_at":"2026-06-09T17:02:42Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"dfe84c9f26cba05d167e8aad8f805beb6b736554","interpretation":"disconfirming","metric_value":0.035,"n_observed":122712,"notes":"FAIL on composite (c2), but DIAGNOSTIC. PRESERVE confirmed strongly: microseism_secondary_band_3s confound_r2=0.0036, cosmic_ray_thule 0.0074 -> feature-level de-confounding is SAFE for the +19sigma precursor (c1 PASS). F10.7 positive control 0.952 -> basis valid. KILL FAILED: the artifact features qgr=0.0016, oulu=0.015, dst=0.035, ae=0.039 are all <4% periodic-confounded (c2 needs median>0.50). KEY REFRAME: the Apr-25 artifact gate found prompt-gravity's STACKED-EPOCH precursor z collapsed under a solar-cycle control because EVENTS cluster in solar-cycle PHASE -- that confound lives in the event SAMPLE, not in the feature's variance spectrum (dst/ae/cosmic_ray vary on storm/day timescales, not the smooth annual+11yr envelope). Feature-level de-confounding removes PERIODIC variance and therefore CANNOT remove a stacked-epoch event-clustering confound. So this test correctly shows feature/eigenmode-level de-confounding is the WRONG tool for stacked-epoch precursors (already handled by the solar-cycle control) and the RIGHT tool for CONTINUOUS cross-feature correlation / eigenmode mining, where shared annual/solar periodicity inflates couplings -- consistent with the tensor diagnostic (PC0 66%, PC1 63%, 21% variance-weighted periodic confounder). RECOMMEND: wire de-confounding into the discovery-engine / cross-correlation / Phase-4 mining inputs (its correct domain); do NOT use it to replace stacked-epoch solar controls. A properly-scoped follow-on pre-reg: de-confounded eigenmode inputs reduce Phase-4 spurious-triple rate on held-out validation. Eval: scripts/deconfound_preserve_kill_eval.py","ots_proof_path":null,"passed":false,"tested_at":"2026-06-09T17:04:48Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"confound_r2_preserve_kill","n_permutations":null},"threshold":{"direction":"negative","metric":"preserve_microseism_r2","value":0.3},"title":"De-confounding preserves microseism precursor + kills solar artifact","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"retro eigenmodes (eigenmode_timeseries bundle 38, 122712 snapshots 2005-2018) raw vs deconfounded (common.deconfound annual+semiannual+F10.7); M7+ targets from USGS FDSN; identical Phase-4 mining code path (tertile, univariate |z|>=2, Bonferroni)","expected_direction":"positive","feature_set":["retro_eigenmodes_bundle38_deconfounded"],"git_commit_at_registration":"0310f72bdd84b29b9d00da23e18eb38e2a8fbaf9","hypothesis_id":"20260609_de_confounding_reduces_phase_4_t_320f3a7a","locked_at":"2026-06-09T17:13:39Z","locked_file_sha256":"2aaa3f680aa6b433a9147309cc132db5c2b5b51dd2bbaeca80569f5f3efe423e","notes":"De-confounding TRUST gate, part 2 -- the CORRECT-domain test. The preserve/kill gate (20260609_de_confounding_preserves_microse_72295a03) showed feature-level de-confounding is the right tool for CONTINUOUS mining (not stacked-epoch). This runs the canonical Phase-4 ternary mining (M7+/168h, tertile states, univariate |z|>=2 pre-filter, Bonferroni) on retro eigenmodes RAW vs DE-CONFOUNDED (each mode residualised on annual+semiannual+real F10.7), SAME M7+ targets + SAME code path -- only the eigenmode input differs. PASS = spurious_triple_reduction=(n_sig_raw-n_sig_dec)/max(1,n_sig_raw) >= 0.50 on the NAIVE full-sample mining (122712 snapshots, the best-powered spurious surface) AND n_sig_raw >= 1. Rationale for naive: de-confounding removes PERIODIC variance (PC0 66%/PC1 63% periodic per the tensor diagnostic), and the naive surface is dominated by periodicity+autocorrelation, so its effect is cleanest+best-powered there; the 168h-block subsample counts are also reported for context but de-confounding removes periodicity, not general autocorrelation. PASS authorizes wiring de-confounding into the discovery-engine/mining inputs (gated, opt-in). FAIL (<50% reduction) means periodicity is not the dominant spurious driver in mining and de-confounding helps less than hoped. Confirmatory on 14-yr static retro data; locked before first run. Eval: scripts/deconfound_mining_contrast.py","ots_proof_path":null,"outcome":{"lookahead_hours":168.0,"min_magnitude":7.0,"type":"earthquake"},"reevaluate_at":"2026-06-09","registered_at":"2026-06-09T17:13:20Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"b74e2191e788ebeb365fedd96eb7431c81e659a4","interpretation":"disconfirming","metric_value":0.033,"n_observed":122712,"notes":"FAIL, and it REFUTES the periodicity hypothesis. Ran on bundle 40 (the materialized retro 14-yr eigenmode timeseries, 122712 snapshots x 50 modes; bundle 38 named in the lock is PCA-only with no materialized timeseries -- load-time substitution, no result seen beforehand). M7+/168h, base rate 0.24, 219 events. Naive full-sample: de-confounding cut Bonferroni-significant triples only 220423->213146 = 3.3% (needed >=50%). Autocorrelation-controlled 168h-block subsample: significant count ROSE 11->79 (Bonferroni threshold = q/n_tests differs between arms, so this is a threshold/small-sample artifact, not a real increase in true signal). CONCLUSION: the Phase-4 mining spurious-triple surface is dominated by TEMPORAL AUTOCORRELATION (hourly non-independent snapshots + overlapping 168h target windows), NOT by shared annual/solar periodicity -- de-confounding removes periodicity (~3% of the surface) and therefore does NOT fix the mining's false-discovery problem. The correct tool for that is the block-subsample + held-out validation the canonical Phase-4 design ALREADY uses (which gave the documented clean null). Combined with the preserve/kill result (stacked-epoch confounds are event-sample solar-clustering, handled by solar controls): in BOTH candidate domains the real confound mechanism turned out to be something de-confounding does not address and the platform already controls. De-confounding is therefore NOT wired into production -- it remains a validated, signal-safe utility (microseism 0.4%) + the honest diagnostic finding that ~21% of tensor variance (PC0 66%/PC1 63%) is periodic. Eval: scripts/deconfound_mining_contrast.py.","ots_proof_path":null,"passed":false,"tested_at":"2026-06-09T17:44:39Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"phase4_mining_raw_vs_deconfounded_significant_triple_count","n_permutations":null},"threshold":{"direction":"positive","metric":"spurious_triple_reduction","value":0.5},"title":"De-confounding reduces Phase-4 ternary-mining spurious triples","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"IRIS SPUD EMTF (8210 MT transfer functions) via cross_physics/emtf_magnetotelluric.apparent_resistivity_at, determinant apparent resistivity @ ~100 s, max 100 km","expected_direction":"negative","feature_set":["emtf_rho_det_ohmm"],"git_commit_at_registration":"d446d95156a4876a875e6cbca1ecbc2345c69897","hypothesis_id":"20260609_emtf_resistivity_discriminates_f_3d4ddf91","locked_at":"2026-06-09T16:31:58Z","locked_file_sha256":"2f111195d039ba4727ed32b9e98cb3c31ff530fee5425f95daccc8b30fe38458","notes":"EMTF tensor-entry GATE 1 of 2 (dataset usability). Real EMTF determinant apparent resistivity at ~100 s should be systematically LOWER at fluid-rich crust (geothermal/volcanic/magmatic: Yellowstone, Long Valley, Geysers, Salton, Coso, Roosevelt, Valles, Newberry, Soda Lakes, Medicine Lake, Imperial, Mono) than at stable cratonic crust (Superior/Great Plains/Ozark/Appalachian/Texas/Iowa/WY craton/Illinois basin/Oklahoma). PASS = Mann-Whitney one-sided (fluid<stable) p<0.05 AND median(stable)/median(fluid)>=2.0 AND n>=8 real-data sites per group. CONTEXT: the originally-pitched 'validate our synthesized MT' framing is infeasible -- telluric_current_monitor returns tel_apparent_resistivity_ohmm=0.0/data_real=0.0 even at Boulder (BOU), so there is no live synthesized MT to validate against. EMTF therefore stands as a standalone real conductivity dataset; this gate tests whether it carries usable fluid/melt signal at all. Module cross_physics/emtf_magnetotelluric.py DELIBERATELY exports no *_FEATURE_NAMES and no build_*_ml_features, so ECM auto-discovery cannot wire it to the tensor. PASS authorizes a follow-on held-out predictive-lift test (vs ablation, CONUS-only coverage caveat) BEFORE any tensor wiring; FAIL stops EMTF here. Confirmatory test on existing static data; locked before first run. Eval: scripts/emtf_discrimination_eval.py","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"volcano"},"reevaluate_at":"2026-06-09","registered_at":"2026-06-09T16:31:43Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"d446d95156a4876a875e6cbca1ecbc2345c69897","interpretation":"disconfirming","metric_value":0.0606,"n_observed":22,"notes":"FAIL (narrow). EMTF determinant apparent resistivity @100s: median fluid-rich 39.7 ohm.m vs stable craton 73.8 ohm.m (n=12 fluid, 10 stable). Direction CORRECT (fluid less resistive) but misses BOTH locked sub-criteria: Mann-Whitney one-sided p=0.0606 (>=0.05) AND median ratio 1.86x (<2.0). EMTF carries the textbook fluid/melt signal but not strongly enough at the locked 100s/determinant summary to clear the bar -> per the pre-registered rule EMTF is NOT promoted to the tensor. Module cross_physics/emtf_magnetotelluric.py stays a standalone ingestion (no *_FEATURE_NAMES). Likely a too-coarse summary (single 100s determinant): a v2 using shorter periods (shallow geothermal conductors) or a broadband min-rho summary would probably sharpen it -- but that is a NEW pre-registration, not a re-judge of this locked test (no post-hoc p-hacking). Eval: scripts/emtf_discrimination_eval.py; data/emtf/emtf_discrimination_eval.json.","ots_proof_path":null,"passed":false,"tested_at":"2026-06-09T16:36:22Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"mannwhitney_u_one_sided_less","n_permutations":null},"threshold":{"direction":"negative","metric":"p_value_one_sided","value":0.05},"title":"EMTF resistivity discriminates fluid-rich vs stable crust","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"IRIS SPUD EMTF (8210 MT transfer functions) via cross_physics/emtf_magnetotelluric.band_resistivity_at; MIN determinant apparent resistivity across full period band, max 100 km","expected_direction":"negative","feature_set":["emtf_rho_min_ohmm"],"git_commit_at_registration":"df85c5751e86f7c7d6d857b877755a3c5e0a0871","hypothesis_id":"20260609_emtf_v2_broadband_min_resistivit_0024a9ef","locked_at":"2026-06-09T16:42:05Z","locked_file_sha256":"fa4a1857b9583636539530b3ce37e1a356540c041031ab2354a1cb0369e04e1a","notes":"EMTF tensor-entry GATE v2 (dataset usability). Supersedes v1 (20260609_emtf_resistivity_discriminates_f_3d4ddf91), which FAILED narrowly (p=0.061, ratio 1.86x) using a single 100s determinant. v2 changes ONLY the per-site summary on PHYSICAL grounds: MINIMUM determinant apparent resistivity across the full period band (the conductivity low = depth of max saline-fluid/partial-melt; stable craton stays resistive at all depths so its min-rho stays high). SAME 12 fluid-rich + 12 stable site lists as v1 (imported, not re-typed) and SAME statistical bar -- only the summary metric changed, so this is not threshold-gaming. min-rho is a quantity NOT inspected in v1 (v1 reported only medians/p/ratio at 100s), so this lock is blind to the v2 outcome. PASS = Mann-Whitney one-sided (fluid<stable) p<0.05 AND median(stable min-rho)/median(fluid min-rho)>=2.0 AND n>=8 real-data sites per group. Module cross_physics/emtf_magnetotelluric.py exports no *_FEATURE_NAMES/build_*_ml_features -> auto-discovery cannot wire EMTF to the tensor. PASS authorizes a follow-on held-out predictive-lift test BEFORE tensor wiring; FAIL stops EMTF. Confirmatory test on existing static data; locked before first run. Eval: scripts/emtf_discrimination_eval_v2.py","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"volcano"},"reevaluate_at":"2026-06-09","registered_at":"2026-06-09T16:41:53Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"df85c5751e86f7c7d6d857b877755a3c5e0a0871","interpretation":"disconfirming","metric_value":0.2655,"n_observed":22,"notes":"FAIL (clearer than v1). Broadband MIN determinant resistivity: median fluid 8.5 ohm.m vs stable 16.1 ohm.m (n=12/10), ratio 1.88x, Mann-Whitney one-sided p=0.2655 -- WORSE than v1's p=0.061. Physically-motivated change degraded separation because the conductivity-low is NOT geothermal-specific: stable interiors also have conductive layers (basin sediments, upper-crustal fluids), so min-rho across the band muddies the fluid-vs-stable ranking. Both pre-registered usability gates (v1 mid-crustal 100s, v2 broadband min) now FAIL -> EMTF does NOT enter the tensor. Honest conclusion: a CONUS-wide single-number-from-nearest-MT-site summary grouped by geology label lacks the power/specificity to make EMTF a useful blanket resistivity feature. Any further EMTF use should be target-specific (depth-resolved 1D inversion at a named site, or dense-array dense-array co-location) -- NOT a tensor feature. Recommend STOP. Eval: scripts/emtf_discrimination_eval_v2.py.","ots_proof_path":null,"passed":false,"tested_at":"2026-06-09T16:46:30Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"mannwhitney_u_one_sided_less","n_permutations":null},"threshold":{"direction":"negative","metric":"p_value_one_sided","value":0.05},"title":"EMTF v2: broadband min-resistivity discriminates fluid-rich vs stable crust","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"microseism_precursor_tensor_monitor z per region + USGS/EMSC M>=5.5 events + regional 36h base rates","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"f5055b211d75cfbac31ae13521f627de10a3fe6c","hypothesis_id":"20260701_microseism_10h_precursor_as_a_sh_47d7b6f5","locked_at":"2026-07-01T15:25:34Z","locked_file_sha256":"5644241fadb2e4a89f6d434125ddb07def004c9f69098f5048bf7c95c37973af","notes":"MICROSEISM -10h -> 24h SHADOW NOTICE. Operationalize the Phase-5 universal-precursor finding (microseism secondary-band 3s +19sigma at lag -10h for M5.5+, survives solar-cycle control) into a per-event 24-36h EQ notice, run in SHADOW (no user-facing alerts) until calibrated. A shadow predictor fires a 24h notice when the per-region microseism precursor z crosses a locked threshold; each notice is scored forward for whether an M>=5.5 follows within 36h (precision) and whether that beats the regional 36h base rate (lift). PASS = precision>=0.15 AND lift>=2x over base rate (bootstrap p<0.05) -> a real short-lead notice for the SUBSET of events the precursor precedes. NULL = no lift -> the stacked-epoch -10h signal does NOT transfer into a reliable per-event alarm (honest, publishable). This is the 24h rung of the graded 30d->7d->36h->24h cascade: the long rungs are always-on regional windows; this short rung fires only when the precursor activates (LOW coverage, HIGH value). Shadow-only; promote to live alerts ONLY after calibration + a passing forward result. Withdraw if <10 notices over the window (underpowered -> re-register looser). Brand: DeepMap AI Inc.","ots_proof_path":null,"outcome":{"lookahead_hours":36.0,"min_magnitude":5.5,"type":"earthquake"},"reevaluate_at":"2027-01-01","registered_at":"2026-07-01T15:25:33Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"673dff0ac5e4841da96c6dd84964911ac8411165","interpretation":"disconfirming","metric_value":0.08,"n_observed":82,"notes":"DISCONFIRMED by independent verification wave (trackAR, 2026-08-24). The feature microseism_secondary_band_3s is MUSTANG pct_above_nhnm -- a BROADBAND whole-day metric that any earthquake's own energy inflates. The z is earthquake self-contamination: the event-day time profile spikes at k=0 with SYMMETRIC foreshock/aftershock shoulders (not a pre>>post precursor), and on a daily feature '-10h' lands on the same UTC day as the event 58% of the time. On the un-biased band_rms (0.14-0.5 Hz, hourly) metric with a clean baseline + declustering, -10h z = +0.08/-0.23/-0.08 across magnitude bins with NO scaling -- decisively excluded (a real precursor at n=82 M7.5+ would give t~13); positive control event-hour z=+30.3. Reproduces the sealed positive control exactly (19.02 sigma), so the effect is real AS COMPUTED but is the earthquake, not an ocean-microseism precursor. Reports: ~/trackAR/trackAR_report.md, ~/trackAN/.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-24T20:27:57Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"baseline":"regional 36h base rate of M>=5.5","basis":"Phase-5 microseism secondary-band 3s +19sigma at lag -10h for M5.5+ (solar-cycle-controlled)","cascade_rung":"24-36h (short rung of 30d/7d/36h/24h; fires only on precursor activation)","event_set":"M>=5.5 within 36h of a shadow notice, bounds-first","forward_window":["2026-07-01","2027-01-01"],"mode":"SHADOW: logs to advisories/shadow table NOT the scored ledger; no user-facing alerts until calibrated","notice_window_hours":36,"null_outcome":"no lift -> -10h stacked-epoch signal does not transfer to a per-event notice","pass":"precision>=0.15 AND hit-rate-lift over regional 36h base rate >=2x, bootstrap p<0.05","trigger":"per-region microseism precursor z (microseism_secondary_band_3s_z) crossing a locked threshold","withdraw":"<10 shadow notices -> underpowered"},"method":"shadow_microseism_precursor_24h_notice_skill","n_permutations":null},"threshold":{"direction":"positive","metric":"notice_precision_and_lift_over_baserate","value":0.15},"title":"Microseism -10h precursor as a SHADOW 24-36h EQ notice: skill over base rate (forward held-out)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"EarthScope IRIS SPUD EMTF impedance tensors (cached)","expected_direction":"positive","feature_set":["mpt_meanphase_deg_crust"],"git_commit_at_registration":"fa7cb0fead63e2330f8a035debad1b2f0a0c86c3","hypothesis_id":"20260705_mt_phase_tensor_crustal_mean_pha_b1dd0fa6","locked_at":"2026-07-05T15:19:29Z","locked_file_sha256":"1a3a745c6f6f49522beb67e1d9cad62bbb629871c1cfbb6804f4ecf6680a93aa","notes":"CLAIM: MT phase-tensor crustal MEAN PHASE (Caldwell 2004) is HIGHER at fluid-fault sites than basin/craton controls (mean phase >45deg = conductor below = fluid at depth). Follow-up to the skew NULL 20260705_mt_phase_tensor_crustal_skew_dis_c577e798 whose ONE suggestive secondary was mean-phase (fault 52 vs control 38, p=0.078, NOT the registered primary). Tested on FRESH sites DISJOINT from the skew 16. FRESH FAULT(11): SaltonSea_Brawley Coalinga_SAF Hollister_SAF OwensValley_fault SurpriseValley_CA BlackRock_NV_geo RaftRiver_ID_geo Yellowstone Socorro_RioGrande VallesCaldera_NM Hayward_fault_CA. FRESH CONTROL(8): DenverBasin_CO ColoPlateau_interior SacramentoValley_CA GreatPlains_KS NevadaBasin_Diamond ColumbiaBasin_E WyomingRedDesert KansasCraton. PASS=one-sided MWU p<0.05 (fault>control). Supporting=fault median>=45 AND control<45. WITHDRAW if <5 either side. CONFIRMATORY on pre-existing EMTF data; run immediately after lock+anchor. Script scripts/mt_phase_tensor_meanphase_test.py.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geophysical_resistivity"},"reevaluate_at":"2027-07-05","registered_at":"2026-07-05T15:19:06Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"286f97bf90860be1e2202080f3d8fcba713b92b2","interpretation":"confirming","metric_value":0.00926,"n_observed":17,"notes":"","ots_proof_path":null,"passed":true,"tested_at":"2026-07-05T15:24:07Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"mann_whitney_u_onesided","n_permutations":null},"threshold":{"direction":"positive","metric":"mwu_meanphase_p_onesided","value":0.05},"title":"MT phase-tensor crustal mean-phase discriminates fluid-associated fault sites from basin/craton controls","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"EarthScope IRIS SPUD EMTF impedance tensors (cached)","expected_direction":"positive","feature_set":["mpt_skew_deg_crust","mpt_meanphase_deg_crust","mpt_ellipticity_crust"],"git_commit_at_registration":"fe29be8b14ead2bad8bf320e26b28224d80578aa","hypothesis_id":"20260705_mt_phase_tensor_crustal_skew_dis_c577e798","locked_at":"2026-07-05T04:06:27Z","locked_file_sha256":"ef5f72058a86813d3013933db06d8aa6861bb7e45d22c001e3c9ec194b8eedc8","notes":"CLAIM: MT phase-tensor crustal-band skew |beta| (Caldwell/Bibby/Brown 2004 GJI 158:457) is HIGHER at fluid/geothermal-associated fault sites than at basin/craton controls -- tests whether phase-tensor GEOMETRY discriminates where the SCALAR EMTF apparent-resistivity FAILED (20260609_emtf_*, p=0.061/0.266). FAULT(8, a-priori geology + EMTF-covered): Wasatch_SLC, DixieValley_NV, SodaLakes_NV, SanAndreas_Parkfield, LongValley_ring, SteamboatSprings_NV, Beowawe_NV, Brady_DesertPeak_NV. CONTROL(9): SnakeRiverPlain_ID, ColumbiaPlateau_WA, CentralValley_CA, WyomingBasin, MojaveInterior_CA, GreatBasin_center_NV, BighornBasin_WY, PowderRiverBasin_WY, SanLuisValley_CO. PASS=one-sided Mann-Whitney p<0.05 (fault skew>control). Supporting=fault median|beta|>=3.0deg AND control<3.0deg (Caldwell2004/Booker2014 3D threshold). WITHDRAW if <5 either side with data_real. Hansel Valley + Delta UT EXCLUDED (glimpsed in smoke-test), reported as named application. CONFIRMATORY on pre-existing EMTF data (not blind-future); run immediately after lock+anchor. Script scripts/mt_phase_tensor_fault_discrimination.py; module cross_physics/mt_phase_tensor.py.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geophysical_resistivity"},"reevaluate_at":"2027-07-04","registered_at":"2026-07-05T04:06:12Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"1d911dfcfe431108792b7730cb87b9ec1de09d3d","interpretation":"disconfirming","metric_value":0.936,"n_observed":16,"notes":"","ots_proof_path":null,"passed":false,"tested_at":"2026-07-05T04:10:39Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"mann_whitney_u_onesided","n_permutations":null},"threshold":{"direction":"positive","metric":"mwu_skew_p_onesided","value":0.05},"title":"MT phase-tensor crustal skew discriminates fluid-associated fault sites from basin/craton controls","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"NCEDC UL network creep channels (RB1, mm) -- 12 Parkfield creepmeters verified live 2026-07-08, open-ended recording; FDSN dataselect","expected_direction":"positive","feature_set":["ngl_gnss_station_pairs","insar_deformation","fault_trace_geometry"],"git_commit_at_registration":"7f0bdc5947dcd2af31042f962abd58e4810fe344","hypothesis_id":"20260708_vq31_virtual_creepmeter_v0_1_gns_ae525a54","locked_at":"2026-07-08T22:04:24Z","locked_file_sha256":"54dca08b73c78f8893b7419938e256c4108f03305d0576a9d8ff2f0ae36048f4","notes":"CONFIRMATORY v0.1, registered before running. Cross-fault GNSS pair differencing + InSAR -> creep rate per creepmeter segment; truth = UL RB1 creep series via NCEDC FDSN. PASS = sign agreement >=90% AND rate within factor 2 at >=60% of creepmeters with >=1yr overlapping data AND beats uniform regional-block-rate baseline. v0.2 (repeating-earthquake catalog) is a SEPARATE later registration. Strategic: fault-loading instrument for the parametric channel. Re-eval 2026-08-15.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"fault_creep"},"reevaluate_at":"2026-08-15","registered_at":"2026-07-08T22:04:09Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"c0a8127ccf688b956d2851d49447d0010916a004","interpretation":"confirming","metric_value":0.64,"n_observed":11,"notes":"v0.1 PASS -- all three locked criteria met: sign agreement 0.91 (>=0.90), within-factor-2 0.64 (>=0.60), median |err| 7.12 mm/yr beats regional-block baseline 16.30. Best pairs: PKMID obs +19.48 vs pred +19.84; PKWKR +14.02 vs +14.20; PKXTA +13.86 vs +14.99 (GNSS cross-fault differential, truth untouched). Failures physically coherent: PKMIR/PKVAR/PKX46/PKXHS read near-zero (locked/transition sections) where GNSS pairs still see elastic loading -- the known v0.1 limitation the InSAR leg refines. HISTORICAL window 2006-2012 (UL archive ends 2012-13); GNSS-only (conservative subset of locked features). Follow-ons: post-2013 live creep truth + InSAR leg + productize as the fault-loading instrument for the parametric channel.","ots_proof_path":null,"passed":true,"tested_at":"2026-07-08T23:00:04Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"baseline":"regional_block_motion_rate","mode":"per_creepmeter_rate_comparison","pass_criteria":{"min_frac_within_factor2":0.6,"min_n_creepmeters":6,"min_sign_agreement":0.9}},"method":"vq_validation_harness.validate_spatial","n_permutations":null},"threshold":{"direction":"positive","metric":"frac_within_factor2","value":0.6},"title":"VQ31 virtual creepmeter v0.1 GNSS+InSAR vs Parkfield UL creepmeters","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-08-15"},{"data_source":"NDBC historical stdmet (held-out buoys, significant wave height Hs) x 3-tier FDSN archive + live PPSD coastal stations","expected_direction":"positive","feature_set":["microseism_secondary_band","microseism_primary_band","coastal_station_psd_3tier"],"git_commit_at_registration":"7f0bdc5947dcd2af31042f962abd58e4810fe344","hypothesis_id":"20260708_vq33_virtual_wave_buoy_from_coas_031574ae","locked_at":"2026-07-08T22:04:24Z","locked_file_sha256":"567e9f70b4233f6ca865ff4afe17f63bae7d232ec640eaacef3b99cb03c37606","notes":"CONFIRMATORY, registered before running. Transfer function per coastal seismic station -> Hs at held-out NDBC buoys. PASS = median per-buoy r >= 0.60 AND >=50% of buoys beat 24h persistence (bootstrap ci_lo>0 each). Physics: secondary microseism amplitude ~ ocean wave energy (established; our storm tracker uses it). WITHDRAW if <5 buoy-station pairs qualify. Re-eval 2026-08-15.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"sea_state"},"reevaluate_at":"2026-08-15","registered_at":"2026-07-08T22:03:45Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"bb3eeebe3fdb3fada92b2b5bbf1a0946800c3672","interpretation":"disconfirming","metric_value":0.233,"n_observed":10,"notes":"v1 NULL as locked: median r=0.233, 0/10 beat persistence-24h. DIAGNOSIS: (1) nearest-buoy rule paired land stations with SHELTERED buoys (46088 Strait of Juan de Fuca etc.) whose local Hs is not the open-ocean wave field that generates secondary microseism -- while the 4 Axial SEAFLOOR stations vs open-ocean buoy 46005 tracked at r=0.85-0.90 leave-site-out (physics clearly present); (2) self-persistence baseline presupposes the buoy exists -- the operational counterfactual for a VIRTUAL buoy is the nearest SURVIVING buoy (spatial), not the target buoy's own history. v2 addendum registered with open-ocean pairing rule + nearest-other-buoy baseline; v1 preserved.","ots_proof_path":null,"passed":false,"tested_at":"2026-07-08T22:31:32Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"baseline":"persistence_24h","mode":"leave_site_out","pass_criteria":{"min_frac_beating_baseline":0.5,"min_median_r":0.6,"min_n_sites":5}},"method":"vq_validation_harness.validate_series","n_permutations":null},"threshold":{"direction":"positive","metric":"median_per_buoy_r","value":0.6},"title":"VQ33 virtual wave buoy from coastal microseism vs held-out NDBC","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-08-15"},{"data_source":"NDBC realtime2 WVHT; buoy inclusion rule: climatological median Hs >= 1.0 m over the window (excludes sheltered waters); pairing: nearest qualifying buoy within 500 km","expected_direction":"positive","feature_set":["secondary_microseism_band_power_16hz","3tier_archive"],"git_commit_at_registration":"bb3eeebe3fdb3fada92b2b5bbf1a0946800c3672","hypothesis_id":"20260708_vq33v2_virtual_wave_buoy_open_oc_494e9af7","locked_at":"2026-07-08T22:31:32Z","locked_file_sha256":"5b163fa96d7258f607276c7ac19432325789dc4fb44ff5f300d28b7f80bd1b85","notes":"ADDENDUM to v1 (recorded NULL). Two DESIGN corrections, no model change: (1) buoy inclusion = climatological median Hs >= 1.0 m (open-ocean; the Longuet-Higgins coupling is to the open-ocean wave field -- v1 sheltered-buoy pairs broke the physics assumption); (2) baseline = nearest OTHER qualifying buoy Hs series (the operational dead-buoy counterfactual), replacing self-persistence which presupposes the target instrument exists. Same transfer function, same criteria. One run, result stands. Re-eval 2026-08-15.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"sea_state"},"reevaluate_at":"2026-08-15","registered_at":"2026-07-08T22:31:32Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"e04d62e9107f973819a17f613fa7e36fe6e1941d","interpretation":"disconfirming","metric_value":0.856,"n_observed":6,"notes":"v2 PARTIAL: r criterion PASSED decisively (median r=0.856 vs 0.60 bar; open-ocean rule fixed v1 pairing, 0.233->0.856; Axial seafloor pairs r=0.85-0.90 leave-site-out) but baseline criterion FAILED 0/6 -- the nearest-surviving-buoy counterfactual is brutal: Pacific swell coherence means a real buoy even 401 km away (46002->46005) beats the seismic amplitude estimate. HONEST SCOPE for the instrument: strong sea-state TRACKER; does NOT beat a real buoy within ~400 km on absolute Hs; niche = coasts/basins with NO buoy (untestable by construction -- no truth there). No v3. VQ#33 status: PARTIAL (tracker-grade, not amplitude-grade).","ots_proof_path":null,"passed":false,"tested_at":"2026-07-08T22:34:29Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"baseline":"nearest_other_buoy_hs","mode":"leave_site_out","pass_criteria":{"min_frac_beating_baseline":0.5,"min_median_r":0.6,"min_n_sites":5}},"method":"vq_validation_harness.validate_series","n_permutations":null},"threshold":{"direction":"positive","metric":"median_per_buoy_r","value":0.6},"title":"VQ33v2 virtual wave buoy open-ocean pairing vs nearest-surviving-buoy baseline","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-08-15"},{"data_source":"data/heat_flow/usgs_heat_flow_wells.json (n=7279, gradient_c_km) + DOGM ut_wells_labeled bht_f (n=134, UT independent check)","expected_direction":"positive","feature_set":["heat_flow_covariates","basement_depth","mt_resistivity","gravity","seismic"],"git_commit_at_registration":"7f0bdc5947dcd2af31042f962abd58e4810fe344","hypothesis_id":"20260708_vq34_virtual_geothermometer_beat_b380f98d","locked_at":"2026-07-08T22:04:23Z","locked_file_sha256":"da34bf77a4613ec8e55bcaa705b2c3e31ad34d67bd6c8179bf0a487c34b42e4c","notes":"CONFIRMATORY on historical truth, registered BEFORE validation runs. PASS = held-out spatial-block R2>=0.30 AND >=10% MAE reduction vs IDW-of-truth with bootstrap ci_lo>0. Supervised retry of the heat-convergence stack that FAILED its positive controls Jun 2026 (unsupervised heuristic); this is the fix that memory file named. NULL = covariates add nothing beyond interpolation -> documented, heat thread closed. Run-by / re-eval 2026-08-15.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geothermal_gradient"},"reevaluate_at":"2026-08-15","registered_at":"2026-07-08T22:03:19Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"a231190ec014994e055cac0dad48843f24f2f89b","interpretation":"disconfirming","metric_value":-0.221,"n_observed":7279,"notes":"v1 NULL on raw file as locked: R2=-0.449, MAE 113.3 vs IDW 92.8, skill -0.221 ci[-0.270,-0.175]. DIAGNOSIS (post-hoc, documented): truth file contains physically-impossible gradients (max 9,205 C/km, p99=1,200); 14% of wells (>120 C/km) own 77% of total absolute deviation -- the garbage-well class documented 2026-06-17 in Phase 65x-HEAT-RESISTIVITY (Pine Valley 3,200 C/km -> cap-120 robust rule). Raw-file MAE test is outlier-dominated. v2 addendum registered with the PRIOR-documented cap-120 QC rule; v1 preserved per Time-A precedent.","ots_proof_path":null,"passed":false,"tested_at":"2026-07-08T22:17:54Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"baseline":"idw_of_truth","block_deg":1.0,"k_folds":5,"pass_criteria":{"min_n_sites":1000,"min_r2":0.3,"min_skill_vs_baseline":0.1}},"method":"vq_validation_harness.validate_spatial","n_permutations":null},"threshold":{"direction":"positive","metric":"skill_vs_idw_mae","value":0.1},"title":"VQ34 virtual geothermometer beats IDW on held-out heat-flow wells","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-08-15"},{"data_source":"data/heat_flow/usgs_heat_flow_wells.json filtered gradient_c_km <= 120 (the cap-120 robust QC rule documented 2026-06-17 Phase 65x-HEAT-RESISTIVITY, PRIOR to the 2026-07-08 v1 lock; excludes physically-impossible garbage-well gradients to 9,205 C/km)","expected_direction":"positive","feature_set":["log10_cmt_rho_5depths","qfault_dist","qfault_count_50km","qfault_age_rank"],"git_commit_at_registration":"a231190ec014994e055cac0dad48843f24f2f89b","hypothesis_id":"20260708_vq34v2_geothermometer_beats_idw__81af9a0b","locked_at":"2026-07-08T22:18:14Z","locked_file_sha256":"4b06c220d4d463c1589e3b1a91832e70f26b3c135227721913ba2d039e0a9931","notes":"ADDENDUM to v1 (20260708_...b380f98d, recorded NULL -- outlier-dominated: 14% of wells >120 C/km owned 77% of deviation). v2 applies ONLY the prior-documented cap-120 QC rule; identical model, covariates, folds, and pass criteria otherwise. One run, result recorded whatever it is. The >120 tail is a separate DETECTION question, not re-tested here. Re-eval 2026-08-15.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geothermal_gradient"},"reevaluate_at":"2026-08-15","registered_at":"2026-07-08T22:18:13Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"64ce29dfda907fe860f2bb7977ae684b4b979baa","interpretation":"disconfirming","metric_value":-0.031,"n_observed":6252,"notes":"v2 CLEAN NULL: cap-120 QC applied (6,252 wells), metrics now physical (MAE 19.27 vs IDW 18.70 C/km), R2=-0.082, skill -0.031 ci[-0.043,-0.017]. INTERPRETATION: CONUS-MT resistivity (smooth crustal model) + fault geometry carry NO gradient information beyond spatial interpolation of the 6,252 wells at 1-deg block scale -- the wells ARE the instrument. Confirms the Jun-2026 heat-convergence failure was not a fusion-method problem: the covariates genuinely lack local heat signal. Heat-from-these-channels thread CLOSED with two anchored results (v1 outlier-dominated null + v2 clean null). No v3 covariate-shopping.","ots_proof_path":null,"passed":false,"tested_at":"2026-07-08T22:19:54Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"baseline":"idw_of_truth","block_deg":1.0,"k_folds":5,"max_gradient_c_km":120,"pass_criteria":{"min_n_sites":1000,"min_r2":0.3,"min_skill_vs_baseline":0.1}},"method":"vq_validation_harness.validate_spatial","n_permutations":null},"threshold":{"direction":"positive","metric":"skill_vs_idw_mae","value":0.1},"title":"VQ34v2 geothermometer beats IDW on QC-capped heat-flow wells","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-08-15"},{"data_source":"data/apollo_moonquakes.db -- Nakamura UTIG TR-018 (2008) catalog, 1,359 classified DMQ events 1969-1977; anomalistic phase per lunar_tidal_phase() (same function at train and test)","expected_direction":"negative","feature_set":["lunar_anomalistic_phase_27p55d","phase_bin_rate_model_trained_1969_1974"],"git_commit_at_registration":"1a0acd61643de8a83f52d58ef3dbbd8b27075e64","hypothesis_id":"20260709_vq41_lunar_tidal_quake_gauge_pre_ee6ba102","locked_at":"2026-07-09T01:01:36Z","locked_file_sha256":"93a3061690ae1af23bfe3a4f417ec4708df8a703d32fcf511797e92501650ec5","notes":"FIRST OFF-WORLD VIRTUAL INSTRUMENT validation. CONFIRMATORY temporal holdout, registered BEFORE the held-out years are scored: fit 12-bin anomalistic-phase rate model on TRAIN DMQs (1969-07..1974-12), score HELD-OUT DMQs (1975-01..1977-09). PASS = held-out Schuster/Rayleigh p < 0.01 AND mean per-event log-likelihood gain vs uniform > 0 with bootstrap ci_lo > 0. WITHDRAW if held-out n < 100. Physics: deep moonquakes are tidally triggered (Lammlein 1974; Bulow 2007; Knapmeyer-Endrun 2014) -- this tests PREDICTIVE phase-locking, instrument-grade, distinct from the unregistered Phase-65s retrospective correlation check. Detection-efficiency drift over the mission affects rate, not phase distribution (stated confound, benign for phase tests). If PASS: live gauge (ephemeris runs forever) = today's moonquake-likelihood state, standing ready for Artemis / Farside Seismic Suite truth. Re-eval 2026-08-15.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"deep_moonquake_timing"},"reevaluate_at":"2026-08-15","registered_at":"2026-07-09T01:01:36Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"e5ca0eeba7799f2c114cfd0e1c637a91ee93f753","interpretation":"confirming","metric_value":0.0651,"n_observed":477,"notes":"PASS -- first off-world virtual instrument validated. Train 882 DMQs (1969-07..1974-12), held-out 477 (1975-01..1977-09): Schuster p=4.17e-08 (bar 0.01), mean LL gain +0.0651 ci[+0.0387,+0.0914] (bar ci_lo>0), 0 phase cross-check mismatches. Train peak phase 0.54 (near apogee; per-cluster phases vary -- aggregate consistent with Bulow 2007 family). Deep moonquakes are PREDICTABLY phase-locked to the 27.5545-d anomalistic month in held-out years. Next: live gauge (ephemeris-driven moonquake-likelihood state, runs forever), research card, standing ready for Artemis / Farside Seismic Suite truth.","ots_proof_path":null,"passed":true,"tested_at":"2026-07-09T01:04:10Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"baseline":"uniform_rate","bins":12,"pass_criteria":{"heldout_schuster_p_max":0.01,"min_heldout_n":100,"min_mean_ll_gain_ci_lo":0.0},"test":"1975-01-01..1977-09-30","train":"1969-07-01..1974-12-31"},"method":"temporal_holdout_phase_prediction","n_permutations":null},"threshold":{"direction":"negative","metric":"heldout_schuster_p","value":0.01},"title":"VQ41 lunar tidal quake gauge predicts held-out Apollo deep moonquakes","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2026-08-15"},{"data_source":"Continuous tri-axial waveforms + hypocenter catalog (already in DeepMap via FDSN), a 1D/3D velocity model for take-off angles and P/S ratio k (already used for coda-Q/dv/v), regional stress orientation from InSAR/moment-tensor climatology (already computed), and a reference set of published multi-station focal mechanisms for scoring.  (already collected: YES)","expected_direction":"","feature_set":[],"git_commit_at_registration":"27c31a902e32b849a210a59c52e1bff84b9dcdb7","hypothesis_id":"20260712_springboard_test_us6985816b2_met_474a72f6","locked_at":"2026-07-12T22:59:30Z","locked_file_sha256":"e56756cf5da47baa62b4ed41e04a363b74d30fe9dc2295b90fbae4af1ffd3701","notes":"SPRINGBOARD LINEAGE (invention loop). Brief-derived hypothesis approved by Mike. CLAIM: For shallow crustal microearthquakes (M 1.5-3.5) recorded on at least one clean tri-axial DeepMap station within 30 km, the patent's time-domain Sv/P–Sh/P–Sv/Sh single-station ratio inversion (with a DeepMap regional-stress prior replacing the frac-azimuth assumption) recovers the failure-plane strike within 30° of the independently derived multi-station/gCMT solution for at least 60% of events with amplitude-ratio residual ≤ 0.3. TEST: Assemble a back-test set of a few hundred M1.5-3.5 events in one well-instrumented region (e.g. California/Japan) that ALREADY have published multi-station focal mechanisms. For each, pick the single closest clean tri-axial station, run the ported extraction+inversion, apply the residual≤0.3 and Sv/Sh-sign constraints, resolve order ambiguity with the regional-stress prior. Compare recovered failure-plane strike/dip and failure mode (strike-slip vs normal) to the reference solution. Score % within 30° strike and mode-agreement rate; pass/fail against the 60% criterion. Ablate the stress prior to quantify how much it beats the fixed 44° rule. PASS: ≥60% of residual-qualified events have single-station strike within 30° of the reference multi-station mechanism AND failure-mode agreement ≥65%; the regional-stress prior must beat the fixed-44° rule by ≥10 percentage points on strike accuracy. DATA: Continuous tri-axial waveforms + hypocenter catalog (already in DeepMap via FDSN), a 1D/3D velocity model for take-off angles and P/S ratio k (already used for coda-Q/dv/v), regional stress orientation from InSAR/moment-tensor climatology (already computed), and a reference set of published multi-station focal mechanisms for scoring.  (already collected: YES). Test-before-file: provisional drafted only on pass.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-12","registered_at":"2026-07-12T22:59:30Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"ddeb060ffccb0e6f1dc198f85d53f3ea7d8f6cdb","interpretation":"confirming","metric_value":0.603,"n_observed":null,"notes":"PASS (marginal, exactly at bar): single-station amplitude-ratio focal-mechanism inversion (US6985816B2) recovers failure-plane strike within 30deg for 60.3% of 136 M1.5-3.5 SCSN events (residual<=0.3), median error 25.4deg. Honest caveat: the DeepMap regional-stress prior (0.601) did NOT beat a fixed regional prior (0.601) at n=138 -- the smoke-scale advantage washed out; the pass is on single-station strike recovery, not on the stress-prior refinement. First sparse-truth reproducibility PASS this session -- the RIGHT bar (focal mechanism is per-event, no smooth global field to interpolate, unlike Moho/geothermometer).","ots_proof_path":null,"passed":true,"tested_at":"2026-07-13T02:47:02Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"springboard_hypothesis_test","n_permutations":null},"threshold":{"direction":"","metric":"","value":null},"title":"Springboard test: US6985816B2 - Methods and systems for determining the orientation of natural fractures","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"GOES-West ABI (NOAA Open Data/AWS) + GRACE-FO GWSA + platform dv/v & tidal-compliance + Open-Meteo airmass covariates + 3DEP terrain","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"c121f12e5de264bd25b3b88aa50d8e13c544a9c7","hypothesis_id":"20260713_deep_water_fault_venting_at_pine_63364a5e","locked_at":"2026-07-13T17:12:53Z","locked_file_sha256":"e806485f5294a145bb5678f0568b43573aa222af4183b75cf9a8753d25102e87","notes":"Primary pre-registered test on Pine Valley ONLY (one clean hypothesis). Killer discriminant = H3 tidal: the atmosphere has no lunar M2, a tidally-pumped fault vent does; M2(12.42h) must be separated from solar S2(12.00h) via the multi-year record. Orographic null controlled with 3 matched Colorado Plateau laccoliths (same intrusive-mountain physics). Verdict: CONFIRMED if H3 + (H1 or H2); SUPPORTIVE if H1+H2 only; NULL if H1 fails. Confirmatory-on-archive (see framing in test.extra). Atlas-wide scan is a SEPARATE follow-on pre-registration. Motivated by the deep-water-through-faults literature (Tyne 2025 Palouse, Sano 2025 Noto pre-M7.6 He shift, Warr 2022, Hallis 2015) + Pine Valley artesian pressurizing through 20yr drought.","ots_proof_path":"hypotheses/locked/20260713_deep_water_fault_venting_at_pine_63364a5e.json.ots","outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"deep_water_venting"},"reevaluate_at":"2026-08-15","registered_at":"2026-07-13T17:12:36Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"e776066ebe0ab569f1431f9da7129a55b9a973be","interpretation":"disconfirming","metric_value":0.5246,"n_observed":73223,"notes":"NULL on the locked rule. H1 SPATIAL FAILS ON DIRECTION, which the rule makes dispositive (\"NULL if H1 fails\"): fault-box summer-mean cloud fraction 0.392 vs laccolith-control mean 0.524, difference -0.132 (95% CI -0.140..-0.124, bootstrap P(diff<=0)=1.000, n_days 3158, n_hourly 73223). The fault box makes FEWER clouds than the matched laccoliths, the opposite of the venting prediction; H1 required p<0.05 AND positive direction.\n\nH3 KILLER also fails. M2 lunar 12.42h: p=0.5246 in the locked M2 test (n=73224) and p=0.25-0.27 on re-measure, against a required p<0.01 -- a fifty-fold miss, not a marginal one. Mf fortnightly reaches p=0.003 at the fault and clears the p<0.01 half of the criterion, BUT H3 also requires fault > controls by a preset margin and Mf is p=0.003 at henry, 0.003 at la_sal, 0.007 at abajo and 0.003 at ESCALANTE FLAT DESERT. Identical fortnightly power over flat ground with no laccolith and no fault means Mf here is regional weather, not venting. The pre-registered control set is the only reason that 0.003 did not read as a discovery.\n\nConsistent with the four sibling preregs locked 2026-07-14: M2 lunar disconfirming (p=0.5246), fortnightly/spring-neap null (snr 1.73 p=0.164; Mf secondary snr 0.95 p=0.512), seasonal hydrologic loading null (p=0.403, best control exceeded the fault), seismic dynamic triggering null (p=0.479, abajo_mtns control EXCEEDED the fault). The one sibling PASS was barometric pumping (beta=-0.02535, block-bootstrap p=0.001), which is a pressure-response result and does not bear on H1 or H3.\n\nH2 (dry-day moisture source) and H4 (20yr loading) were NOT run: the locked rule makes H1 dispositive for NULL, and running further sub-tests after the dispositive one fails would be searching for a rescue.\n\nmetric_value reported is the registered threshold metric H3_tidal_M2_or_Mf_peak_p using the M2 measurement (0.5246) against the locked threshold of 0.01.\n\nFraming per the lockfile: CONFIRMATORY on the existing 2018-2025 GOES archive, which existed at lock time -- the pre-registration prevents metric/control/period selection, it does not establish blindness. Site context: Pine Valley is the site whose customer report was retracted in the 2026-08-07 provenance incident, so a PASS here would have needed to be unimpeachable. It is a clean, well-controlled NULL instead.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-16T17:15:06Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{"atlas_followon":"After the Pine Valley verdict is recorded, apply the IDENTICAL locked method to the full replenishable atlas (including the 3 Great Salt Lake cloud-forming spots Mike flagged in the 2018 radar time-lapse) as a SEPARATE pre-registered test.","cloud_index":"GOES-West ABI cloud-formation index over each box: cold-cloud-top frequency (band 13 BT < threshold) + cloud fraction + WV (bands 8/9/10); hourly, ~2km. GOES-17 2018-2022, GOES-18 2023-present (NOAA Open Data/AWS).","controls":{"airmass_covariates":["upstream 700mb humidity","precipitable water (PWAT)","CAPE","box elevation (3DEP)","wind direction/speed"],"flat_terrain_control":{"lat":37.85,"lon":-113.55,"name":"Escalante Desert flat"},"orographic_laccolith_controls":[{"lat":38.1,"lon":-110.8,"name":"Henry Mountains"},{"lat":38.45,"lon":-109.25,"name":"La Sal Mountains"},{"lat":37.85,"lon":-109.47,"name":"Abajo Mountains"}],"tidal_null":"phase-randomized / AR-surrogate, 1000 draws"},"framing":"CONFIRMATORY on the existing 2018-2025 GOES archive. Pre-registration LOCKS the cloud metric, control set, period, tidal criterion, and PASS thresholds BEFORE any frame is pulled, to prevent metric/control/period selection bias (p-hacking). It does NOT establish blindness to the data -- the archive exists at lock time.","period":"2018-06-01..2025-09-30. Spatial tests on summer JJA; tidal spectral on full-year multi-year series.","primary_target":{"box_deg":0.25,"context":"artesian, 93pct pressurized, -17.68 mGal gravity low, pressurizing through 20yr drought (platform deep-water investigation)","lat":37.2,"lon":-113.5,"name":"Pine Valley Mountain / Hurricane Fault pressurized segment (Target A)"},"tests":{"H1_spatial":{"desc":"fault-box cloud-index excess vs mean of the 3 laccolith controls, AFTER regressing out box elevation + regional airmass","pass":"permutation p<0.05 AND positive effect direction"},"H2_moisture_source_not_ocean":{"desc":"on low-advected-moisture days (regional PWAT < 25th pctile / offshore flow), fault box stays elevated while controls stay clear","pass":"fault-vs-control cloud-index difference significant + positive on dry days"},"H3_tidal_KILLER":{"desc":"lunar M2 (12.42h) and/or fortnightly Mf (13.66d) spectral peak in the fault-box cloud index, SEPARATED from the solar thermal tide S2 (12.00h) using the >=7yr record; absent/weaker in orographic controls","pass":"M2 or Mf peak p<0.01 vs surrogate null AND fault > controls by preset margin"},"H4_20yr_loading":{"desc":"multi-year cloud anomaly vs GRACE-FO GWSA + platform artesian/dv/v signal, controlling ENSO/PDO + regional precipitation (the through-drought test)","pass":"partial correlation significant + positive; SECONDARY (GRACE is basin-scale ~300km, context not fault-localized proof)"}},"verdict_rule":{"CONFIRMED":"H3 (tidal) PASSES AND (H1 or H2 PASSES)","NULL":"H1 fails (fault clouds indistinguishable from orographic laccolith controls)","SUPPORTIVE":"H1 and H2 pass but H3 fails (localized anomalous clouds, no tidal signature)"},"withdraw_gate":"if GOES valid-frame coverage over the fault box <60pct OR <2 control boxes have comparable coverage -> underpowered, re-register with adjusted boxes"},"method":"cloud_formation_tidal_crosscorrelation","n_permutations":1000},"threshold":{"direction":"positive","metric":"H3_tidal_M2_or_Mf_peak_p","value":0.01},"title":"Deep-water fault venting at Pine Valley: anomalous tidally-modulated cloud formation over the Hurricane Fault pressurized segment","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"teleseismic RF (permanent FDSN broadband) + real CRUST1.0 Laske 2013","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"ddeb060ffccb0e6f1dc198f85d53f3ea7d8f6cdb","hypothesis_id":"20260713_vq42_virtual_moho_gauge_h_k_rece_80e2b823","locked_at":"2026-07-13T00:19:19Z","locked_file_sha256":"0546653a3ae095ab0eb9a27c99978e3d670ddf09d908bfa50dc19f2bdbadc8d3","notes":"VQ#42 STACK instrument. PASS = frac_within_5km >= 0.50 AND bootstrap skill-vs-IDW-of-CRUST1.0 CI_lo > 0, over a 30-station permanent-broadband set spanning ~25-50 km Moho. Instrument: RF + Zhu-Kanamori H-k over the FDSN accidental array. Truth: REAL CRUST1.0 (rejected the platform synthetic crust1_grid as fake). Engine validated on IU.ANMO H=32.5 vs 38.3 km, Ps 4.35s. Multi-station skill is the locked outcome.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2026-07-20","registered_at":"2026-07-13T00:19:19Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"ddeb060ffccb0e6f1dc198f85d53f3ea7d8f6cdb","interpretation":"disconfirming","metric_value":-3.227,"n_observed":22,"notes":"PARTIAL: recovers Moho within 5 km for 50% of stations (frac_within_5km=0.50) but skill-vs-IDW-of-CRUST1.0 CI_lo=-3.23 << 0 -> does NOT beat a smooth interpolation of the truth. Per-station RF quality uneven (half good, half off). Fails the locked AND-criterion. Engine validated (ANMO H=32.5 vs 38.3, Ps 4.35s); v2 = per-station RF-fit QC to lift the bad half.","ots_proof_path":null,"passed":false,"tested_at":"2026-07-13T00:57:59Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"hk_receiver_function_zhu_kanamori_vs_crust1","n_permutations":null},"threshold":{"direction":"positive","metric":"skill_vs_idw_ci_lo","value":0.0},"title":"VQ42 Virtual Moho Gauge: H-k receiver-function crustal thickness vs CRUST1.0","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"S-wave broadband displacement spectra via FDSN (response-removed) + USGS catalog Mw (magType mww/mw/mwr) for M4-6 California events, depth<35km, 14yr. Catalog Mw is the per-event truth. (already collected: YES)","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"c127f403474d6327a463aa592795bb08034a81bb","hypothesis_id":"20260713_vq43_virtual_single_station_mw_g_5bfdd766","locked_at":"2026-07-13T03:03:19Z","locked_file_sha256":"d805fa89f3cca12d2a6afaaddfa4b0c0247c3b8f7dd9899d5eebbc7d32b7c433","notes":"VQ#43 Virtual Single-Station Mw Gauge (SPLIT/spectral primitive). CLAIM: a single broadband stations S-wave DISPLACEMENT spectrum, response-removed and Brune omega-2 fit (Omega0,fc) with M0=4pi rho Vs^3 R Omega0/(Rad_S FS), recovers the USGS catalog moment magnitude within 0.3 Mw units for at least 60% of M4-6 California events (median over up to 5 nearest permanent broadband stations 60-400km). SPARSE-TRUTH bar: catalog Mw is per-event, well-measured, heterogeneous -- no smooth global field to interpolate (the RIGHT test, unlike Moho/geothermometer IDW wall). Smoke: event nn00888580 ours Mw=5.58 vs cat 5.71 dMw=0.13 (fc=0.14Hz, single station). Full ~55-event run in progress.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"reproducibility"},"reevaluate_at":"2026-07-13","registered_at":"2026-07-13T03:03:03Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"f5eadb3213ade1adb54eaba0d03c613fc93bb286","interpretation":"confirming","metric_value":0.78,"n_observed":null,"notes":"PASS: single-station Brune omega-2 displacement-spectrum moment recovers catalog Mw within 0.3 for 78% of 36 M4-6 California events (bar >=60%), median error 0.15 Mw, bias +0.04 (near-unbiased). Top recoveries within 0.01-0.05 Mw from ONE stations spectrum. Second sparse-truth reproducibility PASS this session (catalog Mw is per-event + well-measured, no smooth field to interpolate).","ots_proof_path":null,"passed":true,"tested_at":"2026-07-13T04:49:22Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"brune_omega2_spectral_moment","n_permutations":null},"threshold":{"direction":"positive","metric":"frac_dMw_within_0.3","value":0.6},"title":"vq43 virtual single-station Mw gauge Brune spectral moment reproducibility","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"Teleseismic deep-event (M6+, depth>=100km, 88-125deg) BH?/HH? waveforms via FDSN + published SWS-DB (Liu/EarthScope SWS-DB-MST) per-station fast-azimuth+delay reference. (already collected: YES)","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"97a1f4d040b8858077844da6a996dfc6be26b91d","hypothesis_id":"20260713_vq45_virtual_anisotropy_field_sk_ffb62804","locked_at":"2026-07-13T02:48:33Z","locked_file_sha256":"6732813a8727d3c89742ecbac2f017f19b7a785894d76b008768e6553e4ceb1b","notes":"VQ#45 Virtual Anisotropy Field (SPLIT primitive). CLAIM: Silver-Chan (1991) transverse-energy minimization on teleseismic SKS at a permanent broadband station, rotated to ZNE via inventory then R/T by back-azimuth over a multi-year deep-event catalog, recovers the SWS-DB published fast-axis azimuth (phi) within 25deg for at least 50% of permanent-network reference stations (>=4 published A/B measurements each) AND median delay-time error <=0.7s. This is a SPARSE-TRUTH reproducibility bar (per-station mantle anisotropy is heterogeneous, no smooth global field to interpolate -- the RIGHT test, unlike Moho/geothermometer IDW wall). Smoke n=3: HLID dphi=15/ddt=0.00, EYMN dphi=32/ddt=0.15, CMB dphi=31/ddt=0.50. Full ~40-station run in progress.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"reproducibility"},"reevaluate_at":"2026-07-13","registered_at":"2026-07-13T02:48:21Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"24b64e03b15ffe536fc1cf4721574bb6cdc3dd9b","interpretation":"disconfirming","metric_value":0.43,"n_observed":null,"notes":"PARTIAL (real instrument, below the >=0.50 bar): Silver-Chan SKS splitting recovers the SWS-DB fast-axis within 25deg for 43% of permanent stations (23 scored), median delay-time error 0.83s (bar 0.7). Delay recovery is decent; fast-axis has genuine scatter because automated single-event SKS is noisy vs the SWS-DB curated multi-event/backazimuth station averages (many are 2-layer). Machinery validated (HLID dphi 14.8, dt exact). v2 path: stack multiple events per station + backazimuth binning before comparing to the published average.","ots_proof_path":null,"passed":false,"tested_at":"2026-07-13T15:31:28Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"silver_chan_transverse_energy_minimization","n_permutations":null},"threshold":{"direction":"positive","metric":"frac_dphi_within_25","value":0.5},"title":"vq45 virtual anisotropy field SKS shear-wave splitting reproducibility","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"GOES-West ABI-L2-ACMC cloud_index.db 2017-2025 + Open-Meteo/USGS forcing","expected_direction":"negative","feature_set":["barometric_excess_p"],"git_commit_at_registration":"aba6cf27323e8dfd3e1bf1bd60be278881a7611b","hypothesis_id":"20260714_pine_valley_barometric_pumping_c_5d6c8b00","locked_at":"2026-07-14T22:36:04Z","locked_file_sha256":"f8a9f795fa2901dd112a3db0183b1b6a72412059f238dad76a3c4c330aac7b4f","notes":"PRE-REGISTRATION. CONFIRMATORY (data exists in the cloud/forcing/USGS caches; this analysis has run ZERO times on real data before this lock -- only --selftest ran). PASS = excess (fault beta - mean control beta) < 0 AND block-bootstrap p<=0.05. Low pressure eases fault confinement -> extra venting; discriminator is the fault's EXCESS negative pressure sensitivity beyond the shared weather effect. Script sha256: b4659bfa9067b6f7fb7db8a6bac93f6987e00683bd4f94ed5c21af555be1da79. Part of the Pine Valley release-mechanism series following the M2 null (20260714 ...b26f1d93). NULL is a real, publishable constraint.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-14","registered_at":"2026-07-14T22:36:04Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"b87aee0cf1deb545102252e3aa43496c1c749313","interpretation":"confirming","metric_value":0.001,"n_observed":null,"notes":"PASS on the LOCKED criterion: pine_valley_fault pressure-sensitivity beta=-0.02535 (the MOST negative box) exceeds the mean control (excess -0.00497, block-bootstrap p=0.001) -- the fault vents extra when barometric pressure drops. HONEST CAVEAT that tempers the venting interpretation: escalante_flat (the flat WEST control at -113.55) is statistically TIED at beta=-0.02529, while the 3 laccolith controls (all EAST, -0.018 to -0.019) are notably less sensitive. So the barometric sensitivity splits WEST(fault + escalante) vs EAST(laccoliths), not fault vs non-fault. The effect may be a regional west-east barometric-cloud-response difference rather than fault-specific venting. The pre-registered criterion (fault > mean control) genuinely passed and is not moved post-hoc, but a clean fault-venting confirmation needs a WEST non-fault control other than escalante, or an EAST fault, to disentangle. Barometric pumping is the ONE non-null of the four release tests.","ots_proof_path":null,"passed":true,"tested_at":"2026-07-14T23:26:51Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"partial regression of cloud residual on pressure residual per box; block-bootstrap of fault-minus-mean-control excess","n_permutations":null},"threshold":{"direction":"negative","metric":"barometric_excess_p","value":0.05},"title":"Pine Valley barometric-pumping cloud response (fault exceeds controls)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"GOES-West ABI-L2-ACMC cloud_index.db 2017-2025 + Open-Meteo/USGS forcing","expected_direction":"negative","feature_set":["fault_spring_neap_p"],"git_commit_at_registration":"aba6cf27323e8dfd3e1bf1bd60be278881a7611b","hypothesis_id":"20260714_pine_valley_fortnightly_spring_n_b97d3ad8","locked_at":"2026-07-14T22:36:04Z","locked_file_sha256":"a6c515918bcbf5d4e5b5cda847181be92212f257e7cd7a9bb77ce2b7fb7fdc89","notes":"PRE-REGISTRATION. CONFIRMATORY (data exists in the cloud/forcing/USGS caches; this analysis has run ZERO times on real data before this lock -- only --selftest ran). PASS = fault spring-neap p<=0.05 AND fault spring-neap snr > max control snr. Resolves the M2 null: a threshold-triggered fault release responds at the spring-neap ENVELOPE, not the M2 carrier, so a clean M2 null can coexist with a real fortnightly signal. Script sha256: b5531f369393f36898295aef4b159e657b73db39e37b7047f6a8de448eb5a7eb. Part of the Pine Valley release-mechanism series following the M2 null (20260714 ...b26f1d93). NULL is a real, publishable constraint.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-14","registered_at":"2026-07-14T22:36:04Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"8f02c39c4004b609738eca5d6c196f58095efa7a","interpretation":"disconfirming","metric_value":0.164,"n_observed":null,"notes":"NULL: fault spring-neap snr 1.73 p=0.164 (marginally above controls max 1.495 but NOT significant); Mf secondary null (snr 0.95 p=0.512). The tidal null DEEPENS -- neither the M2 carrier nor the spring-neap envelope paces the venting. Threshold-triggered tidal fault release disconfirmed at this floor.","ots_proof_path":null,"passed":false,"tested_at":"2026-07-14T22:46:01Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"LS spring-neap (14.7653d) + Mf (13.6608d) candidate line vs local null grid; fault-exceeds-controls","n_permutations":null},"threshold":{"direction":"negative","metric":"fault_spring_neap_p","value":0.05},"title":"Pine Valley fortnightly/spring-neap cloud modulation (threshold-tidal fault venting)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"GOES-West ABI-L2-ACMC cloud_index.db 2019-2025, 5 boxes","expected_direction":"positive","feature_set":["goes_acmc_cloud_frac_hourly"],"git_commit_at_registration":"f92b8ed1d203667b2828af6783a1f59a97d8af84","hypothesis_id":"20260714_pine_valley_m2_lunar_tidal_cloud_b26f1d93","locked_at":"2026-07-14T18:43:11Z","locked_file_sha256":"097cfbee32e78e9c0dfe0512aaf4c353444e323ce39839035cd0d32e251d8816","notes":"CONFIRMATORY not blind: the cloud index existed at lock time but the analysis ran zero times on real data pre-lock (only --selftest on synthetic; selftest PASS: injected 2% M2 detected snr 11.3 p 0.016, control clean null). Protocol locked verbatim in scripts/deep_water_cloud/pine_valley_m2_test.py sha256=1dd234aab8723585685ce71b26e36fcc9edda7fd83cae93ae597e7c6f249c5a8. Solar weather can only make 24h harmonics; 12.4206h is lunar-only. Global atmospheric lunar tide (Kohyama & Wallace 2016, ~1% rain modulation) hits all 5 boxes equally -- hence the controls-exceedance requirement. Test box pine_valley_fault vs henry_mtns/la_sal_mtns/abajo_mtns/escalante_flat. A NULL is a real publishable constraint on tidally-modulated fault venting. Replication set: the 43-box UT+NV cloud DB (building now) -- a separate lock before any UT+NV run.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"deep_water_venting"},"reevaluate_at":null,"registered_at":"2026-07-14T18:42:55Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"c9dbd0b84843dca02b1a7f3f386b9260ce219dd1","interpretation":"disconfirming","metric_value":0.5246,"n_observed":73224,"notes":"","ots_proof_path":null,"passed":false,"tested_at":"2026-07-14T18:45:15Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"base_model":"DC+trend+annual+semiannual+S1..S4","n_null_m2":60,"n_null_o1":40,"pass":"fault M2 p<=0.05 AND fault M2 snr > max(4 controls)","secondary_period_h":25.8193417,"target_period_h":12.4206012,"withdraw":"fault n<30000"},"method":"ls_harmonic_line_vs_local_null_grid","n_permutations":null},"threshold":{"direction":"positive","metric":"m2_empirical_p","value":0.05},"title":"Pine Valley M2 lunar-tidal cloud modulation (fault venting)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"GOES-West ABI-L2-ACMC cloud_index.db 2017-2025 + Open-Meteo/USGS forcing","expected_direction":"negative","feature_set":["seasonal_lagged_excess_p"],"git_commit_at_registration":"aba6cf27323e8dfd3e1bf1bd60be278881a7611b","hypothesis_id":"20260714_pine_valley_seasonal_hydrologic__e82c2420","locked_at":"2026-07-14T22:36:04Z","locked_file_sha256":"67db91f0611d86610039730cfcb1ffd6f4fab31474a7bf3c4eb290c8c936024f","notes":"PRE-REGISTRATION. CONFIRMATORY (data exists in the cloud/forcing/USGS caches; this analysis has run ZERO times on real data before this lock -- only --selftest ran). PASS = fault peak-lagcorr > every control AND phase-randomised p<=0.05. Interannual loading anomaly -> lagged venting anomaly, fault-attributable. WEAKEST of the four (snow_depth is a loading proxy, broad lag band) -- a null here is weaker evidence than #1/#2/#4. Script sha256: c5359b30ab170ca229d73615e7eae0e332ea0357da6d24cfb781adc8df42078f. Part of the Pine Valley release-mechanism series following the M2 null (20260714 ...b26f1d93). NULL is a real, publishable constraint.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-14","registered_at":"2026-07-14T22:36:04Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"b87aee0cf1deb545102252e3aa43496c1c749313","interpretation":"disconfirming","metric_value":0.4032,"n_observed":null,"notes":"NULL: fault peak lagged (7-84d) loading cross-correlation +0.068 does NOT exceed the best control (+0.075, la_sal_mtns); excess -0.0064, phase-randomised p=0.403. No fault-specific seasonal-loading venting response detectable (the weakest test, as pre-flagged: snow_depth proxy + broad lag band).","ots_proof_path":null,"passed":false,"tested_at":"2026-07-14T23:26:51Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"peak lagged (7-84d) cross-correlation of cloud anomaly vs regional snow-load anomaly per box; phase-randomised null","n_permutations":null},"threshold":{"direction":"negative","metric":"seasonal_lagged_excess_p","value":0.05},"title":"Pine Valley seasonal hydrologic-loading lagged venting (fault exceeds controls)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"GOES-West ABI-L2-ACMC cloud_index.db 2017-2025 + Open-Meteo/USGS forcing","expected_direction":"negative","feature_set":["seismic_postevent_excess_p"],"git_commit_at_registration":"aba6cf27323e8dfd3e1bf1bd60be278881a7611b","hypothesis_id":"20260714_pine_valley_seismic_dynamic_trig_300290ad","locked_at":"2026-07-14T22:36:04Z","locked_file_sha256":"e5fa0a35b8d0d9a0f14cf6ec64adbab274377d366981c72b912adcecc20fc4e5","notes":"PRE-REGISTRATION. CONFIRMATORY (data exists in the cloud/forcing/USGS caches; this analysis has run ZERO times on real data before this lock -- only --selftest ran). PASS = fault post-event (0-3d) minus pre-event (-7..-1d) excess p<=0.05 AND fault excess > max control excess. Earthquakes transiently boost fault permeability -> a venting burst. Script sha256: 106eb239488fda034ef01db70f7d1b4a0dff032dfd5ea04e31a869395048c810. Part of the Pine Valley release-mechanism series following the M2 null (20260714 ...b26f1d93). NULL is a real, publishable constraint.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":"2027-07-14","registered_at":"2026-07-14T22:36:04Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"8f02c39c4004b609738eca5d6c196f58095efa7a","interpretation":"disconfirming","metric_value":0.479,"n_observed":null,"notes":"NULL: fault post-event (0-3d minus -7..-1d) cloud excess +0.0014 p=0.479 on 569 events (local M>=2.5 + teleseismic M>=6.5); a control (abajo_mtns +0.0051) EXCEEDS the fault. No detectable seismic dynamic-triggering of venting-clouds over the fault vs controls.","ots_proof_path":null,"passed":false,"tested_at":"2026-07-14T22:46:01Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"superposed-epoch post-minus-pre cloud residual around local M>=2.5 + teleseismic M>=6.5 events; event-time-shuffle null","n_permutations":null},"threshold":{"direction":"negative","metric":"seismic_postevent_excess_p","value":0.05},"title":"Pine Valley seismic dynamic-triggering venting burst (superposed-epoch, fault exceeds controls)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"GOES-West ABI-L2-ACMC clear-sky-mask 2019-2025 summers (JJA); sites locked in scripts/deep_water_cloud/utnv_sites.json sha256 ca775ad5455d97f8affd78f098fce4c2b74579b0963f474235cfbdc1a786adf6","expected_direction":"negative","feature_set":[],"git_commit_at_registration":"f3a627c803962ce2d6d8b8cc0b3e0c42528f651c","hypothesis_id":"20260714_ut_nv_fault_venting_flatter_summ_4e5e8a1c","locked_at":"2026-07-14T13:32:05Z","locked_file_sha256":"f84656b64421be201490c7646e14a3adf7d8a69c0ff66c22c2aa644a8a138bf1","notes":"PASS = Mann-Whitney one-sided p(diurnal amplitude test < control) < 0.05 AND median(test) < median(control), over 22 fault-venting-candidate sites vs 21 matched flat-basin controls in scripts/deep_water_cloud/utnv_sites.json (sha256 ca775ad5455d97f8affd78f098fce4c2b74579b0963f474235cfbdc1a786adf6). All boxes are basin/valley terrain so amplitude isolates venting (non-solar flat diurnal) from topography. Secondary: same test on afternoon_excess. WITHDRAW if <12 sites/group with >=2 summers of data. Analysis: scripts/deep_water_cloud/utnv_analysis.py on data/deep_water_cloud/cloud_index_utnv.db. Locked before the UT+NV cloud data was touched.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"deep_water_venting"},"reevaluate_at":"2026-07-21","registered_at":"2026-07-14T13:32:05Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"421cb401a15152c29e3acf11f46cdc34bbc39d94","interpretation":"disconfirming","metric_value":0.1626,"n_observed":43,"notes":"NULL (well-powered, not withdrawn). Primary: MWU one-sided p(amplitude test<control)=0.1626 (not <0.05) AND median(test)=0.2198 > median(control)=0.2181 (opposite of the hypothesized direction) -> fails BOTH PASS conditions. 22 test / 21 control basin sites, ~14,900 days / 7 years each. Secondary afternoon_excess: test 0.1035 < control 0.1282 in the hypothesized direction but p=0.0644 (trend, n.s.). Geothermal/hot-spring sites do NOT show a distinctive flatter (venting-suppressed) summer diurnal cloud-amplitude signature vs matched flat-basin controls. Result: reports/deep_water_cloud/utnv_result.json. Confirmatory null -> Methodological Nulls Bundle candidate.","ots_proof_path":null,"passed":false,"tested_at":"2026-07-15T03:10:44Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"mann_whitney_u_one_sided","n_permutations":null},"threshold":{"direction":"negative","metric":"diurnal_amplitude_mwu_p","value":0.05},"title":"UT+NV fault-venting flatter summer diurnal cloud cycle vs matched flat-basin controls","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"regional_microseism_stack_monitor + scan_convergence(prediction_ledger, min_models=3) + USGS FDSN M5.5+","expected_direction":"positive","feature_set":["regional_microseism_z","n_converged_models"],"git_commit_at_registration":"4135e7b7cdf2ae8002b145ade16a8a2cc0764aa7","hypothesis_id":"20260715_convergence_gated_microseism_10h_0eff9c3a","locked_at":"2026-07-15T01:18:46Z","locked_file_sha256":"aa862cfbf81a390b415a15e9af4eec12f563209428b58706e48a8888bd8c92de","notes":"SHADOW-only convergence-gated 10h/36h microseism notice (cross_physics/microseism_gated_notice.py). Gates the Phase-5 regional microseism-z excursion (z>=3.0) on the region being convergently loaded (>=3 distinct models at HIGH/CRITICAL via scan_convergence). PASS = gated forward precision >= 0.20 at n>=20 RESOLVED gated notices AND gated > ungated control precision over the same window. WITHDRAW if < 10 resolved gated by re-eval. Ungated notice (pre-reg 20260701..47d7b6f5, ~3.4pct at n=29) is the control. Never touches the public ledger. Gate calibration: min_models=2 loads 13/16 monitored (loose), =4 loads 1/16 (=72h tier), =3 loads ~6/16 (selective).","ots_proof_path":null,"outcome":{"lookahead_hours":36.0,"min_magnitude":5.5,"type":"earthquake"},"reevaluate_at":"2026-10-11","registered_at":"2026-07-15T01:18:46Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"91fc6a7840a514aba8337e181af76b3c020e6cc9","interpretation":"disconfirming","metric_value":0.0368,"n_observed":null,"notes":"DISCONFIRMED / PREMISE VOID by independent verification wave (trackAR, 2026-08-24). This convergence-gated 10h notice product GATES on the Phase-5 microseism z-excursion, which is now retracted: the feature microseism_secondary_band_3s (= MUSTANG pct_above_nhnm) is a BROADBAND, whole-day, daily-cadence metric that any earthquake energy inflates; on a daily feature \"-10h\" lands on the same UTC day as the event 58% of the time, so the notice is concurrent, not precursory. On the un-biased band_rms metric (0.14-0.5 Hz hourly, clean baseline, GK74 declustering) the -10h z is +0.08/-0.23/-0.08 across magnitude bins with NO scaling -- decisively excluded (positive control event-hour z=+30.3). With no valid pre-event signal to gate on, the gated forward precision is not distinguishable from the in-region M5.5+ base rate (~0.0368), far below the 0.20 threshold. metric_value stamped = base rate 0.0368 (an INFERENCE from the retraction, NOT a fresh forward-precision measurement). Seal + hash chain + OTS anchor preserved as the record of what was claimed.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-26T18:34:02Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"forward precision of convergence-gated microseism notices (z>=3.0 AND >=3 distinct models HIGH/CRITICAL in region); a notice HITS iff an M>=5.5 occurs in the region bbox within 36h; ungated microseism_shadow_notices.db is the live control","n_permutations":null},"threshold":{"direction":"positive","metric":"precision","value":0.2},"title":"Convergence-gated microseism 10h notice beats ungated","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"USGS FDSN triggers (CONUS+AK+HI, 8yr) + multi-catalog M2.5+ neighbourhoods (aftershock_behavior_study cache)","expected_direction":"positive","feature_set":["max_early_mag_ratio","rate_accel_late_over_early","n_early"],"git_commit_at_registration":"7426d3399fd965870aad927a68d25f686f75234b","hypothesis_id":"20260715_escalation_signature_scale_invar_3a64d89f","locked_at":"2026-07-15T02:32:03Z","locked_file_sha256":"0e6fa7407796c27fd32739633d73b13b99c2d3c435c211d6c5cb40f41d1cea0a","notes":"The cascade-vs-pre-slip question operationalized: is the escalation signature scale-invariant? PASS = min(transfer AUC H->L, L->H) >= 0.60 AND each transfer >= 0.85x the target band's own 5-fold in-band CV AUC. WITHDRAW if < 150 feature-sequences or < 12 escalation positives in either band. CONFIRMATORY not blind: the catalog exists at lock, but this cross-band transfer analysis has never been run on real data (selftest synthetic-only, all 3 verdicts verified). Script committed pre-lock at blob sha 4e48aedcbe582af9a457b5baedfdfd86bea3f586 (commit 7426d339). DECISION RULE: PASS -> build the short-retention M3.5-4.5 event-library tier (small events are valid training data for large-event precursor learning); FAIL -> do not build it and document the scale-dependence. Secondary (report-only): Jones-Molnar base-rate scale-independence (two-proportion z), in-band CV AUCs per band. Same-window/same-boxes design kills the temporal+regional confound; per-band z-scoring tests direction+relative weighting of the signature, not raw scales (productivity trivially scales with M).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"earthquake"},"reevaluate_at":"2026-07-22","registered_at":"2026-07-15T02:32:03Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"e7f52ce0def62eb4eb23a39337041d64220ec121","interpretation":"confirming","metric_value":0.838,"n_observed":685,"notes":"PASS_scale_invariant. Primary min transfer AUC = 0.838 (>= 0.60 bar); HIGH->LOW 0.890, LOW->HIGH 0.838; relative transfer/in-band 0.986 and 0.969 (both >= 0.85) -- the early-window escalation signature transfers ~97-99pct of each band's own 5-fold CV skill (in-band LOW 0.903 / HIGH 0.865). Well-powered: LOW 535 seq/209 esc, HIGH 150 seq/18 esc (both > WITHDRAW gate). HONEST NUANCE: the SIGNATURE is scale-invariant but the base RATE is not (LOW escalates 39pct vs HIGH 12pct, z=6.22) -- partly a label-definition artifact ('comparable-or-larger later event' is easier to satisfy at low mag). Supports cascade over pre-slip nucleation for the early-window signature. DECISION per locked rule: PASS -> M4-class catalog signature predicts M5+ escalation, so small events ARE valid training data -> BUILD the short-retention M3.5-4.5 waveform library tier (a now-justified experiment to test whether WAVEFORM precursors also transfer, which this makes plausible). Result: reports/escalation_scale_invariance/result.json.","ots_proof_path":null,"passed":true,"tested_at":"2026-07-15T03:22:03Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"cross-band transfer AUC of the 24h escalation logistic: fit LOW (M4.0-4.9) frozen -> test HIGH (M5.0+), and vice versa; per-band z-scoring; escalation = subsequent event >= trigger-0.05 within max(WC,30km) in 20d","n_permutations":null},"threshold":{"direction":"positive","metric":"min_transfer_auc","value":0.6},"title":"Escalation-signature scale invariance M4 vs M5 plus (US)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"dvv_reference.db + grace_fo_gwsa_cache.db + snotel_swe_cache.db","expected_direction":"positive","feature_set":["dvv_loading_slope","grace_gwsa","snotel_swe"],"git_commit_at_registration":"44f6b50db77982e1ce685cc4915e5d8e723f8710","hypothesis_id":"20260715_replenishable_basin_loading_is_d_06cf1059","locked_at":"2026-07-15T21:36:49Z","locked_file_sha256":"32a9c67c770fc9dd5233002953af5563a6d1dcf8a3f79d6e7409af1f98375c51","notes":"PRIMARY: Spearman(loading_rate, surface-decoupling) across replenishable UT basins >= 0.60 AND perm p<0.05, >=6 basins with >=24 overlapping months. loading=OLS slope of monthly dv/v anomaly (%/yr); decoupling = -mean(r(dvv,GRACE_GWSA), r(dvv,SWE)). Deep fault-fed recharge loads while surface depletes -> decoupled, strongest where loading fastest. NULL (rho~0/neg) is publishable. CONFIRMATORY (data exists, analysis not yet run). script sha256(16)=375025627093ad0c. selftest PASS (planted rho=1.0 p=0.0005; noise rho=-0.17 p=0.70).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"hydrology"},"reevaluate_at":"2027-07-15","registered_at":"2026-07-15T21:36:33Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"44f6b50db77982e1ce685cc4915e5d8e723f8710","interpretation":"disconfirming","metric_value":0.3214,"n_observed":7,"notes":"","ots_proof_path":null,"passed":false,"tested_at":"2026-07-15T21:39:42Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"spearman_loading_vs_surface_decoupling","n_permutations":2000},"threshold":{"direction":"positive","metric":"spearman_rho","value":0.6},"title":"Replenishable-basin loading is deep-decoupled from surface mass, strongest where loading rate is highest","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"prediction_ledger.db resolved EQ forecasts, pre vs post 2026-07-16","expected_direction":"positive","feature_set":["mag_scaled_window_hours","expected_mag_max"],"git_commit_at_registration":"425f8d09590dd61df788109eb944509e631e7699","hypothesis_id":"20260716_magnitude_scaled_forecast_window_fb872e19","locked_at":"2026-07-16T17:53:19Z","locked_file_sha256":"a2f994ebca7385a4d7822ee42fa2f7d26a6e8f5d6f423449db61aa345c30469b","notes":"The magnitude-scaled forecast window (M6.5+ 21d / M6-6.5 14d / M5.5-6 10d / M<5.5 7d, live 2026-07-16) is predicted to RAISE the resolved hit rate for high-mag EQ forecasts (expected_mag_max>=6.0) -- by crediting the true longer lead AND eliminating premature-expiry misses (early forecasts that used to expire before the quake) -- WITHOUT moving the M<6 bulk. PASS: hit_rate(M6+, forecasts logged on/after 2026-07-16) > hit_rate(M6+, pre-change 60d baseline) AND |hit_rate(M<6, post) - hit_rate(M<6, pre)| <= 2pp, at n>=30 resolved M6+ forecasts post-change. Resolved = matched OR expired; binned by expected_mag_max. WITHDRAW if <30 resolved M6+ post-change by re-eval. Honest guardrail: if M<6 moves >2pp, the change had an unintended side effect and must be revisited.","ots_proof_path":"hypotheses/locked/20260716_magnitude_scaled_forecast_window_fb872e19.json.ots","outcome":{"lookahead_hours":504.0,"min_magnitude":6.0,"type":"earthquake"},"reevaluate_at":"2026-10-16","registered_at":"2026-07-16T17:53:03Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"0bfce8aec2ed9763ddf8ecccfa629e23bc239b72","interpretation":"disconfirming","metric_value":-0.017,"n_observed":2784,"notes":"Graded 2026-09-02 (n>=30 met 90x over, before the 2026-10-16 re-eval). FULLMAG M6+ hit rate post 0.612 vs pre-60d 0.629 (-1.7 pp, NOT UP); LEDGER any-observation rate +6.8 pp (coverage from 21-day windows, not catches). M<6 guardrail exceeded (+29.8 pp LEDGER / +37.0 pp FULLMAG; post n=75, partly compositional). FAIL under both definitions. Full table: reports/validation/forecast_window_ladder_grade_2026-09-02.md. Ladder retired for new rows, kept as shadow key.","ots_proof_path":null,"passed":false,"tested_at":"2026-09-02T18:20:25Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"hit_rate_comparison_pre_post","n_permutations":null},"threshold":{"direction":"positive","metric":"m6plus_resolved_hit_rate_delta_pp","value":0.0},"title":"Magnitude-scaled forecast window raises M6+ hit rate without moving the M<6 bulk","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"USGS FDSN catalogue + retro/live tensor, regions japan indonesia chile_peru","expected_direction":"positive","feature_set":["ETAS Ogata 1998 rate -> P(>=1 event) vs tensor_eq_per_region calibrated probability"],"git_commit_at_registration":"dd2e854bdafe5d996d31f0279c93510aaab5dd22","hypothesis_id":"20260719_etas_derived_per_region_probabil_0ad8c552","locked_at":"2026-07-19T19:24:05Z","locked_file_sha256":"9fb3d2fa04480c5a2acabf44e39265b4ea471589b5f50d931f3f7db3ce61a552","notes":"PRE-REGISTERED BEFORE the per-region temporal validation finished. At lock time 26 of ~103 regions were evaluated showing mean k-fold 0.992 vs mean HELD-OUT 0.496, i.e. the incumbent is at chance out-of-sample. The criterion is fixed HERE precisely because beating a chance-level incumbent is a LOW BAR and could be satisfied by anything mildly non-random. PASS THEREFORE REQUIRES BOTH: (a) ABSOLUTE BAR -- ETAS held-out AUC >= 0.60 in at least 2 of the 3 regions; AND (b) RELATIVE BAR -- ETAS held-out AUC strictly greater than the incumbent held-out AUC in at least 2 of the 3 regions. Both must hold. Beating the incumbent while below 0.60 absolute is an explicit FAIL and means BOTH models are unusable, which is itself a publishable result. WITHDRAW AS UNDERPOWERED iff any region yields fewer than 30 test windows or fewer than 10 positive labels. Regions japan, indonesia, chile_peru, the same three used for every CSEP comparison this week. A PASS gates a BROADER validation across all regions BEFORE any production swap -- passing 3 regions does not by itself authorise replacing the live model. A NULL IS PUBLISHABLE and would mean per-region EQ forecasting needs a different object entirely.","ots_proof_path":"hypotheses/locked/20260719_etas_derived_per_region_probabil_0ad8c552.json.ots","outcome":{"lookahead_hours":168.0,"min_magnitude":4.5,"type":"earthquake"},"reevaluate_at":"2026-07-26","registered_at":"2026-07-19T19:23:47Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"dbfe90630b0f15ca182ae4f0d5059ab4b3b6b3c5","interpretation":"disconfirming","metric_value":1.0,"n_observed":3,"notes":"","ots_proof_path":null,"passed":false,"tested_at":"2026-07-20T00:06:26Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"Both models emit P(>=1 event of M>=region floor within the 7-day window). Scored on an IDENTICAL temporal holdout (first 70pct train, last 30pct test), identical labels, identical windows. Metric: held-out AUC, plus Brier and calibration reported. ETAS P = 1-exp(-lambda) with parameters fit on TRAIN windows only; incumbent scored with its own production pipeline on the same test windows.","n_permutations":null},"threshold":{"direction":"positive","metric":"regions_where_etas_beats_incumbent_AND_clears_absolute_bar","value":2.0},"title":"ETAS-derived per-region probability beats the incumbent tensor_eq_per_region classifier on a temporal holdout","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"retro_tensor.db 2005-2018 + USGS FDSN catalogue, 3 CSEP regions","expected_direction":"positive","feature_set":["retro_tensor_112_features_pca10"],"git_commit_at_registration":"2043b83ee58b477b790467489bfcbe82d2f033a6","hypothesis_id":"20260719_tensor_adds_magnitude_informatio_3b3bf3bc","locked_at":"2026-07-19T18:25:43Z","locked_file_sha256":"609a3009e4c861a1535040b79c773712b3ae5ed16463fa84638db58f41039c50","notes":"PRE-REGISTERED BEFORE THE ANALYSIS RAN. Analysis fixed by scripts/csep_tensor_magnitude_on_etas.py sha256 8748731c144b667b3d87c058438767921e7be91b32106a3d30ef0f2660dd4fc0 (committed before this lock). Third framing after two nulls: rate (japan +0.057, indonesia -0.009, chile_peru +0.008) and spatial (japan -0.032, indonesia +0.031 but shuffle p=0.53, chile_peru -0.015). Choosing a third test after two failures is itself a multiple comparison, so the criterion is fixed here and not decided from output. REGIONS japan, indonesia, chile_peru. A region PASSES BOTH GATES iff per-window 95pct CI lower bound > 0 AND shuffle-null p < 0.05. PASS iff (1) at least 2 of 3 regions pass both gates AND (2) pooled per-event IG summed across the 3 regions > 0 AND (3) within each passing region positive in at least 2 of 3 n_pca configurations (5,10,20). FAIL otherwise, including the case where exactly one region passes -- that is the non-replication pattern both prior framings produced. WITHDRAW AS UNDERPOWERED iff any region yields fewer than 40 usable windows or fewer than 200 test events. Count and location are not modelled so the gain is magnitude by construction. Control is a recalibrated b, not the historical b. A NULL IS A PUBLISHABLE OUTCOME and completes the three-framing sweep.","ots_proof_path":"hypotheses/locked/20260719_tensor_adds_magnitude_informatio_3b3bf3bc.json.ots","outcome":{"lookahead_hours":720.0,"min_magnitude":4.5,"type":"earthquake"},"reevaluate_at":"2026-07-20","registered_at":"2026-07-19T18:25:29Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"d7dd8e0c94ba7046a2ba34540d1a4b763923e424","interpretation":"disconfirming","metric_value":0.0,"n_observed":3,"notes":"","ots_proof_path":null,"passed":false,"tested_at":"2026-07-19T19:00:50Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"GR magnitude log-likelihood, beta_w = exp(a0 + a.z_w) fit on train windows by ML; scored per-event vs a single ML-refit constant b; circular row-shuffle null","n_permutations":200},"threshold":{"direction":"positive","metric":"regions_passing_both_gates","value":2.0},"title":"Tensor adds magnitude information on top of ETAS beyond a recalibrated b-value","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"USGS FDSN catalogue, chile_peru region bounds","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"6e4323f67da59656a7fc7a493a48d00e088b9a52","hypothesis_id":"20260720_chile_peru_etas_7_day_probabilit_0aa760c4","locked_at":"2026-07-20T01:09:10Z","locked_file_sha256":"f5fcb472e53414b6b91d1812daad5039df4c0a3d064c70edfaf977df8b8d5501","notes":"CONFIRMATORY test of a lead, run ONCE. Lead origin: Track A (hyp 20260719_etas_derived_per_region_probabil_0ad8c552) FAILED at 1 of 3 regions; the one that cleared was chile_peru 0.695. The rarer-label sweep then found chile_peru 0.523-0.695 in all 9 of its configurations while japan and indonesia stayed at chance. Label-independent and looks real, but noticed inside a failure and since examined across TWELVE configurations, which is a search. HOLDOUT IS GENUINE: all prior chile_peru configurations were scored inside the retro tensor window 2005-2018 because the incumbent arm needed tensor features; ETAS needs only the catalogue, so 2019-01-01 to 2026-07-01 has never been examined for this question. Nothing refitted: ETAS formulation, Reasenberg-Jones parameters, 7-day window, 365-day lookback and region magnitude floor all carry over from Track A unchanged. ANALYSIS FIXED BY scripts/chile_peru_etas_holdout.py sha256 d9e7e7fe33dc0072fa10b715d14de0b7b35a256b03ee38009b4a021274d346af committed before this lock. PASS = chile_peru AUC >= 0.60. FAIL = below, and the lead is dropped. WITHDRAW GATE: fewer than 30 positive windows in the held-out period. SECONDARY ARM (reported, NOT gating): the same fixed configuration on ten unexamined subduction regions, fixed in the script before running, testing generalisation only -- a failure there refutes a subduction-wide claim but says nothing about chile_peru specifically. SCOPE ON PASS: report as possible regional skill in ONE subduction zone. Never a platform claim. Re-eval is same-day because the holdout data already exists; the discipline here is that the criterion is locked before the run, not that time must pass.","ots_proof_path":null,"outcome":{"lookahead_hours":168.0,"min_magnitude":5.0,"type":"earthquake"},"reevaluate_at":"2026-07-20","registered_at":"2026-07-20T01:09:00Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"0a6cc78373b89e6e98ca0da8ebe268cdadeccd4e","interpretation":"disconfirming","metric_value":0.493,"n_observed":338,"notes":"","ots_proof_path":null,"passed":false,"tested_at":"2026-07-20T01:11:46Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"ROC AUC of ETAS P(>=1 event >= region floor in 7d) vs observed, single confirmatory run","n_permutations":null},"threshold":{"direction":"positive","metric":"auc","value":0.6},"title":"chile_peru ETAS 7-day probability retains AUC >= 0.60 on held-out 2019-2026","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"live microseism secondary-band z (GSN median, data/live_microseism.db) vs USGS global M7+","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"238dc1da0780734ecd81040170511ff20a8ec3f8","hypothesis_id":"20260720_microseism_precursor_operating_p_61a24c94","locked_at":"2026-07-20T01:15:08Z","locked_file_sha256":"5871865aa5b178755e032022b9480a9853527b1e673c1028987014494c687140","notes":"FORWARD test of the Track B operating characteristic. Track B chose z>=10 on a strict era split (fit 2005-2014, scored 2015-2018) giving precision 0.222, recall 0.607, 28.5 false alarms per year against a base rate of 0.0368. An era split is still retrospective; this applies the SAME point forward to data that did not exist when it was chosen. NOTHING RE-TUNED: threshold, 10-hour lag, 24-hour claim window, 24-hour debounce and M7.0 target all carry over unchanged. Analysis fixed by scripts/track_b_forward_shadow.py sha256 fae0bfa01401dbfee04bb952ac49c21db1614bd2cc0018a9665e5edf2e4abd6e committed before this lock. GLOBAL BY CONSTRUCTION: the secondary band is a median across GSN stations, so an alert claims an M7+ somewhere in the window, not where -- matching what the retrospective scored rather than quietly claiming something narrower. PASS = precision >= 0.10 over at least 30 resolved alerts. Bar is 0.10 rather than the measured 0.222 because 0.10 was the threshold declared in advance in Track B and moving it to the observed value would be fitting the bar to the result. WITHDRAW GATE: fewer than 30 resolved alerts by the re-evaluation date means underpowered, extend rather than report. HONEST EXPECTATION: at 22 percent roughly one alert in 4.5 is real, so a run of misses is the detector behaving normally and precision must NOT be read before n reaches 30. SHADOW ONLY: no ledger entry, no public surface, no advisory. Not for customer use until this passes. Re-eval 2027-09-01 is ~13.5 months, the time needed to accrue 30 alerts at the measured 28.5 per year.","ots_proof_path":null,"outcome":{"lookahead_hours":24.0,"min_magnitude":7.0,"type":"earthquake"},"reevaluate_at":"2027-09-01","registered_at":"2026-07-20T01:14:53Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"91fc6a7840a514aba8337e181af76b3c020e6cc9","interpretation":"disconfirming","metric_value":0.0368,"n_observed":null,"notes":"DISCONFIRMED / PREMISE VOID by independent verification wave (trackAR, 2026-08-24). This Track-B operating point (z>=10) reads directly off the retracted feature microseism_secondary_band_3s (= MUSTANG pct_above_nhnm), a BROADBAND, whole-day, daily-cadence metric inflated by each event own energy; \"-10h\" on a daily feature samples the event own UTC day 58% of the time (concurrent, not precursory). The era-split precision 0.222 in this manifest is inflated by that same-day self-contamination. On the un-biased band_rms metric (0.14-0.5 Hz hourly, clean baseline, declustered) the -10h z is ~0 with no magnitude scaling -- decisively excluded (n=82 M7.5+ would give t~13; measured -0.09; positive control event-hour z=+30.3). Forward precision on the clean channel is not distinguishable from the M5.5+ base rate (0.0368), below the 0.10 threshold. metric_value stamped = base rate 0.0368 (an INFERENCE from the retraction, NOT a fresh measurement). Seal + hash chain + OTS anchor preserved.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-26T18:34:03Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"forward shadow: alert when z>=10, claims M7+ globally within lag 10h +/- 12h, 24h debounce; precision over resolved alerts","n_permutations":null},"threshold":{"direction":"positive","metric":"precision","value":0.1},"title":"Microseism precursor operating point z>=10 holds forward at precision >= 0.10","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":"2027-09-01"},{"data_source":"retro_microseism_hourly.db + GSN IU/II BHZ backfill (phase6_v2); USGS M>=5.5 GK74-declustered","expected_direction":"negative","feature_set":["microseism_band_rms_secondary"],"git_commit_at_registration":"31c6f2e93d75ea1756b73a7691f5c210e143bc83","hypothesis_id":"20260724_microseism_precursor_spatial_mec_ea481761","locked_at":"2026-07-24T19:35:25Z","locked_file_sha256":"59d98e836c821d9ebfed08906bbf2333ed648bccb5f443a951cc8bf69f43dfcf","notes":"MICROSEISM PRECURSOR -- SPATIAL MECHANISM (coherence) EXPERIMENT. Pre-registered BEFORE the global-station data is built or any spatial statistic is computed.\n\nBACKGROUND. The ~10h microseism precursor (feature microseism_band_rms_secondary, secondary band 0.14-0.5 Hz, lag -10h) survived the Jul-2026 audit (Gardner-Knopoff declustered 9.40 sigma, N=3565; magnitude gradient Cohen's d 0.057 at M5.5-6.0 rising to 0.869 at M7+). The audited headline was computed on a GLOBAL AGGREGATE (median across a ~7-12 station western-US/Pacific GSN pool), so it cannot say WHETHER the -10h anomaly is (i) an epicentre-organized SOURCE/nucleation signal, (ii) an AMBIENT ocean-storm microseism field that merely correlates with large-event occurrence, or (iii) STATION-LOCAL noise averaged into an apparent signal. This experiment discriminates the three by the SPATIAL structure of the per-station anomaly. It can DEFLATE our own signal; that is the point.\n\nDATA (frozen). Per-station secondary-microseism band RMS (0.14-0.5 Hz, BHZ, m/s^2) computed by scripts/phase6_v2_microseism_backfill.py, hourly, 2010-01-01..2026-04-16. STATION-SELECTION RULE (fixed here, a COVERAGE decision made before any outcome): all GSN-backbone stations in networks IU and II with a BHZ channel and >=70% hourly microseism coverage over the span. This EXPANDS the existing 7-station pool (retro_microseism_hourly.db: ANMO COR ISA KIP PASC PFO PGC -- too few and too Pacific-clustered to resolve spatial structure) to a globally-distributed pool via an idle-compute backfill using the SAME builder. Catalog: USGS M>=5.5 GLOBAL with coordinates, Gardner-Knopoff-1974 declustered (common.declustering), restricted to the coverage span.\n\nANOMALY (faithful to the audit's stacked_z). For event E and station s, a(s,E) = (v_s(t_E - 10h) - mu_s)/sigma_s, where v_s is station s's hourly microseism_band_rms_secondary and mu_s,sigma_s are the mean/std of station s's OWN full series. d_epi(s,E) = great-circle station-to-epicentre distance (km).\n\nPRIMARY TEST 1 (source/near-field signature). Pooled regression of a(s,E) on d_epi(s,E) with per-event centring (each event's station-mean removed). A source/nucleation signal predicts a NEGATIVE slope: nearer stations have a larger -10h anomaly. Significance vs a season-matched CIRCULAR TIME-SHUFFLE null of event epochs (1000 draws; preserves each station's microseism seasonal/diurnal climate and the catalog's own recurrence). One-sided.\n\nPRIMARY TEST 2 (ambient/ocean signature). Cross-station correlation of a(.,E) over events per station pair, regressed on INTER-STATION distance. An ocean-storm-driven field predicts correlation DECREASING with inter-station distance AND independence from epicentral geometry. Same shuffle null.\n\nINTERPRETATION MATRIX (locked). T1 significantly negative -> SOURCE / near-field (nucleation or path): the precursor is epicentre-organized (strengthens the operational product; weight near-epicentre stations). T1 null + T2 significant -> AMBIENT / ocean-driven (a shared-noise field; deflates the causal precursor -- a publishable partial-retraction). Both null -> STATION-LOCAL / artifact (guts the precursor: the stacked signal is independent station noise averaged). Both significant -> MIXED (a source signal on an ocean field); report both. MAGNITUDE COROLLARY: a genuine source signal should make Test 1's negative slope STRONGER for the M7+ cohort than the M5.5-6.0 cohort (the signal scales with magnitude, as the audited effect does).\n\nPASS / primary outcome = Test 1. CONFIRM if the pooled epicentral-distance slope is significantly negative vs the shuffle null (one-sided p<0.05). Otherwise the interpretation matrix assigns AMBIENT / STATION-LOCAL / MIXED as the honest mechanism reading.\n\nWITHDRAW GATES (coverage, decided before any outcome). (a) >=30 IU/II stations clear the >=70% coverage rule; else the pool is too sparse -> widen the coverage window or the network set and RE-REGISTER (do not proceed). (b) >=100 declustered M>=5.5 events, each with >=10 covered stations spanning >=5000 km of epicentral-distance range within the pool; else insufficient distance leverage. (c) M7+ cohort >=30 covered events; else the magnitude corollary is reported DESCRIPTIVELY only.\n\nRUNNER. The GSN microseism backfill (idle compute) and the analysis run BLIND after this lock is Bitcoin-anchored. No parameter is tuned after unblinding. A NULL, a deflation, or an ambient-artifact verdict is a valid and publishable outcome. This lock precedes the global data, so the expanded-pool analysis is a genuinely blind test.","ots_proof_path":null,"outcome":{"lookahead_hours":-10.0,"min_magnitude":5.5,"type":"earthquake"},"reevaluate_at":"2026-10-24","registered_at":"2026-07-24T19:35:13Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"91fc6a7840a514aba8337e181af76b3c020e6cc9","interpretation":"disconfirming","metric_value":0.0259,"n_observed":null,"notes":"DISCONFIRMED (mechanism inference VOID) by independent verification wave (trackAR/trackAN, 2026-08-24). EXPLAINED OVERRIDE: this seal tests for an epicentre-organized source via a distance-slope p, where p<0.05 was meant to CONFIRM the effect. The pre-event (k=-1, holds -10h) near-field slope IS still significant AFTER GK74 mainshock-only declustering -- contrast/p_gt: 0.124/0.0259 (<500 km), 0.104/0.0040 (<1000 km), 0.053/0.0100 (<2000 km) -- so the metric mechanically \"passes\" its 0.05 threshold. It is recorded FAILED anyway because that surviving slope is EARTHQUAKE SELF-CONTAMINATION (small foreshock/sequence seismicity near the epicentre, recorded by a BROADBAND whole-day metric microseism_secondary_band_3s = MUSTANG pct_above_nhnm), NOT an epicentre-organized ocean-microseism source. Decisive: on the un-biased band_rms metric (0.14-0.5 Hz hourly, clean baseline, declustered) the -10h z is ~0 with NO magnitude scaling -- excluded (n=82 M7.5+ would give t~13; measured -0.09; positive control event-hour z=+30.3), so there is no microseism signal for a source to organize. The raw (non-declustered) gradient is far larger (contrast 2.47, p 0.002 at 500 km) and is Omori aftershock contamination. metric_value stamped = 0.0259 (the real declustered k=-1 <500 km slope p). Seal + hash chain + OTS anchor preserved as the record of what was claimed.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-26T18:41:25Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"per_station_anomaly_epicentral_and_interstation_regression_vs_circular_shuffle_null","n_permutations":1000},"threshold":{"direction":"negative","metric":"epicentral_distance_slope_one_sided_p","value":0.05},"title":"Microseism precursor spatial mechanism: epicentre-organized source vs ambient ocean field vs station-local artifact","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"","expected_direction":"negative","feature_set":[],"git_commit_at_registration":"102843d2923011f12fb2c6d4405196bebb6273f2","hypothesis_id":"20260725_yellowstone_abiogenic_hc_negativ_bae9aa51","locked_at":"2026-07-25T16:57:44Z","locked_file_sha256":"841622a450c2456800a19c6390bfa0393b66d87fd1dcf5cbe103edb50a3536b5","notes":"NEGATIVE CONTROL for abiogenic_hc_discovery_monitor at Yellowstone. Confirmatory baseline (ran 2026-07-25): known-thermogenic Calcite Springs oil seep (biomarker-bearing) correctly NOT flagged -- composite=19.7 class=0 p_abi=0.05 R/Ra=0.02. ARTIFACT CAUGHT: extreme mantle-He caldera center (16 R/Ra) over-flagged composite=52.8 class=2 p_abi=0.36, because the thermal axis has no UPPER cutoff for hydrocarbon destruction (graphitization >~250C -> CO2/graphite, per USGS YVO). Positive refs correct: Lost City 68.9/class3, Bourakebougou 35.1/class1. FORWARD TEST: after adding an upper-T (graphitization) disqualifier so abiogenic-HC potential DROPS above the HC-destruction window, PASS = caldera-center class<=1 AND Calcite Springs stays class 0 AND Lost City stays class>=3 (corrects the false-positive without breaking true-abiogenic detection). MT (independent): Calcite Springs 26 ohm-m = 4.5x more resistive than Norris brine 5.9 = HC-consistent.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"hydrocarbon"},"reevaluate_at":"2026-08-20","registered_at":"2026-07-25T16:57:04Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"d3a5fc3c6a8d921c6bf1e9f33c02bcfd389b964f","interpretation":"disconfirming","metric_value":2.0,"n_observed":null,"notes":"FORWARD TEST FAILED (honest disconfirmation). Upper-T graphitization cutoff implemented (commit d3a5fc3c) + regression-safe (Lost City 68.9/class3, Bourakebougou 35.1/1, Calcite Springs 0 -- unchanged). BUT caldera center UNCHANGED at 52.8/class 2, so PASS criterion (class<=1) NOT met. CORRECTED DIAGNOSIS: the thermal axis was already 0 at the caldera center (reservoir_t=97.9C -- geothermal_gradient_trainer UNDERESTIMATES the magma-chamber gradient, so the upper-T cutoff never engages). Real drivers of the 52.8: he3_mantle=100 x0.30=30 (CORRECT, real 16 R/Ra mantle He) + continuous_flow=100 x0.15=15 (the (44.43,-110.59,Yellowstone_active) SELF-ANCHOR in _CONTINUOUS_FLOW_ANCHORS). LESSON: the score is a GENERATION-favorability index; a volcanic degassing center IS abiogenic-generation-favorable (real mantle input), so class 2 is not wrong for GENERATION -- only for recoverable RESOURCE (no preservation/trap). Generation-favorability and resource-prospectivity DIVERGE at a magmatic center. Real fixes: (a) gradient trainer must see the magma so the preservation gate engages; (b) reconsider the Yellowstone_active continuous-flow self-anchor; (c) add an explicit preservation/trap gate to separate generation from resource.","ots_proof_path":null,"passed":false,"tested_at":"2026-07-25T17:07:52Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"abiogenic_negative_control","n_permutations":null},"threshold":{"direction":"negative","metric":"caldera_center_composite_class","value":1.0},"title":"Yellowstone abiogenic-HC negative control: upper-T fix corrects caldera-center false positive","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"PhysioNet BIDMC PPG+Respiration 1.0.0 (53 records, 8 min, 125 Hz; PPG payload, impedance-respiration ground truth); downloaded ONLY after seal+lock+anchor","expected_direction":"positive","feature_set":["bidmc_ppg_cardiac_phase_demod_channels"],"git_commit_at_registration":"5d5fdb43d15f840c63d4d1578c35845248c662a4","hypothesis_id":"20260807_cosmos_biomedical_transfer_cardi_7b8ae746","locked_at":"2026-08-07T03:17:38Z","locked_file_sha256":"d37523e9672ca88d7f6ad9ef016895541c2019739d8e9f0e7d957b722e95d75c","notes":"SEALED PRE-REGISTRATION OF RECORD: strategy/cosmos_biomedical_prereg_v1_SEALED.md, SHA-256 b692273bb70eba52bc83e2d90278efe4aed625c211a1d215dc4023a5a20af6fb (sealed BEFORE any BIDMC byte was downloaded). Working copy _active_today/CoSMoS_Biomedical_Prereg_v1.md (identical bytes). Implementation of record scripts/cosmos_biomedical_bidmc.py SHA-256 219bb6d80baf6e4b2b2b2adadd2b602b15236210a8d7c31c9e53d2c666fe968c. SCOPE DISCIPLINE: exactly ONE cheap pre-registered experiment, NOT a pivot (Grok verdict reports/cosmos_biomedical_grok.md; Mike approved 2026-08-07); no clinical-utility claim. FROZEN BARS -- PRIMARY PASS requires ALL: (1) Wilcoxon signed-rank one-sided p < 0.05 on per-record improvement MAE_FFT - MAE_CoSMoS; (2) median improvement > 0 brpm; (3) split-halves: median improvement > 0 in BOTH record halves. Otherwise NULL (publishable). NULL VALIDITY: random-permutation phase scramble (NO circular shift -- Aug-6 near-periodic lesson), 5 draws/record seed 20260806; median degradation > 2.0 brpm AND Wilcoxon p < 0.05, else verdict INVALID. SECONDARY (reported, does not gate H1): injected mains-like additive tone at fc+0.55 Hz, 30 pct cardiac-band rms; SECONDARY PASS iff median CoSMoS MAE inflation < 1.0 brpm AND median FFT inflation > CoSMoS inflation. Sideband amplitude SYMMETRY reported per record (it does the discriminating); demod-collapse gain is AM-vs-PM only, REPORTED never gated (measured CFC-veto limits 2026-08-06). Inclusion: >= 4 valid reference windows per half; n >= 30 records else INVALID. Phase 0 synthetic gate PASSED pre-seal: recovery median MAE 0.069 brpm; null moves statistic +8.96 brpm; artifact inflation 0.00 vs FFT +2.18 (reports/cosmos_biomedical_phase0.json). Kill criteria binding: if indistinguishable from standard practice, CoSMoS adds only nomenclature here -- say so and stop. null > false data.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"methods_benchmark"},"reevaluate_at":"2026-11-06","registered_at":"2026-08-07T03:17:33Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"f1cf1586c04c84a8ea4988c61d63414676510bfd","interpretation":"confirming","metric_value":1.974242649002974e-8,"n_observed":53,"notes":"PASS on all three frozen primary bars. n=53/53 records analyzed, 0 excluded. PRIMARY: Wilcoxon signed-rank one-sided p = 1.97e-08 (bar < 0.05); median per-record MAE improvement (FFT - CoSMoS) = +3.676 brpm (bar > 0); split-halves BOTH positive (h1 +4.104, h2 +2.875 brpm). Median MAE: CoSMoS 6.618 vs fixed-FFT 11.074 brpm. NULL VALIDITY MET BUT NARROWLY: random-permutation phase-scramble degradation median +2.041 brpm (bar > 2.0 -- a 0.041 brpm margin), Wilcoxon p = 0.0024 (bar < 0.05); the null does move the statistic but the margin is thin and is reported as the main fragility. SECONDARY artifact rejection PASS: injected fc+0.55 Hz tone inflates CoSMoS MAE by only +0.257 brpm vs fixed-FFT +2.069 brpm. REPORTED NOT GATED: median sideband symmetry 0.866 clean / 0.839 artifact; median AM-vs-PM SNR ratio 0.434 (collapse gain is an AM-vs-PM discriminator only, per the measured 2026-08-06 CFC-veto limits). HONEST CEILING: absolute skill is POOR in both arms (CoSMoS median 6.6 brpm error vs a ~8.2 brpm scrambled-null floor) -- this is a relative win over a deliberately naive fixed-frequency sideband stack, NOT a clinically useful respiratory-rate estimator and NOT a claim of clinical utility. The pre-registered question (does the CoSMoS machinery transfer to a wandering biological carrier and buy artifact rejection) answers YES; the Grok kill-criterion caveat stands: instantaneous-phase estimation may be doing most of the work, and no comparison to standard PPG-derived-respiration practice was pre-registered or run. Result JSON sha256 a48f73044f36835c29ab40e045f19cb5e3af6c2182452d6a39712f1ddfed33e2, OTS-anchored on AX41 (4 calendars). Runner sha256 219bb6d8... verified byte-identical to the sealed implementation on AX41 before the run; prereg sha256 b692273b... likewise. Data downloaded ONLY after the seal commit 5abb7dfa was pushed. ONE-AND-DONE per the sealed scope-discipline clause: no follow-on without a fresh explicit decision.","ots_proof_path":null,"passed":true,"tested_at":"2026-08-07T03:43:54Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"paired_per_record_MAE_wilcoxon_with_phase_scramble_null_and_artifact_secondary","n_permutations":5},"threshold":{"direction":"positive","metric":"wilcoxon_p_one_sided_improvement","value":0.05},"title":"CoSMoS biomedical transfer: cardiac-phase demodulation vs fixed-FFT sideband stack on BIDMC PPG+Respiration","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"INTERMAGNET raw 1-min B-field, 11-station curated subset, dm_raw_bfield.db (springboard)","expected_direction":"positive","feature_set":["E_amp_sid_nT","z_sid","p_empirical_sid"],"git_commit_at_registration":"395e7622b44da31252adfe78a811248f4d30a389","hypothesis_id":"20260807_dm_magnetometer_channel_v1_sider_296ee71c","locked_at":"2026-08-07T02:45:39Z","locked_file_sha256":"5090ff310f2d1c1dfc3ea0292491e881c6c364a95ebefb583e4c624f2e57e424","notes":"Magnetometer sibling of the DM gravimeter-channel v2 certified null. A NULL is a publishable outcome (upper limit on quasi-static inertial-frame equatorial B amplitude). Any bar-passing line is a CANDIDATE pending separately pre-registered K1 adjudication, never a detection claim.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_line_search"},"reevaluate_at":null,"registered_at":"2026-08-07T02:45:33Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"7fd445546e00a53dae634b03ebcf7905ffd43693","interpretation":"confirming","metric_value":147.14556003587398,"n_observed":11,"notes":"CAPPED OUTCOME -- SIDEREAL_CANDIDATE_PENDING_K1_ADJUDICATION. This is NOT a detection, and per the sealed prereg's K1 clause it CANNOT become one from this analysis alone. Frozen runner scripts/dm_mag_channel_v1.py (sha256 7db981e1...) executed on springboard (CPX32) 2026-08-07T03:13Z under sealed prereg strategy/dm_mag_channel_prereg_v1_SEALED.md (sha256 3fcad0c7...). 11 INTERMAGNET observatories (BOU CMO CNB HAD HER HUA KAK KOU NGK SJG TAM). All 4 pre-registered bars PASSED: (1) z_sid = 147.15 vs dummy grid (bar >= 5); (2) empirical p = 0.005 dummy-grid (station-perm p 0.002, block-scramble p 0.005; circular-shift surrogates deliberately EXCLUDED per the measured 2026-08-06 near-periodic lesson); (3) split-halves 4/4 pass (stations even/odd, time first/second half, amplitude ratio + phase agreement); (4) injection gate pass. Sidereal-line equatorial amplitude 1.711 nT, phase -74.3 deg; 95 pct upper limit 1.711 nT. WHY CAPPED AT CANDIDATE: the tidal/ionospheric K1 line sits EXACTLY at the sidereal frequency (K1 argument = sidereal time; expectation ~5.32 nT here) and the run's frequency resolution cannot split them -- the observed 1.711 nT is fully consistent with ionospheric Sq/K1 leakage. The prereg pre-committed that ANY bar-passing sidereal line is a CANDIDATE pending a SEPARATELY pre-registered K1 adjudication, never a detection claim. REQUIRED NEXT STEP: a K1-adjudication pre-registration (being drafted separately for Mike's review) must be sealed and anchored BEFORE any further analysis of this candidate; no follow-on analysis until then. Result JSON sha256 b6b768839cdc4f4b2d78a2f2d1b7e14cb165835aa73ccfb91be6a87c093176ed, repo copy reports/dm_mag_channel_v1_RESULTS.json, copied springboard -> AX41, OTS-anchored on AX41. Registry lock committed a476dd2f. Sibling program records: DM gravimeter-channel v2 certified null (A_UL 2.96e-13 @ 5.82 mHz) + DM oscillating-G v2 NO_CANDIDATE_LIMITS (same-day). Magnetometer sibling of the gravimeter channel; a downgrade to NULL at K1 adjudication would still yield the publishable upper limit.","ots_proof_path":null,"passed":true,"tested_at":"2026-08-07T13:48:28Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"bars":{"block_null_p":0.05,"p_dummy":0.05,"perm_null_p":0.05,"split_phase_deg":45,"split_ratio":[0.5,2.0],"z":5.0},"code":"scripts/dm_mag_channel_v1.py","code_sha256":"7db981e1ab3e1ff6ef044f17ad023471a81e66bdaa82e369cdfc62de9b7f1a40","n_dummy":200,"no_circular_shift_surrogates":true,"null_battery":["dummy_frequency_grid","station_label_permutation","30day_block_phase_scramble"],"phase0":"PASS 9/9 pre-seal (synthetic only)","prereg_doc":"strategy/dm_mag_channel_prereg_v1_SEALED.md","prereg_sha256":"3fcad0c7a51bf8bc43eea31d0de61c7c71191b55c798488d2e44a6bde5290cfa","scope":"11-station curated subset, NOT the 154-obs network","seed":20260806},"method":"geometric_coherent_stack_dummy_frequency_empirical_null","n_permutations":500},"threshold":{"direction":"positive","metric":"robust_z_vs_dummy_grid","value":5.0},"title":"DM magnetometer-channel v1 sidereal-line coherent-stack search (CoSMoS)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"gsn_microseism_hourly.db (2010-2026, v1 archive) + gsn_microseism_hourly_1995_2009.db (C3 back-extension, launched 2026-08-07, merged before analysis); USGS M>=5.5 GK74-declustered","expected_direction":"negative","feature_set":["microseism_band_rms_secondary"],"git_commit_at_registration":"28284ccf9bccbe2b5712654f8a7a5f9242d8fd4d","hypothesis_id":"20260807_microseism_precursor_spatial_mec_eb705438","locked_at":"2026-08-07T00:15:41Z","locked_file_sha256":"42f7d708de0cd5da2ed035150c99de5c056ff177460d378566d94664ea672145","notes":"SEALED PRE-REGISTRATION OF RECORD: strategy/microseism_spatial_mechanism_prereg_v2_SEALED.md, SHA-256 baf4d61b9837b347a3a280325d56d88bd0e975ad12d980c0212b54824f982e6f (sealed 2026-08-07). This manifest is the registry pointer; the sealed document is the binding specification (frozen data, gates, analysis plan, combiners, season-matched null, pass criteria B1-B5, confound gates, forbidden combination). DECISIONS (Mike, 2026-08-07): (a) PRIMARY combiner = WEIGHTED SIGN TEST per the pre-committed Section 6.3 rule (meta nominally better by exactly one grid step, inside the 200-realisation binomial tie band -> tie -> sign test; the rule, not post-hoc preference, made the choice); (b) confirmatory set = C3 PRIMARY: pre-2010 declustered M>=5.5 events from the 1995-2009 back-extended GSN IU/II archive, fully disjoint from the v1 archive; C1 (M5.0-5.5) secondary, C2 (post-2026-04-16) supplementary; (c) C3 backfill funded + launched 2026-08-07 on AX42-U (separate output DB gsn_microseism_hourly_1995_2009.db + own state DB; sqlite single-writer discipline; the running 2010-2026 job untouched). HARD PREREQUISITE (unchanged at seal): B4 cross-event-correlation correction requires the 30-day BLOCK-YEAR null calibrated from >=1000 draws BEFORE any confirmatory run; until then the confirmatory run does not proceed. The v1 event set (2010-01-01..2026-04-16 declustered M>=5.5) is BARRED from the confirmatory statistic; the v1 verdict (20260724_microseism_precursor_spatial_mec_ea481761, FAIL/NULL p=0.0731) is FINAL and is not revised. Power study: reports/microseism_v2_power_study.md (MDE at 80% eff: weighted sign 1.6e-07 z/km mag-scaled / 7.5e-07 flat). N_DRAWS=2000 frozen; every null draw archived or the run is invalid.","ots_proof_path":null,"outcome":{"lookahead_hours":-10.0,"min_magnitude":5.5,"type":"earthquake"},"reevaluate_at":"2026-10-24","registered_at":"2026-08-07T00:15:35Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"91fc6a7840a514aba8337e181af76b3c020e6cc9","interpretation":"disconfirming","metric_value":0.0259,"n_observed":null,"notes":"DISCONFIRMED (mechanism inference VOID) by independent verification wave (trackAR/trackAN, 2026-08-24). Spatial-mechanism v2 (weighted sign test, 1995-2009 back-extended archive) rests on the same retracted feature microseism_secondary_band_3s (= MUSTANG pct_above_nhnm), a BROADBAND whole-day daily-cadence metric inflated by each event own energy. PROVENANCE NOTE: this seal declared metric (weighted_sign_test_one_sided_p on the 1995-2009 archive) was NOT independently recomputed; the metric_value stamped (0.0259) is the trackAN 2005-2018 declustered epicentral-distance slope p (<500 km, k=-1) used as the nearest measured SPATIAL evidence. EXPLAINED OVERRIDE: like seal 20260724, the pre-event near-field slope stays significant even after GK74 declustering (p 0.004-0.026), but that significance is EARTHQUAKE SELF-CONTAMINATION (foreshock/sequence seismicity), NOT an epicentre-organized microseism source. Decisive disconfirmation: on the un-biased band_rms metric (0.14-0.5 Hz hourly, clean baseline, declustered) the -10h z is ~0 with no magnitude scaling -- excluded (positive control event-hour z=+30.3). Seal + hash chain + OTS anchor preserved.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-26T18:41:25Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"per_event_ols_slope_weighted_sign_test_vs_per_event_season_matched_whole_year_null","n_permutations":2000},"threshold":{"direction":"negative","metric":"weighted_sign_test_one_sided_p","value":0.05},"title":"Microseism precursor spatial mechanism v2: weighted sign test on 1995-2009 back-extended archive (C3 primary)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"INTERMAGNET BGS GIN raw 1-min definitive B-field (frozen station set, 80-pct-coverage rule, 2006-2018) + IGS final IONEX global TEC from CDDIS (2005-2018, native cadence); triggers: frozen 71-event GOES X-class CSV (NCEI XRS + DONKI) + frozen 259-trigger HEASARC fermigtrig SGR CSV","expected_direction":"positive","feature_set":["intermagnet_sfe_dhdt_z","ionex_dayside_vtec_z"],"git_commit_at_registration":"eb9be802550a25af8d99a2b16103d9406a90ab3b","hypothesis_id":"20260807_multi_messenger_arm_a_triggered__18d235db","locked_at":"2026-08-07T03:47:47Z","locked_file_sha256":"9d37d9a431a89e28ddf5c0d897c56c9449ecf0eb5aff920e39352d82fc5bf3bd","notes":"SEALED PRE-REGISTRATION OF RECORD: strategy/mm_arm_a_validation_prereg_v1_1_SEALED.md, self-hash SHA-256 4e143b368098ce1aca40b667e36037887f413de19f9d1dfa9ebed78d02bf6778 (self-hash convention: hash field replaced by literal SELFHASH before hashing). PROVENANCE: v1 DRAFT sha256 ade3482c96a90071c6dfc67462ea0fe43d71648a30352cd86791aa6167e2bb7c (commit 8237f1d0, drafted + hashed 2026-08-06; Mike pre-approved the anchor on landing -- standing order). v1.1 = that draft + the pinned SGR list + seal-status header/footer; NO frozen content changed. FROZEN EVENT LISTS: strategy/mm_arm_a_event_list_v1_DRAFT.csv sha256 92450a459cdd40c48cb87d5ebeed26645db830b766d38d5f5b45acef971333ce (71 GOES X-class flares 2005-2018); strategy/mm_arm_a_sgr_list_v1.csv sha256 57cc16e723ee9487334b90c7d463820728e942432441cc3ff259f81347b2ae5f (N=259 Fermi GBM SGR triggers 2008-07-12..2018-12-31; frozen query executed verbatim: catalog HEASARC fermigtrig, classification=SGR, window as above, no other filters, sorted by trigger time). OBJECT UNDER TEST: the triggered-stack pipeline itself (epoch-fold + event-time-scrambled null + split-halves), NOT new physics -- a failure is a pipeline verdict. PRIMARY CHANNELS: CH-TEC SID (IGS final IONEX dayside mean VTEC, N=71 events) + CH-GEOMAG SFE magnetic crochet (INTERMAGNET raw 1-min dayside abs-dH/dt, N=53 in-window events); 24-h de-cluster rule applied mechanically. FROZEN BARS: p_emp < 1e-3 with S > 0 in EACH primary channel, PLUS all 4 split-halves (events odd/even by chronological rank; CH-GEOMAG stations odd/even by longitude rank; CH-TEC dayside N/S hemisphere) at p_emp < 0.05 with S > 0. Physics-consistency secondary (Spearman per-event response vs log10 GOES class positive at p < 0.01 in at least one primary channel) required for the clean-PASS label only, not the pipeline license. NULL: matched pseudo-event scrambled-time battery, M=10,000 draws, each pseudo-event matched on UT hour (+/- 1 h) + calendar month (+/- 1) in a different year, >= 3 d from all real and pseudo events -- NOT circular-shift (invalid on diurnal series, cosmos_veto_limits item 5). SCHUMANN EXCLUDED from all criteria at N < 30 (live archive N=3, sr_hourly N=5, WSPR N=2; data-driven honesty floor) -- conditional non-gating annex only; re-enters solely via a re-registered addendum after an institutional backfill covers >= 30 in-window events. SECONDARY STACK (own bar, exploratory): SGR short bursts, bar p_emp < 0.01 in either channel with S > 0; a null is an honest stacked upper limit, NOT a pipeline failure; bright-subset row = SGR J1550-5418 Jan-Feb 2009 storm. ANNEX: SGR 1806-20 giant flare 2004-12-27 21:30 UT single-event positive control (non-gating); cddis_tec.db coarse-archive cross-check row. VERDICT SEMANTICS: any primary-criteria failure = pipeline NOT VALIDATED and ALL Arm B/C stacked-limit publications are EMBARGOED until a revised pipeline passes a re-registered v2; no partial-credit path. Analysis seed = the OTS block hash of this anchor. FROZEN ORDER: CH-GEOMAG station set enumerated + logged before any response fetch; the Arm-A run awaits the native-cadence IONEX TEC refetch the prereg specifies. Verdict goes on the ledger either way. null > false data.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"pipeline_validation"},"reevaluate_at":"2026-11-06","registered_at":"2026-08-07T03:47:44Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"885795b4e2742fa61da920a8618f105eb0c7a766","interpretation":"disconfirming","metric_value":0.0357964,"n_observed":101,"notes":"NOT VALIDATED (NOT VALIDATED). CH-GEOMAG S=+1.4955 p=9.999e-05; CH-TEC S=+13.2246 p=0.0358; split-halves FAIL. Seed=BTC block 961374. Full result: reports/science/mm_arm_a_validation_result.md","ots_proof_path":null,"passed":false,"tested_at":"2026-08-07T18:46:26Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"epoch_fold_triggered_stack_vs_matched_pseudo_event_scrambled_time_null","n_permutations":10000},"threshold":{"direction":"positive","metric":"p_emp_per_primary_channel","value":0.001},"title":"Multi-messenger Arm A: triggered-stack pipeline validation on solar-flare SID/SFE (CoSMoS)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"IGETS Level-2 CORMIN 1-min SG data, 21 stations >= 5 yr (frozen cut), house Slichter-validated residualization (drift + 15 tidal constituents + barometric admittance) declared in the addendum; achieved band 1e-6..8.333e-3 Hz (1-min Nyquist)","expected_direction":"positive","feature_set":["igets_sg_l2_cormin_residual_spectra"],"git_commit_at_registration":"59780ea08c7c276965bb6934b668df4adc38fcf0","hypothesis_id":"20260807_ultralight_scalar_dm_oscillating_e252f6bd","locked_at":"2026-08-07T02:39:07Z","locked_file_sha256":"972dcbd603607cfa5d96a6627e5365451ad4c5bac4be89cc09f2d1f65b4ee367","notes":"SEALED PRE-REGISTRATION OF RECORD v2: strategy/dm_oscillating_g_prereg_v2_SEALED.md SHA-256 9feedf7f53d673602a2d884446bb05fb3e25cf46f161b816ee2494e62ab13bea + METHODS ADDENDUM strategy/dm_oscillating_g_methods_addendum_v1_SEALED.md SHA-256 bcbe90de33b43d33076c9a4a00859073163b151d1229d2dab07b870f242dbad6, both OTS-anchored on AX41 2026-08-06 BEFORE any Phase-1 unblinding (receipts committed beside the docs; hashes re-verified against the AX41 originals and byte-verified after commit). SUPERSEDES v1 (6affe3071d05e9bad7e0fe06fd3f2c24ef0d551d17a9a61b7eae0470c313ab86, locked as 20260807_ultralight_scalar_dm_oscillating_4d41c614, commit 59780ea0) -- v1 artifacts preserved untouched, its lock manifest is OTS-anchored and must not be edited; this note is the supersession record. FROZEN CRITERIA (all four): p_LEE<1e-3 (>=1e4 draws); coherence z_C>5.0 vs shuffled-inter-station-phase null; annual sideband amplitude [1%,10%] of halo prediction, phase +/-45deg of June maximum, p_annual<0.05; demod-collapse gain >=1.3x. ADDENDUM DECLARES: L2 CORMIN residualization is ours (house cleaner), achieved band 1e-6..8.333e-3 Hz, TIDE-ADJACENT veto (any candidate within 3 Rayleigh widths of a fitted tidal line or its +/-1/yr sidebands is ineligible, per memory/cosmos_veto_limits.md), 4 red-team implementation corrections, sealed stats in ONE shared module cosmos_dm_stats.py imported by both Phase-0 validator and Phase-1 runner. PHASE-0 v1 (phase0_dm.py, 12 sta / 6 yr hourly, seed 1feedf7f): VERDICT PASS 4/4 -- B null calib realized FA 1.2e-3 vs target 1e-3 (thr 17.0) PASS; A recovery eff 92.5% at SNR8 (emp SNR 7.94, need >=90%) PASS; C fingerprint sideband z>=3 fires DM 97.5% (med z 5.64) vs monochromatic 0% PASS; D annual-phase within 45deg DM 100% vs null 22.5% PASS. HONEST GATE STATUS: the addendum (Sec 4) requires PHASE-0 v2 PASS on the shared module before Phase 1 unblinds; Phase-0 v2 (phase0_dm_v2.py, 21 sta / 8 yr @ 60s) currently returns VERDICT REVIEW (B PASS FA 1.5e-3; A 85%<90% FAIL; C collapse 0% fire FAIL; D annual gate 10% FAIL; E p_annual median 0.85 FAIL; coherence z reachability CONFIRMED z_perfect 8.09 > 5.0) -- Phase 1 is BLOCKED until the pipeline is fixed openly and a Phase-0 v2 PASS report is re-anchored, per the prereg's own frozen Phase-0 clause. Additional Phase-1 preconditions: Bitcoin confirmation of the parent anchor (null seed = confirming block hash; receipts still pending attestation) and a corrected Phase-1 runner importing cosmos_dm_stats.py (the archived phase1_dm_runner.py is the pre-review v1, preserved for the record in reports/dm_oscillating_g_phase0/). Data READY: 21 IGETS stations >= 60 monthly CORMIN files on AX41. Run ONCE when gates clear; limits curve (default) or sealed candidate goes to the ledger either way. null > false data.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geophysical_measurement"},"reevaluate_at":"2026-11-06","registered_at":"2026-08-07T02:39:02Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"7fd445546e00a53dae634b03ebcf7905ffd43693","interpretation":"disconfirming","metric_value":0.0,"n_observed":50,"notes":"CLEAN NULL -- VERDICT NO_CANDIDATE_LIMITS. Sealed Phase-1 runner scripts/nobel_shots/dm_oscillating_g_v2_run.py (commit 276f91e3, module sha256 4098abad7d424eba...) executed ONCE on AX41 2026-08-07T03:19Z, unblinding seed = BTC block 961333 hash (data/dm_v2_seed.json). 21 IGETS SG stations, band 1e-6..8.333e-3 Hz, 4,556,034 bins, 10,000 permutations, LEE amplitude threshold S >= 4494.87. Of the top 50 lines evaluated: 3 vetoed tide-adjacent (per the sealed addendum's 3-Rayleigh-width tidal veto), 0 passed all four frozen criteria (c1 LEE + c2 coherence z_C>5 + c3 annual-sideband fingerprint + c4 demod-collapse >=1.3x). Several lines cleared amplitude+coherence alone but FAILED the velocity-modulation fingerprint -- exactly the monochromatic-systematic signature the fingerprint battery exists to reject. metric_value records n_pass_all_criteria = 0 (the runner applies the frozen LEE bar as an amplitude threshold; no per-candidate p_LEE scalar is emitted). Default pre-registered outcome delivered: per-frequency upper-limit curve on the DM-matter coupling (limits_curve in the result JSON) -- LIMITS-PAPER INPUT, sibling of the DM gravimeter-channel v2 certified null (strain A_UL 2.96e-13 @ 5.82 mHz). Result JSON sha256 2b088575f5766ad177b1e3b3213384efe6e01dc9868fe3e7155d43bd6f7ef6e4 (verified against the AX41 .sha256 sidecar), repo copy reports/dm_oscillating_g_v2_phase1_result.json, OTS-anchored on AX41. Run log ~/prereg/phase1_v2_run.log on AX41. This v2 result SUPERSEDES v1 (prereg 6affe307..., manifest 20260807_ultralight_scalar_dm_oscillating_4d41c614, preserved untouched per its supersession record). null > false data.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-07T13:48:08Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"coherent_network_stack_LEE_band_scan + shuffled-inter-station-phase coherence z + annual sideband amp/phase permutation + phase-aligned complex demod-collapse (cosmos_dm_stats.py shared module)","n_permutations":10000},"threshold":{"direction":"positive","metric":"global_line_p_LEE","value":0.001},"title":"Ultralight scalar-DM oscillating-coupling search v2 + methods addendum: coherent SG-network stack, corrected sealed-criteria implementation (HYP-9feedf7f; SUPERSEDES v1 6affe307)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"ComCat via fdsn_multi, HOLDOUT 1964-1989 (WWSSN era; 1990-2026 already examined and exploratory)","expected_direction":"positive","feature_set":["antipode_distance_km","trigger_mag","follower_mag","lag_days"],"git_commit_at_registration":"40307398966dae6537b3280eeed9b6a227449412","hypothesis_id":"20260815_antipodal_triggering_at_pkikp_fo_9982109b","locked_at":"2026-08-15T21:30:21Z","locked_file_sha256":"b77be0159c7cad99e59085a9cb678aa1f18d919f6dbf655c27fba0c15a27fdaa","notes":"EXPLORATORY ORIGIN, stated in the lock: 12-cell grid on 1990-2026, best cell 200km/30d lift 4.23 p=0.0125 uncorrected on FOUR events; Bonferroni ~0.15. Registered for the SHAPE (lift falls monotonically with radius, 14/15 pairs) which is what PKIKP focusing predicts. POWER COMPUTED BEFORE LOCKING: ~0.60 on the 26-year holdout, so a non-significant result is NOT a refutation and will be reported as underpowered. Wide-radius result was a clean null (lift 0.99, p=0.60); 16.6pct of M7+ have an antipodal follower by geography alone. Prior expectation NOT REPLICATED. Full prereg hypotheses/antipodal_focusing_prereg_v1.md","ots_proof_path":null,"outcome":{"lookahead_hours":720.0,"min_magnitude":6.5,"type":"earthquake"},"reevaluate_at":"2026-08-16","registered_at":"2026-08-15T21:30:00Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"ef9f5169c6de5cf6cc7998ec296c6b517958384b","interpretation":"disconfirming","metric_value":1.0,"n_observed":0,"notes":"UNRESOLVED per the lock's own criterion, and the reason matters. Observed k=0 in 1964-1989 (995 M6.5+ events, 307 M7.0+ triggers; declustered 893/290), against a time-shuffle null of mean 0.119 over 2000 draws. p=1.0. Declustered and undeclustered agree exactly, so the lock's unspecified declustering parameters did not decide anything. The lock names common.null_quality.is_alive as an UNRESOLVED trigger and it fired (3 distinct values), but the diagnostic shows the null DID have resolution: k>=2 would have given p<0.05. The real problem is POWER: my pre-lock estimate assumed a null rate of 0.68 extrapolated from 1990-2026, and the actual 1964-1989 rate is 0.119 -- a factor of 5.7 lower -- so true power was ~9%, not the ~60% stated in the lock. The null rate did not transfer across epochs and I did not check that it would. Verdict stands as UNRESOLVED: this holdout could not have detected the effect it was built to test.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-15T22:02:20Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"count M7.0+ triggers followed by a declustered M6.5+ within 200km of the antipode in 30d; null = time-shuffle preserving every epicentre and permuting times, 2000 draws","n_permutations":2000},"threshold":{"direction":"positive","metric":"count_and_permutation_p","value":0.05},"title":"Antipodal triggering at PKIKP focusing scale: M6.5+ within 200km of a M7.0+ antipode in 30d","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"NASA OMNI2 hourly 1996-2025 + ComCat via fdsn_multi (global, M6.5+, declustered 500km/30d)","expected_direction":"positive","feature_set":["omni_vsw","omni_np","hss_onset","declustered_m65_count"],"git_commit_at_registration":"cdbd38b054aab9570192fee4ab402ea29bb2fb5a","hypothesis_id":"20260815_coronal_hole_high_speed_streams__d599a50d","locked_at":"2026-08-15T20:22:35Z","locked_file_sha256":"06cafff0efa2d6b2e2cef171e54d7fa34190304747f94666b755b3ec42bc3e03","notes":"Tests the PHYSICAL claim, not Burns's forecasting skill (no alert scrape). Prior expectation NULL. ONE primary analysis; 24h/72h, M6.0/M7.0, speed-only and cycle-phase splits are declared EXPLORATORY in advance and cannot promote a fail. UNRESOLVED conditions fixed in advance: <30 HSS onsets, dead null, >20pct OMNI gaps, <100 declustered events. Full prereg hypotheses/t1_hss_earthquake_prereg_v1.md","ots_proof_path":null,"outcome":{"lookahead_hours":48.0,"min_magnitude":6.5,"type":"earthquake"},"reevaluate_at":"2026-08-16","registered_at":"2026-08-15T20:22:25Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"95a278a440894c183f294129e95cac8a691c64e5","interpretation":"disconfirming","metric_value":0.811,"n_observed":6,"notes":"NULL. 31 HSS onsets (Vsw>=600 AND Np<=1.0 for >=6h) over 1996-2025; OMNI coverage gap 2.2%; 1566 M6.5+ fetched, 1310 after declustering. Observed 6 events in exposure windows against a circular-shift null mean of 7.4 (95% CI [2,13]); k=0.81, p=0.755 -- observed sits BELOW the null, no hint of elevation. SENSITIVITY: the pass threshold was >13, so this excludes lift >= 1.76 at 97.5% and cannot exclude smaller effects. 31 onsets is just above the lock's floor of 30, so the test is at the edge of its declared power and that is a real limit, not a caveat added afterwards. All UNRESOLVED conditions in the lock were checked and none tripped.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-15T21:35:41Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"HSS onset = Vsw>=600 km/s AND Np<=1.0 cm-3 sustained >=6h, onsets merged within 48h. Null = circular time shift of the EQ catalogue vs the solar record, lag uniform in [1yr, span-1yr], 2000 draws, preserving the solar cycle and EQ clustering while destroying only the alignment.","n_permutations":2000},"threshold":{"direction":"positive","metric":"observed_count_vs_null_p97.5","value":0.975},"title":"Coronal-hole high-speed streams elevate the global M6.5+ rate within 48h","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"data/igets_cleaned_v1 (manifest sha256 021e7c1cbadce138917ed6fe9fe4f99e00f1757f3fef2a294cf8983cca672361) + data/igets_constituent_table_v1.json (sha256 4906e5d37d7fbd0165292dc3df9155001681e75b85719ff0bfffbf3c4b29242e) via pipeline commit 298a4c60","expected_direction":"positive","feature_set":["slichter_S_m_neg1","slichter_S_m_0","slichter_S_m_pos1","slichter_neff_inband","slichter_a95_nm_s2"],"git_commit_at_registration":"298a4c60b4a4a511a3f1f063743fb4da5f1e3870","hypothesis_id":"20260822_slichter_1s1_triplet_search_v2_o_9af67bec","locked_at":"2026-08-22T22:46:48Z","locked_file_sha256":"822635eaa2e17287021c61d1ff9d0fe929e28aa08739fc19d1a8a042ed6b52d8","notes":"D.1-SPRINT flagship deliverable. Sealed BEFORE any scored statistic touches the 3-9 h band. Any C1+C2+C3 pass is a DETECTION-CANDIDATE only; expected outcome is a methods-clean per-channel A95 bound at injection-measured gain. UNRESOLVED is a third outcome. Full procedure in the prereg doc (sha in test.extra).","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"solid_earth_normal_mode"},"reevaluate_at":"2026-09-30","registered_at":"2026-08-22T22:46:29Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"c462facfbb5b1194cc7d369cb4ed99e06c111558","interpretation":"disconfirming","metric_value":1.0,"n_observed":18,"notes":"OUTCOME CLASS: NULL_A95 (the pre-declared expected case; 'failed' here means no detection-candidate, not an error). All three Y1 channels: A95 median 0.100/0.102/0.103 nm/s2 (~10 nGal), band worst-case 0.141/0.149/0.145, injection-measured gain 0.971-0.978 (stability <=2.8%), N_eff in-band 10.9/9.3/10.6 of 18. C1 fired at 8.3853 h in all channels (S up to 6175) and C3 geometry rejection KILLED it: MO3/2MK3 compound terdiurnal tide, permuted-weight survival 50/0/5 of 100 vs >=95 required -- the matched-filter-vs-power-meter gate working as sealed. Smylie-1992 frequencies: nothing above threshold; A95 sits ABOVE the claimed few-nGal amplitudes so NO exclusion language per the sealed caveat. Null quality: is_alive PASS x3, tail check 0.888/0.915/0.932 in [0.5,2.0], N_trials ~150-194k measured Lee-style. One substantive deviation (D2, dummy-null tidal exclusion extended to the Doodson species-2 band by fixed catalog boundary, pre-scoring) + 4 minor: hypotheses/deviations/slichter_v2_run_deviations_v1.md. Full report: _active_today/slichter_v2_null_a95_v1.md. Self-test 19/19 laptop+AX41 bit-identical.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-22T23:33:20Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"a95_rule":"injection ladder brackets crossing 0.5-4.0x, interpolate never extrapolate, gain stability 15% else UNRESOLVED","band_period_h":[3.0,9.0],"channels":[-1,0,1],"criterion":"C1 trials-corrected p_FA 0.001 dummy-frequency null AND C2 split-half amp ratio [0.5,2.0] phase<=45deg AND C3 weight-permutation geometry rejection 95/100 below 50%","null":"empirical dummy-frequency >=200, null_quality in-path, tail-checked","outcome_classes":["DETECTION-CANDIDATE","NULL_A95","UNRESOLVED"],"prereg_doc":"hypotheses/slichter_v2_search_prereg_v1.md","prereg_doc_sha256":"c4ad30b28c2e41721ffbf97c3f3b6b82ccac50f2d21260663d1c2af35140bfc4","station_exclusion_rule":"in-band median whitened floor > 5x network median, applied before scoring"},"method":"sealed_matched_filter_search","n_permutations":null},"threshold":{"direction":"positive","metric":"trials_corrected_p_fa","value":0.001},"title":"Slichter 1S1 triplet search v2 on the piece-3 cleaned IGETS store: sealed matched-filter criterion, detection-candidate or A95 bound","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"data/igets_cleaned_v2r (manifest sha256 29c2a4f7c6e3be3ada244393128d3396ca342c2505f58a285951ba05c098e81d, 20 piers / 36 segments / 303.3 station-years) via scripts/igets/slichter_search_v4.py (self-test 53/53)","expected_direction":"positive","feature_set":["slichter_v4_a95_persistent","slichter_v4_a95_damped_q500","slichter_v4_a95_damped_q1000","slichter_v4_a95_damped_q2000","slichter_v4_a95_damped_q5000","slichter_v4_neff_inband"],"git_commit_at_registration":"57a98eb23776e7844849b43f730f3597c4f885bd","hypothesis_id":"20260823_slichter_1s1_search_v4_the_first_84e85c3b","locked_at":"2026-08-23T17:54:34Z","locked_file_sha256":"85cb78b3d59ed279033f72a36f2126111fac0bdaa45ad46293b79330ea822f18","notes":"The damped tier is the reason v4 exists: every prior bound we located, including our own v2/v3, calibrates on an UNDECAYED sinusoid while the mode is damped. It is expected to make the headline number WORSE by 3.8-9.6x, and that is correct -- a decaying signal is harder to detect. Q list declared here and not extendable; onset randomised so no favourable onset can be chosen; S_thr never recomputed per Q. Limitations declared in the doc rather than discovered by a reader: the two calibrator-flagged segments were cleared on out-of-band constituents only, the species-2 null exclusion was adopted after seeing the tail, Smylie exclusion language stays forbidden, and Ding & Chao remain tighter.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"solid_earth_normal_mode"},"reevaluate_at":"2026-11-30","registered_at":"2026-08-23T17:54:09Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"e8f8900de0fd010360f20aa73a3921b9040203e2","interpretation":"disconfirming","metric_value":1.0,"n_observed":20,"notes":"OUTCOME: persistent tier NULL_A95 x3; damped tier UNRESOLVED in ALL 15 channel x Q cells. Persistent A95 0.0804/0.0847/0.0809 nm/s2, N_eff 11.6/9.3/11.5 of 20, gain 0.926-0.956, u95 all bracketed and interpolated. PRE-STATEMENT 1 MISSED IN THE SAFE DIRECTION and the red flag did not fire: predicted 0.0604/0.0649/0.0618, measured x1.31-1.33 WORSE -- and x1.09-1.11 worse than v3 DESPITE 26% more data. Accounting via a 2x2 store x weighting census with a positive control that reproduces v3 phase A to every digit: the store repair itself cost x1.09-1.11 (the repaired segments carry a heavier noise tail -- more data is not a better bound when the added data is noisier), while the 1/sigma^2 weighting's x1.071 kappa gain is CONFIRMED (x1.072-1.078) and WORTH NOTHING because the empirical S_thr rises x1.065-1.091 with it. Null empirical max 25.8->38.3 is a fit-free confirmation the tail really got heavier. Same cancellation mechanism the v3 pier-rule control found. PRE-STATEMENT 2 satisfied: exclusion did not fire (Moxa 50.00x->0.83x, Onsala 82.43x->1.00x, Pecny 60.16x->1.22x), realised 303.256 st-yr/20 piers, restated prediction equals the budgeted one to three digits. PRE-STATEMENT 3 CONFIRMED without one inversion across 9 rows: nothing came back tighter. C3 KILLED THE SAME FALSE ALARM A THIRD TIME -- C1 fired at 8.3854 h at 168/68/107x threshold, C3 43/0/27 of 100. All five top peaks are M3 satellites, and v3's MO3/2MK3-Mm label is RELABELLED to M3-2Mm (disclosed post-hoc). THE DAMPED TIER IS THE FINDING: not weak, NO A95 EXISTS -- peak detection fraction collapses 1.000 (undecayed) -> 0.31-0.54 (Q=5000) -> 0.008-0.11 (Q=2000) -> 0.000 (Q<=1000). MECHANISM MEASURED with a two-armed frozen-vs-self whitening control over a 3000x sweep: the estimator WHITENS THE LORENTZIAN AWAY (HWHM 0.037 blocks at Q=5000, 0.372 blocks at Q=500); frozen-whitening gain stays flat <=x1.7 while self-whitened collapses x880/x304/x103. THE PREREG'S OWN x3.8-9.6 ESTIMATE IS REFUTED -- true penalty x100-1700 -- because it assumed a filter MATCHED to the transient (~sqrt(Q)) while the sealed estimator keeps the persistent filter (~Q). The persistent path saturates by the same mechanism but at 1085x, so u95~1.38 sits at gain 0.96 and THE HEADLINE A95 IS UNAFFECTED. Smylie rows p_corr=1, no exclusion language; Ding & Chao remain tighter and nothing is presented otherwise. Self-tests 53/53 x2 weightings both machines; v4 under invsigma reproduces the graded module's self-test 19/19 lines byte-identical. All sealed artifacts byte-identical before and after every phase. One declared judgement call: the damped rung ladder had to be scaled because the sealed 0.5-4.0x rungs cannot bracket a crossing at gain ~1e-3 and extrapolation is forbidden; it could not have changed the outcome, since the response DECREASES with amplitude. Report: _active_today/slichter_v4_null_a95_v1.md; deviations: hypotheses/deviations/slichter_v4_run_deviations_v1.md.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-23T22:13:34Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"Q_list_DECLARED_not_extendable":[500,1000,2000,5000],"band_period_h":[3.0,9.0],"onset":"randomised uniform over the record; A95 across onsets, never a chosen onset","outcome_classes":["DETECTION-CANDIDATE","NULL_A95","UNRESOLVED"],"predecessors":["20260822_slichter_1s1_triplet_search_v2_o_9af67bec","20260823_slichter_1s1_triplet_search_v3_i_8a6809e7"],"prereg_doc":"hypotheses/slichter_v4_search_prereg_v1.md","prereg_doc_sha256":"ad00fb1a5c1492182cd0ecf45f73734baa60d8df4103d928eb6f0f7786802235","prestated":{"damped":"MUST be WEAKER than persistent for every finite Q; tighter = DEFECT -> UNRESOLVED","multiplier":"restate 1/sqrt(T) from the REALISED scored station-years, both stated","persistent":"~0.060/0.065/0.062 nm/s2; materially better is a RED FLAG"}},"method":"sealed_matched_filter_search","n_permutations":null},"threshold":{"direction":"positive","metric":"trials_corrected_p_fa","value":0.001},"title":"Slichter 1S1 search v4: the first Q-aware (damped-transient) bound, plus exclusion-radius fix and inverse-variance weighting","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"data/igets_cleaned_v2 (manifest sha256 ba58096597c667f233f6eba49e8879e59764d73cdfb0caa5223306ef5772552f, 45 npz / 20 stations / 45 segments) via estimator scripts/igets/slichter_search.py at commit dac19367","expected_direction":"positive","feature_set":["slichter_v3_S_m_neg1","slichter_v3_S_m_0","slichter_v3_S_m_pos1","slichter_v3_neff_inband","slichter_v3_a95_nm_s2"],"git_commit_at_registration":"bc1481233eda46c07377dc53af5583a09b8b0c67","hypothesis_id":"20260823_slichter_1s1_triplet_search_v3_i_8a6809e7","locked_at":"2026-08-23T03:10:54Z","locked_file_sha256":"779d3da858732c4deb1a650087bee3e7152ec52602412a5e24d573ee24ee5879","notes":"Pure data upgrade: same estimator, same band, same C1/C2/C3, same null machinery as the graded v2. Exactly three declared deviations, all in the prereg. The pier rule is load-bearing -- 8 of 20 stations are multi-code (dual-sphere or successive epochs) and two spheres on one pier see the SAME ground, so counting them as separate channels would inflate the array. Expectation stated in advance so the improvement cannot be judged after the fact.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"solid_earth_normal_mode"},"reevaluate_at":"2026-10-31","registered_at":"2026-08-23T03:10:25Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"b1c2579813a64a4e33108519dc9210ce425211fa","interpretation":"disconfirming","metric_value":1.0,"n_observed":17,"notes":"OUTCOME CLASS: NULL_A95 all three channels (the pre-declared expected case; 'failed' = no detection-candidate, not an error). A95 median 0.0726 / 0.0780 / 0.0743 nm/s2 (7.3-7.8 nGal) vs v2's 0.100/0.102/0.103 -- improvement x1.377/1.307/1.387. PRE-STATED RED FLAG DID NOT FIRE: the improvement is WORSE than the predicted x1.59, the safe direction, and is fully accounted for -- the sealed exclusion rule dropped 3 piers so the REALISED data multiplier is x2.057 not x2.54 (116.9 -> 240.5 scored station-yr), whose 1/sqrt(T) prediction is x1.434, and measured is 96/91/97% of that. The bound behaves as noise-limited; no pier double-counting. Exclusion fired on Moxa 50.0x / Onsala 82.4x / Pecny 60.2x, diagnosed producer-side: 5 segments of the frozen store carry un-removed glitches (max|g| 54,000-94,000 nm/s2 against ~250 clean). C1 fired again at 8.3854 h (S 7672/9489/8046 vs thr 45.7/66.0/46.8) and C3 killed it again at 28/14/32 of 100 -- and the identification IMPROVED to M3-Mm at delta 2.8e-6 cpd, ~100x closer than v2's MO3/2MK3 label for the same line; 4 of the top 5 are M3 satellites, i.e. the cleaning removes the M3 central line but none of its satellites (highest-leverage v4 fix). PIER RULE MEASURED not asserted: the wrong split overstates apparent array size by x1.24/1.31/1.24 and raises max single-site weight share 0.074 -> 0.159; a third hazard was REFUTED -- the floor does not drop, because the empirical null's threshold rises with the inflated array and cancels the gain. Smylie rows below threshold, NO exclusion language (no checked citation for a claimed amplitude). Gates: is_alive PASS x3, tail 0.912/0.916/0.935, 103,520 dummy draws, N_trials 167-280k. Self-test 19/19 on both boxes plus 5/5 pier unit tests. Deviations: hypotheses/deviations/slichter_v3_run_deviations_v1.md. Report: _active_today/slichter_v3_null_a95_v1.md.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-23T03:47:26Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"band_period_h":[3.0,9.0],"channels":[-1,0,1],"deviations_from_v2":["pier rule: one geometric weight per pier regardless of sphere/epoch count","split halves re-declared for 20 stations","species-2 null exclusion promoted from deviation to method"],"outcome_classes":["DETECTION-CANDIDATE","NULL_A95","UNRESOLVED"],"predecessor":"20260822_slichter_1s1_triplet_search_v2_o_9af67bec","prereg_doc":"hypotheses/slichter_v3_search_prereg_v1.md","prereg_doc_sha256":"10f2ed93e2443d7556ceab8e8a99eb97fe8150d1bb76c88ae26f768b90a4d677","prestated_expectation":"A95_v3 ~ A95_v2/1.59 ~ 0.063 nm/s2; a ratio materially better than 1.59 is a RED FLAG not a triumph"},"method":"sealed_matched_filter_search","n_permutations":null},"threshold":{"direction":"positive","metric":"trials_corrected_p_fa","value":0.001},"title":"Slichter 1S1 triplet search v3: identical v2 instrument on the full-archive store (306 station-years, x2.54) -- pure data upgrade","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"IGETS SG network, 21 stations (frozen). trackQ Aug-23 screened products sha dcffe8b5; constituent table sha 4906e5d3; NPZ 18 v1 + 3 q13; loading TPXO10v2a sha ae8075e6 (model of record) + FES2014b sha a955fc7d (sigma_load); sealed-fit script sha c2bf5467","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"0e138fa39a110f821a104a0c1a412efa93f8ce43","hypothesis_id":"20260829_l_4_wahr_1981_admixture_in_the_m_2cb6bcb0","locked_at":"2026-08-29T01:32:18Z","locked_file_sha256":"92cf0cc31399907f847dcfe85c864bbb337de7a3ce6549d6617977cbee6c10cb","notes":"METHOD-FROZEN REPRODUCIBLE MEASUREMENT, not forward-blind: the scored fit was RUN to validate the controls (eps identical before/after a negative-fixture bug fix, so no p-hacking). The anchor certifies procedure+frozen-input-shas -> deterministic result (seed 20260828), timestamped. Onsala adjudicated by TPXO (44% raw loading diff = FES2014b Kattegat artifact) and down-weighted to 0.009x via sigma_load, not excluded. Controls all pass: shuffle-null alive, injection recovery +-0.7012 for +-0.7, negative fixture recovers 0.20% (correctly fails). Script scripts/igets/l4_sealed_scored_fit.py.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":null,"registered_at":"2026-08-29T01:32:08Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"0e138fa39a110f821a104a0c1a412efa93f8ce43","interpretation":"confirming","metric_value":0.076,"n_observed":21,"notes":"eps_4 = +0.076% +- 0.30%; boot 2sigma [-0.45,+0.76] includes zero -> NULL_BOUND; |eps| < 0.675% at 95% (below Wahr 0.7%; Wahr=2.34 sigma). First observational constraint on the l=4 admixture. Controls 3/3 pass. PASSED = a valid publishable bound was obtained with all controls alive (the prereg's declared bound outcome), NOT a Wahr detection.","ots_proof_path":null,"passed":true,"tested_at":"2026-08-29T01:32:18Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"weighted LS of A/cos^2(phi) on [1, g(phi)], g=(7sin^2phi-1)/6; eps=c4/c0; weights 1/(sigma_floor^2+sigma_load,i^2); sigma_load=|A_FES-A_TPXO|/A","n_permutations":1000},"threshold":{"direction":"positive","metric":"epsilon_4_pct","value":0.7},"title":"l=4 Wahr-1981 admixture in the M2 gravity tide -- method-frozen reproducible measurement (NULL_BOUND)","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"IGETS SG network, M3-satellite-cleaned store igets_cleaned_v3sat (manifest sha 003403a3ba76f318); constituent table sha 4906e5d3; graded harness slichter_search_v6_graded.py (v4-based, pointed at v3sat) sha ea7c71cfd369d812","expected_direction":"negative","feature_set":[],"git_commit_at_registration":"68572a31cec19feb8fe8956a9d3f8ef7f08e1cb8","hypothesis_id":"20260829_slichter_1s1_search_v6_m3_satell_b3cd85d5","locked_at":"2026-08-29T02:06:51Z","locked_file_sha256":"792a004ad5ec0391cc54c4e690c3340ac8f26e5ebbd8ee750156e4109e3af118","notes":"FORWARD-BLIND: --phase C has NOT been run; the search refuses it pre-seal by design, so this is a genuine prediction. DECLARED EXPECTATION (armed red flag if wrong): with the M3 satellite forest removed, C3 NO LONGER trips on the M3 false alarm, so the scored Slichter-band statistic can be honestly evaluated; the persistent Slichter bound STAYS systematics-limited (A95/sigma ~6-9) -- v6 is a DETECTABILITY fix, NOT a bound change. Outcomes: C3_PASS + no candidate -> the detectability fix works and the null stands honestly (expected); C3_PASS + a candidate surviving all gates at sealed significance -> a real Slichter detection (would be extraordinary); C3_STILL_TRIPS -> the M3 removal did not clear the gate (surprise, red flag). Graded run via v6_graded --phase C on v3sat AFTER this seal + OTS anchor.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":null,"registered_at":"2026-08-29T02:06:45Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"ffa85548a295ac54bd5869caa800996a8cdef08d","interpretation":"disconfirming","metric_value":7.73,"n_observed":3,"notes":"FORWARD-BLIND RESULT (run AFTER the OTS seal, via the corrected trackAB v4-caller on v3sat -- deviation 2). Outcome NULL_A95 all 3 channels. The declared expectation ('M3 forest removed -> C3 no longer trips -> clean evaluation') is REFUTED: C1 STILL fires on the top peak at 8.3712 h = the M3-Mm SATELLITE residual (S_max 538/264/422 vs S_thr 70/98/73, ~7.7x over -- down 20x from v5's ~140x, so the M3 cut is real but PARTIAL), and C3 still rejects it (pass=False). The residual is the 18.6-yr NODAL-MODULATION sidelobe that a constant-line M3-satellite basis cannot clear (slichterM3 flagged this). So v6 improves detectability 20x but does NOT achieve a clean Slichter evaluation; the null stands (NULL_A95) for the same systematics/residual reason, not because the store is clean. RED FLAG (C3_STILL_TRIPS) partially fired -- honest, informative refutation. NEXT (v7): clear the on-line residual with a nodal (f,u) correction, NOT more constant lines. Bound unchanged (systematics-limited); v6 was a detectability attempt, not a bound change.","ots_proof_path":null,"passed":false,"tested_at":"2026-08-29T02:29:40Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"graded scored pass (v6_graded --phase C): C1 stack -> C2 significance -> C3 geometry gate -> tail check -> N_eff, on the M3-satellite-removed store. The M3 forest that tripped C3 140x in v4/v5 is removed (transparency peakfinder confirmed 20x suppression, top peak 17717->869).","n_permutations":null},"threshold":{"direction":"negative","metric":"C3_geometry_over_threshold","value":1.0},"title":"Slichter 1S1 search v6 -- M3-satellite-cleaned store, forward-blind C3 + scored pass","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"data_source":"IGETS SG network, NODAL-corrected store igets_cleaned_v3nod (full re-clean from v2r, 20 piers/36 segments/303.3 station-yr, manifest sha ee79c1e56dc4389b3a9ab75bc75111c461be34ff12a8b7aaaaf8e2867fbbfa22); built by v7 fit module slichterM3_tidal_harmonic_fit_v7.py sha 6109e6c2a0a2060c (per-constituent Schureman node factors on the M3 family + the M3+Mm-MSf combination line, node modulation IN the joint fit_model, node freq 18.6136 yr); graded estimator slichter_search_v4.py sha 2438dc37cf8a4631 (UNEDITED); graded caller trackAB_slichter_v4.py sha 5b8d0a94d31e933a (v3nod store-registration edits only).","expected_direction":"negative","feature_set":[],"git_commit_at_registration":"bb44b90847bae3a14f5a513621ef2797664b6445","hypothesis_id":"20260829_slichter_1s1_search_v7_nodal_f_u_4d21a75a","locked_at":"2026-08-29T03:24:19Z","locked_file_sha256":"49f95b7805580485e02747e555e3724ebc07ab0634a837122e104a942d318a9e","notes":"FORWARD-BLIND: --phase C has NOT been run on v3nod; the search refuses it pre-seal by design, so this is a genuine prediction. v7 removes the M3 ter-diurnal FOREST via NODAL (f,u) modulation of each M3-family line (M3 carrier + 5 satellites + M3+Mm-MSf), applied IN the original joint fit_model (not an incremental pass -> no carrier injection). TRANSPARENCY (allowed pre-seal, top-5 peakfinder, NOT the scored statistic): the M3 carrier is GONE from every channel's top-5, the dominant in-band peak is now a solar S3/S4 harmonic (154-174) not an M3 line, M3-satellite fraction 8/15 -> 4/15, and the remaining M3+Mm (100-128) sits at/below the noise floor. DECLARED EXPECTATION (armed red flag if wrong): with the M3 forest removed, C3 NO LONGER trips on the M3 residual, so the scored Slichter-band statistic can be honestly evaluated; the persistent Slichter bound STAYS systematics-limited (A95/sigma ~6-9) -- v7 is a DETECTABILITY fix for the M3 forest, NOT a bound change. HONEST UNCERTAINTY declared BEFORE the run: the peakfinder shows solar S3/S4 harmonics now top the band; if C3 relocates to and trips on S3/S4, that is a NEW nuisance finding (a v8 target), NOT a v7 failure -- v7's claim is specifically that the M3 FOREST no longer blocks C3. Outcomes: C3_PASS + no candidate -> M3 detectability fix works, null stands honestly (expected); C3_PASS + a candidate surviving all gates at sealed significance -> a real Slichter detection (extraordinary); C3_STILL_TRIPS_ON_M3 (top peak an M3-family line) -> nodal removal did not clear the M3 gate (surprise, red flag); C3_TRIPS_ON_NON_M3 (S3/S4) -> M3 cleared but a new nuisance blocks (partial success, directs v8). Graded run via trackAB v4-caller --store v3nod --phase C AFTER this seal + OTS anchor.","ots_proof_path":"hypotheses/locked/20260829_slichter_1s1_search_v7_nodal_f_u_4d21a75a.json.ots","outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":null,"registered_at":"2026-08-29T03:24:06Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"72c95f4335a3e6a861572dfb66e905dfa2209877","interpretation":"confirming","metric_value":77.0,"n_observed":3,"notes":"FORWARD-BLIND RESULT (graded via trackAB v4-caller --store v3nod --phase C AFTER the OTS seal; unedited estimator sha 2438dc37). The sealed PRIMARY prediction HELD and the armed red flag (C3_STILL_TRIPS_ON_M3) did NOT fire: the C1 top peaks are now m_neg1 6.210 h (S4 solar), m_0 8.007 h (S3 solar), m_pos1 6.210 h (S4 solar) -- NONE at M3 (8.28 h) or M3-Mm (8.37 h). The M3 ter-diurnal forest is fully cleared from the scored top peaks; the nodal (f,u) correction did exactly what it claimed (a DETECTABILITY fix). This is the seal's PRE-DECLARED 'C3_TRIPS_ON_NON_M3 (S3/S4) -> partial success, directs v8, NOT a v7 failure' outcome: C3 still fails overall (below50 m_neg1 66/100, m_0 46/100, m_pos1 77/100 vs need 95) but now on the SOLAR S3/S4 harmonics, not M3. Progression across the lineage: v5 C1 fired ~140x over threshold on M3, v6 ~7.7x on M3-Mm, v7 M3 entirely off the top peaks with C3 below50 improved from v6's 6-40 to 46-77. Outcome NULL_A95 all three channels (u95_inband ~1.37) -- the persistent Slichter bound STAYS systematics-limited exactly as predicted (a detectability fix, not a bound change; no detection). INTEGRITY: the first (foreground) scored pass froze S_max then hit the wall-clock timeout before grading; the run was resumed to completion (the killed run's partial artifact is a pre-normalization intermediate, NOT a valid determinism reference). The estimator is deterministic (SEED=20260822 + frozen store/estimator shas); the resumed scored S_max (97.03/127.17/110.03) matches the C1 grading exactly. metric_value = 77 (best-channel C3 below50, need 95). V8 TARGET (now unambiguous): remove the solar S3/S4 (thermal/diurnal-harmonic) nuisance from the store-conditioning, the same way v7 removed the M3 forest, then re-seal forward-blind.","ots_proof_path":null,"passed":true,"tested_at":"2026-08-29T03:50:09Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"eval_C3 permutation-geometry gate at kstar (sealed v4 estimator C3_FRAC/C3_NEED); C1 peak S_max/S_thr per channel","n_permutations":null},"threshold":{"direction":"negative","metric":"C3_pass_all_channels_and_M3_not_top_peak","value":1.0},"title":"Slichter 1S1 search v7 -- nodal (f,u) M3 forest correction (full re-clean)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"IGETS SG network, store igets_cleaned_v3sol (M3 forest + M4/S3/S4/MS4/MK4/SK3 cleaned, 20 piers/36 segments/303.3 station-yr, 38 constituent lines, manifest sha aec6b54c7a76cdeccb37c4e7f9622c2566e4fd58da4612daec47e83537584e58); NEW estimator slichter_search_v9.py CANONICAL (LF/git) sha b4089bce04d7d4058001e2de31cc2091b2bf001fd65175e024fc3eabb039cb49 (v4 + a constituent-line mask on the C1 kstar search + find_peaks; self-test 53/53); the base GRADED estimator slichter_search.py sha 0f43f909... is UNCHANGED; graded caller trackAB_slichter_v9.py sha fffb1b61a59e9a97 (imports v9, v3sol store-registration). The SEALED v4 estimator (2438dc37) is byte-untouched -- v9 is a NEW lineage.","expected_direction":"negative","feature_set":[],"git_commit_at_registration":"740b0f86655a61326f5478be8c78c96bbaa5e3c6","hypothesis_id":"20260829_slichter_1s1_search_v9_line_mask_8579f6b3","locked_at":"2026-08-29T14:03:32Z","locked_file_sha256":"d882abb27233ef67752ba80c93e721d89f1877b7c1f00348816ec415b3582c73","notes":"FORWARD-BLIND: --phase C has NOT been run on v3sol through v9; sealed BEFORE any scored pass. (Re-seal of the withdrawn v9_line_mask_6432d7ef, whose recorded estimator sha was the Windows CRLF working-copy hash; this seal uses the CANONICAL LF/git sha b4089bce. No scored pass ran under the withdrawn seal -- it REFUSED on the sha mismatch -- so forward-blindness is intact.) v7/v8 hit a SPECTRAL-LEAKAGE WALL: each strong removed tidal/thermal line (M3 carrier, then M4, then S3) leaves a WHITENED peak that captures C1's kstar, so C3 tests a line's geometry (strong at all stations, not Slichter-like) and fails -- and cleaning one line just moves kstar to the next. v9 is the ESTIMATOR-LEVEL fix: exclude every constituent-line neighborhood (+-one whitening block) from the C1 kstar search + the transparency peak search, so kstar lands in the clean GAPS between lines, where the Slichter triplet (3.43/4.36/5.33 cpd) lives. CONSERVATIVE by construction: masking only SHRINKS the search space, so keeping v4's sealed null threshold makes C1 strictly HARDER to pass (no false-alarm inflation); C3's permutation null self-calibrates at whatever kstar survives. DECLARED EXPECTATION (armed red flag if wrong): C1's kstar now lands OFF every tidal/thermal line (> the exclusion radius from all 38 constituents); the persistent Slichter bound STAYS systematics-limited (a search-geometry fix, NOT a bound change). Since these 20 piers carry no known Slichter mode, the honest expected outcome is NULL_A95 with kstar OFF-LINES and C3 correctly failing on the residual noise geometry -- i.e. the estimator is no longer FOOLED by line leakage, which is the endpoint of the cleaning+estimator program. Outcomes: KSTAR_OFF_LINES + NULL_A95 -> the fix works, null stands honestly (expected); KSTAR_OFF_LINES + C3 PASS + candidate surviving all gates -> a real Slichter detection (extraordinary); KSTAR_STILL_ON_A_LINE -> the mask failed (surprise, red flag). Graded via trackAB v9-caller --store v3sol --phase C AFTER this seal + OTS anchor.","ots_proof_path":"hypotheses/locked/20260829_slichter_1s1_search_v9_line_mask_8579f6b3.json.ots","outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":null,"registered_at":"2026-08-29T14:03:19Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":{"git_commit_at_test":"8b9a746c6dc79d0d650e335a1103079564463b7c","interpretation":"confirming","metric_value":31.4,"n_observed":3,"notes":"FORWARD-BLIND RESULT (graded via trackAB v9-caller --store v3sol --phase C AFTER the OTS seal; v9 estimator sha b4089bce verified pre/post-run). The sealed PRIMARY prediction HELD and the armed red flag (KSTAR_STILL_ON_A_LINE) did NOT fire: C1 kstar is OFF all 38 masked constituents in every channel -- m_neg1/m_pos1 at 2.7944 cpd (31.4 mcpd past the 27.5 mcpd mask edge), m_0 at 4.9398 cpd (4.859 h) a full 0.94 cpd from any line, DEEP in the clean Slichter-band gap where the triplet (3.43/4.36/5.33 cpd) would live. The estimator-level line-mask BROKE the v7/v8 spectral-leakage wall: kstar moved from ON the tidal/thermal lines (v7/v8, d<2 Ray) to OFF them. Outcome NULL_A95 all three channels -- with the estimator finally searching the real off-line geometry, there is NO detectable Slichter 1S1 mode at these 20 piers (C1 fails in m_0 at the in-gap kstar, 67.4 vs S_thr 98.4; the one C1-passing channel m_neg1 fails C3 at 37/100). The persistent bound STAYS systematics-limited, as predicted (a search-geometry fix, not a bound change). HONEST nuance for a future v10: m_neg1/m_pos1 kstar at 2.7944 cpd sits at ~M3-Mf, a tidal SATELLITE not in the 38-constituent masked set -- so the mask cleared every KNOWN constituent but an unmasked satellite remains the strongest off-mask peak; adding the M3-Mf family would tighten it, but C3 already rejects it. This is the honest endpoint of the cleaning+estimator program: the leakage capture is broken and no Slichter mode is present. metric_value=31.4 (min off-line mcpd, all > the 27.5 mcpd mask radius).","ots_proof_path":null,"passed":true,"tested_at":"2026-08-29T14:08:59Z","tested_by":"Mljessop@Ubuntu-2404-noble-amd64-base"},"status":"tested","test":{"extra":{},"method":"line-masked C1 kstar (exclude +-1 whitening block around all 38 constituents); C3 permutation-geometry at kstar; sealed v4/v9 null threshold","n_permutations":null},"threshold":{"direction":"negative","metric":"C1_kstar_off_all_constituent_lines","value":1.0},"title":"Slichter 1S1 search v9 -- line-masked kstar (estimator-level fix)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"INTERMAGNET 11-observatory K1 synthetic injection campaign (33 yr, n=500) with graded oceanic weight from GEBCO2020 bathymetry + Natural Earth 50m coastlines","expected_direction":"","feature_set":["k1_sidereal_amplitude","rho_e_arm_e_spatial"],"git_commit_at_registration":"4b57bcb10231d8261969502956b62f4dc69d8459","hypothesis_id":"20260903_dm_magnetic_k1_adjudication_v2_4_cf01e1fb","locked_at":"2026-09-03T16:16:24Z","locked_file_sha256":"1117aeab521839836fac2dfb0de16d43b8c419da931e15f43abc920dce85570c","notes":"DM Magnetic K1 Adjudication v2.4 (graded oceanic injector + arm (e) spatial vote). Supersedes the v2.3 exploratory battery, which was NOT sealed as an adjudication because a purely oceanic K1 was returned DM-flavoured in 10.4% of realizations and the D9 48x separation was a boolean-injector tautology. v2.4 replaces the boolean COASTAL step with a PHYSICAL graded weight w_phys(s) (GEBCO bathymetry conductance x K1 skin depth; corr(w_phys, distance)=-0.41, which breaks the arm-(e) circularity), restores Sq in the oceanic arm, and promotes D9 to a VOTING spatial discriminant rho_e = corr(per-station sidereal amplitude, w_phys) with threshold T9 = p99 of rho_e over the pure-DM arm (forward-blind, DM-derived, never the oceanic gap). PRE-COMMITTED: oceanic false-DM at T9 < 2% (from 10.4%), DM false-demote <= ~2%, rho_e disjoint, base gates pass; FALSIFIER stated. Frozen code scripts/dm_mag_k1_adjudication_v24seal.py sha256 44ea9cdab31c68e18b5397c11019d714315e0c6949ef77d7500251b538aff889 (git 4b57bcb1) derived from frozen v2.3 097face5 by nine anchor-asserted edits; prereg _active_today/dm_mag_k1_adjudication_prereg_v2_4_AMENDMENT.md sha256 8bfbb2b855c1aaabfc966deabf25d2978bef8efc5bedbd852d04d61b27ed43f9. seed = sha256(EARLIEST attesting Bitcoin block hash of THIS anchor), no realization scored before confirmation. expected_direction=OCEANIC_HOLE_CLOSED_OR_BOUNDED.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dm_magnetic_k1_adjudication"},"reevaluate_at":null,"registered_at":"2026-09-03T16:15:57Z","registered_by":"Mljessop@deepmap-compute","result":{"git_commit_at_test":"b5e8ce2a9392e5a95d523f5b83c4f0ad08739b1b","interpretation":"confirming","metric_value":0.0,"n_observed":500,"notes":"SEALS AS ADJUDICATION -- oceanic false-DM hole closed. Forward-blind grade seed 1003605714 n=500: oceanic_false_dm_at_T9=0.0 (bar 0.02, v2.3 baseline 0.104), dm_false_demote=0.01 (<=0.02), rho_e disjoint (d.p05 0.7356 > b.p95 0.3013), base gates pass. v2.3 boolean-injector D9 tautology replaced by non-circular arm-(e) spatial vote (graded w_phys). Legacy recovery_gates R_cal check reads false = known night-retention finding, not a sealed criterion.","ots_proof_path":null,"passed":true,"tested_at":"2026-09-04T16:48:59Z","tested_by":"Mljessop@deepmap-compute"},"status":"tested","test":{"extra":{"amendment_v2_4":{"arm_e_not_circular":"geometry_g is spatially FLAT (unit norm every station) so DM -> rho_e~0; oceanic induction -> amplitude tracks conductance -> rho_e high. shape-corr(w_phys, geometry_g) = -0.069.","changes":["graded oceanic coupling w_phys(s) injected at every station (continuous, non-zero inland), replacing the boolean COASTAL step -- kills the v2.3 tautology","Sq restored in the oceanic arm (k1_ion no longer 0): ocean measured against a real background","arm (e) VOTING statistic rho_e = corr(per-station sidereal amplitude, w_phys)","threshold T9 = p99(rho_e | pure-DM arm b), forward-blind, DM-derived, never the oceanic gap"],"seed_rule":"K1_SEED_CAMPAIGN = int(sha256(EARLIEST attesting Bitcoin block hash of THIS anchor)) mod 2^31, set at seal, passed via env; dev seed 20260807 is fallback only.","w_phys":{"BOU":0.016,"CMO":0.0193,"CNB":0.5304,"HAD":0.0186,"HER":0.166,"HUA":0.2962,"KAK":0.3438,"KOU":0.0924,"NGK":0.0084,"SJG":1.0,"TAM":0.0061},"w_phys_derivation":"C_ocean(s)=sigma_sw*mean GEBCO2020 nearby ocean depth, times exp(-d/delta), delta=K1 skin depth 1476 km (rho=100 ohm.m). corr(w_phys, distance)=-0.414 vs -0.996 for bare exp(-d/L) -- this is what makes arm (e) non-circular."},"caveats":["residual circularity (bounded): oceanic rho_e ~+1 is partly the injector read back; discrimination rests on DM != +1. Arm (h) (mis-specified ocean model) is the named follow-up.","DM centers at rho_e ~ -0.41 not 0 (benign per-station recovery systematic); T9 self-calibrates on the DM arm so it is handled, not tuned around."],"code_git_commit":"4b57bcb10231d8261969502956b62f4dc69d8459","code_path":"scripts/dm_mag_k1_adjudication_v24seal.py","code_sha256":"44ea9cdab31c68e18b5397c11019d714315e0c6949ef77d7500251b538aff889","dev_probe_directional_only":{"T9":-0.125,"dm_false_demote":0.05,"dm_rho_e_p50":-0.41,"n_real":20,"note":"development seed + short span; superseded by the forward-blind graded campaign","oceanic_false_dm_at_T9":0.0,"oceanic_rho_e_p50":0.999,"seed":20260807,"span_yr":6.0},"expected_direction":"OCEANIC_HOLE_CLOSED_OR_BOUNDED","frozen_v23_sha256":"097face5ea7d7d836a80c67415d39188bf80578f9ab50144462421538eb05254","n_realizations":200,"pre_committed_expectation":{"PRIMARY_oceanic_false_dm_at_T9_lt":0.02,"base_gates":"geophysical false-DM-pass < 0.02 (arm a) AND pure-DM false-demote < 0.10 (arm b)","dm_false_demote_at_T9_lte":0.02,"falsifier":"oceanic false-DM at T9 >= 0.02, OR rho_e distributions overlap, OR a base gate fails -> does NOT seal as an adjudication; reported as a bounded methods reduction. Widening any bound after the graded result is refused.","rho_e_separation":"oceanic arm (d) p05 > pure-DM arm (b) p95 (disjoint)"},"prereg_path":"_active_today/dm_mag_k1_adjudication_prereg_v2_4_AMENDMENT.md","prereg_sha256":"8bfbb2b855c1aaabfc966deabf25d2978bef8efc5bedbd852d04d61b27ed43f9","span_yr":33.0,"supersedes":["v2.3 exploratory battery (NOT sealed as an adjudication: base gates passed but arm (d) measured a purely oceanic K1 returned DM-flavoured in 13/125 = 10.4%, no bar registered; the D9 48x separation was a boolean-injector tautology)"],"unchanged":"D1/D2/D5 gating + controls, geometry_g, _dm_group_ratio_expectation, discovery/verdict vocabulary, base gates, R_cal diagnostic, leakage tripwire, 33-yr span, n=500.","v1_core_sha256":"7db981e1ab3e1ff6ef044f17ad023471a81e66bdaa82e369cdfc62de9b7f1a40"},"method":"graded_injection_campaign_arm_e_spatial_vote_v2_4","n_permutations":500},"threshold":{"direction":"","metric":"oceanic_false_dm_at_T9","value":0.02},"title":"DM Magnetic K1 Adjudication v2.4 -- graded oceanic injector + arm (e) spatial vote (closes the v2.3 oceanic false-DM hole)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"data_source":"Mars InSight XB.ELYSE.02.BHU/BHV/BHW -> Z (2020-01-01..2021-12-31 core) via IRIS/IPGP/EarthScope FDSN; Earth GSN {IU.ANMO,II.BFO,IU.TSUM,IU.CCM}.00.BHZ -> ACC; prepared store manifest 23b964f7b9d0 (SAME as v1.2)","expected_direction":"","feature_set":["gamma2_cross_coherence","S_x_cross_power"],"git_commit_at_registration":"0d1d0ac5ac46e872a71537087234aa5d54cc2b11","hypothesis_id":"20260904_earth_x_mars_cross_planetary_ult_2a4273ed","locked_at":"2026-09-04T00:03:20Z","locked_file_sha256":"05b6194cc014306d5a8931122edb1f0377f5bb5e743df53ce6ebf458d39f1546","notes":"DM Earth x Mars cross-planetary ultralight-DM Compton-line coherence search v1.3 (MEMORY-ONLY amendment; byte-identical to v1.2 01fea7ac, which was sealed, Bitcoin-confirmed at block 965248, ran and passed BOTH gates then OOM-killed in the scan stage at 123.8 GB before scoring). THE ONLY CHANGE vs v1.2: the raw ~56 GB prepared store is freed after precompute_fine_spectra (the scan uses only the ~16 GB fine spectra), dropping peak 124 -> ~72 GB. SCAN_CHUNK_BINS is left untouched because it reorders the phase-scramble null rng draws. PROVEN byte-identical before sealing (validate_v13.py: every science array/candidate/calibration identical, only the wall-clock timer differs). Frozen code scripts/dm_earthmars_coherence_v13.py sha256 4958f6560c8ed12fa9c79b034ffaab4ab8d792c69ef36946bac79d61656b0d93 (git 0d1d0ac5) derived from frozen v1.2 c1d04a58 by four anchor-asserted edits; prereg _active_today/dm_earthmars_coherence_prereg_v1_3_AMENDMENT.md sha256 4d9754a0968ed077f05fec4aa1c7ed40fea674d1a150017ef2d34ba14c138cec. FRESH forward-blind seed = sha256(EARLIEST attesting Bitcoin block hash of THIS anchor), no real bin scored before confirmation. expected_direction=NULL_BOUND.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_line_search"},"reevaluate_at":null,"registered_at":"2026-09-04T00:03:04Z","registered_by":"Mljessop@deepmap-compute","result":{"git_commit_at_test":"d404e7728d11247188c5d154f49064511f367582","interpretation":"disconfirming","metric_value":4.98,"n_observed":0,"notes":"NULL_BOUND as pre-registered. v1.6 out-of-core sealed scan of 0.01-1 Hz, 4605173 bins, K median 1442; max Sidak z 4.98 at 0.483 Hz, BELOW the 5.593 discovery threshold, so 0 candidates. Null calibration alive 1.000, empirical FAR 0.0483 (expected 0.05). Positive controls PASSED (FFT-equivalence + injection recovered/injected 1.000, joint z 157.2), so this is a TESTABLE null, not untestable-as-frozen (cf v1.1). Result = a 95pct exclusion on the cross-planetary per-body DM Compton-line coherence; coupling deferred. Ran to completion on the v1.6 memmap fix (peak ~25 GB, 3.1 h). failed denotes no DM candidate met the discovery rule -- the expected NULL_BOUND.","ots_proof_path":null,"passed":false,"tested_at":"2026-09-05T14:48:00Z","tested_by":"Mljessop@deepmap-compute"},"status":"tested","test":{"extra":{"amendment_v1_3":{"change":"free the raw store after precompute_fine_spectra (its only remaining use is the scalar res_hz, computed first). Peak 124 GB -> ~72 GB (precompute) -> ~56 GB (chunk loop).","edits":["add free_patches=False to scan_grid_fft signature (phase0fft reuses pM0 across passes; untouched)","compute res_hz right after precompute; if free_patches, clear the patch lists (~56 GB freed)","drop the redundant res_hz recompute in the candidate-cluster block","pass free_patches=True at the ONE real scoring scan"],"kind":"MEMORY-ONLY (byte-identical to v1.2; the science is UNCHANGED)","scan_chunk_bins_untouched":"SCAN_CHUNK_BINS is NOT a legal lever: changing it reorders the rng draws in the per-candidate phase-scramble nulls (candidate + calibration bins interleave at chunk granularity), which feed promotion.","seed_rule":"FRESH forward-blind seed = sha256(EARLIEST attesting Bitcoin block hash of THIS anchor); NOT v1.2's now-public block 965248. The byte-identical proof makes the memory change neutral, so a fresh seed is the cleanest provenance.","why":"v1.2 OOMs because the ~56 GB raw prepared store stays resident through the chunked scan while the scan needs only the ~16 GB fine spectra. Measured RSS: flat 56 GB through both gates, then 56 -> 123.8 GB in the scan stage, OOM-killed still rising."},"byte_identical_proof":{"method":"scan_grid_fft on IDENTICAL synthetic input, free_patches=False (v1.2) vs True (v1.3), same scan-rng seed","result":"every science array IDENTICAL (gamma2, z_analytic, K, S_x, meanpow_E, meanpow_M, p_analytic, f_hz via np.array_equal); candidates identical; calibration identical; the ONLY differing summary field is scan_seconds (wall-clock); store freed only when asked (run A kept 16 patches, run B freed to 0)","script":"scripts/validate_v13.py"},"code_git_commit":"0d1d0ac5ac46e872a71537087234aa5d54cc2b11","code_path":"scripts/dm_earthmars_coherence_v13.py","code_sha256":"4958f6560c8ed12fa9c79b034ffaab4ab8d792c69ef36946bac79d61656b0d93","discovery_rule":"candidate iff analytic Sidak z >= 5.593 over N_bins=4.605e6 AND empirical phase-scramble p <= 0.05 with a live null AND single-body HARD rule AND in-run positive control: per-arm recovered/injected in [0.75,1.33] using DIFFERENTIAL recovery (4 phases); else NULL. IDENTICAL to v1.2.","expected_direction":"NULL_BOUND","expected_outcome":"FIRST cross-planet DM line search; a NULL reported as a 95% exclusion on the observable per-body line amplitude (m/s^2) + the cross-coherence bound + the Earth-Mars coincidence veto; coupling DEFERRED.","frozen_config":{"amendment_v1":{"empirical_null_policy":"phase-scramble on every candidate (>= 500 draws) + 10000-bin calibration subsample (200 draws) reported, never promoting","equivalence_bins":600,"equivalence_patches":40,"evaluation":"FFT path (patch_fine_spectrum + cubic interpolation) reproducing patch_amplitude; gated by --stage equivalence in-run","fft_equiv_tol":0.001,"fft_pad":32,"injection_amplitude":"amp_E such that expected gamma2 = 2.0 x gamma2*(K); amp_M = amp_E / ratio","injection_floor":"LOCAL: 60 log-uniform dummy freqs within +/-20% of f_inj","scan_chunk_bins":200000,"scored_population":"prepared store of the frozen core (MARS_BASELINE_START..END), daily-chunked, wind percentile applied over the whole baseline","store_version":"v1","unchanged":["band","df_over_f","n_bins","statistic","bar_sidak_z","single_body_hard","inj_tol","xi","ratio","projection","wind_keep_percentile","seed_rule"]},"band_hz":[0.01,1.0],"bar_sidak_z":5.593,"df_over_f":1e-6,"earth_duty_floor_rule":"drop a station BEFORE scoring if its 2020-2021 duty < the pre-declared floor (frozen rule, not post-hoc)","earth_stations":["IU.ANMO.00.BHZ","II.BFO.00.BHZ","IU.TSUM.00.BHZ","IU.CCM.00.BHZ"],"global_alpha":0.05,"injection_tol":[0.75,1.33],"mars":{"chans":["BHU","BHV","BHW"],"loc":"02","net":"XB","sta":"ELYSE"},"mars_baseline":["2020-01-01T00:00:00","2021-12-31T23:59:59"],"n_bins":4605170.185988092,"null_battery":{"circular_shift":"EXCLUDED","dummy_floor":150,"phase_scramble":500,"time_shift":500},"projection":"Z","ratio_E_over_M":1.8796282637557162,"seal_anchor_id":null,"seed_rule":"science RNG seed = sha256(confirmed OTS Bitcoin block hash of the recorded anchor); forward-blind, set AT SEAL","single_body_rejection":"HARD","wind_keep_percentile":50.0,"xi_elastic":1.0},"prereg_path":"_active_today/dm_earthmars_coherence_prereg_v1_3_AMENDMENT.md","prereg_sha256":"4d9754a0968ed077f05fec4aa1c7ed40fea674d1a150017ef2d34ba14c138cec","scored_population":"the SAME prepared store v1.2 scored: frozen 2020-01-01..2021-12-31 core, 730 day files, 1442 kept, manifest sha256 23b964f7b9d0. No real bin scored by any version (v0/v1.1 halted at their positive control; v1.2 OOM-killed before scoring).","supersedes":["20260903_earth_x_mars_cross_planetary_ult_01fea7ac (v1.2, SEALED + BTC-confirmed block 965248, RAN and passed BOTH gates -- FFT-equivalence + injection positive control 1.000/1.000 -- then OOM-killed in the scan stage at anon-rss 123.8 GB before scoring any real bin; the science is unchanged and v1.3 makes the sealed scan runnable on the 125 GB box)"],"unchanged":"band 0.01-1 Hz, df/f 1e-6, N_bins 4.605e6, gamma2, Sidak z 5.593, INJ_TOL (0.75,1.33), xi, ratio, projection, wind percentile, FFT path + equivalence gate, differential positive control (4 phases), empirical phase-scramble null policy, single-body hard rule, coincidence veto, SCAN_CHUNK_BINS, FFT_PAD, discovery rule, prepared store (manifest 23b964f7b9d0)"},"method":"fft_cross_coherence_sidak_scan_v1_3","n_permutations":500},"threshold":{"direction":"","metric":"sidak_z","value":5.593},"title":"Earth x Mars cross-planetary ultralight-DM Compton-line coherence search v1.3 (memory-only amendment, byte-identical to v1.2 01fea7ac)","withdrawn":null,"_ots_anchored":true,"_status_dir":"tested","_re_eval_date":null},{"atlas_channel":null,"data_source":"USGS ComCat Mww>=5.0 GK-declustered targets; EMSC M>=3.3 EMSC-only (magnitude-aware dedup) as Arm B extra triggers; Indonesia lat -11..6 lon 95..141; catalog 2013-now; train_end 2023-01-01","expected_direction":"positive","feature_set":["armA_triggers_usgs_ge4.5","armB_plus_emsc_only_ge3.3","mainshock_only_dTIMING"],"git_commit_at_registration":"41ced2020df964c81348576a9c226ffd34162a52","hypothesis_id":"20260908_emsc_low_mc_extra_trigger_admiss_919d04af","locked_at":"2026-09-08T16:30:43Z","locked_file_sha256":"ed91e5783282560faa6e9502c065615522abf9a43abe242ea353f7c42a7529b0","notes":"SHADOW v0-core, RETROSPECTIVE. Imports sealed earthquake.point_forecast primitives READ-ONLY; never mutates MC=4.5 or the sealed +0.94 OTS. Holds MC=4.5 in BOTH arms (F4: MC/b/ETAS_A are absorbed by the a_tr/a_bg calibration MLE, so the ONLY lever is trigger admission). Arm A triggers = USGS M>=4.5. Arm B = + EMSC-only M>=3.3 (magnitude-aware dedup |dM|<=0.75 / 30 km / 30 s). Shared ONE frozen declustered smoothed-seismicity background (low-M events act ONLY as triggers, never as background rate) and ONE shared USGS-authority GK-declustered MAINSHOCK target list; both arms N-calibrated to the same train targets. Decision metric = paired block-bootstrap of mainshock-only dTIMING = mean_mains[(lnc_B-lnn_B)-(lnc_A-lnn_A)]. KILL CRITERIA (ANDed; any miss -> UNRESOLVED or NULL, never re-tuned): (1) paired dTIMING 95pct CI lower bound > 0; (2) survives small-events time-shuffle (real-minus-shuffled CI > 0); (3) mainshock-specific positive control fires; (4) n_seq >= 30 hard floor else UNRESOLVED. EXPECTED RESULT stated IN ADVANCE: mainshock-only dTIMING SMALL or NULL, likely +0.05..0.20 IG/mainshock and probably UNRESOLVED at one region; the real gain is aftershock-only. HONEST FRAMING: the OOS window 2023-2026 is already past, so this is a METHOD-FREEZE seal (config/controls/kill/expected frozen before the estimator first runs on real Indonesia data, preventing tuning to the answer), NOT forward-blind; the result is a RETROSPECTIVE UPPER BOUND and forward confirmation on the shadow ledger is owed before any operational claim (the l=4 precedent). DEFERRED CONTROLS not in v0-core: F5 as-known-at-t availability (a look-ahead-leak guard; a MAINSHOCK_TIMING_GAIN verdict from v0-core is NOT a claim -- it triggers building F5 plus magnitude homogenization plus admission ladder plus ETAS_B sweep plus era-split before any statement), M3 magnitude homogenization to Mw (currently identity) plus 0.2 sensitivity, M4 within-catalog admission ladder, M2 ETAS_B sweep, F6 era-split OOS flatness. LANGUAGE FROZEN: baseline = fitted causal ETAS; unit = IG per M>=5 mainshock over that baseline; NEVER we-forecast-mainshocks, never an accuracy or hit-rate; coverage != skill; IG-over-baseline != an alarm. The 41pct EMSC-only-foreshock census (245 Indonesian M>=5 mainshocks, ~2pct PPV) is the publishable ceiling independent of whether dTIMING moves.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":5.0,"type":"earthquake"},"reevaluate_at":"2026-12-31","registered_at":"2026-09-08T16:30:12Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"5f23613443a92278c57897cca417a44575ef1c81","interpretation":"disconfirming","metric_value":0.0071,"n_observed":341,"notes":"VERDICT UNRESOLVED. Pre-registered PASS = paired dTIMING CI95 lower > 0; observed dTIMING = +0.0071 IG per mainshock, CI95 [-0.0443, +0.0702] spans zero -> criterion NOT met, recorded --failed. UNDERPOWERED not refuted: MDE band 0.0572 > |effect| 0.0071; n_mainshocks=341, n_usgs=12323, n_emsc_only_admitted=46379 of 58669 EMSC, b=1.273. Controls: positive control FIRES (an M6.2 mainshock with 4 EMSC-only foreshocks raised Arm B pre-event lambda 440x over Arm A -- mechanism real at individual events); time-shuffle gap +0.1514 [0.1034, 0.2483] excludes 0 (genuine WHEN-structure in the EMSC-only triggers). But the NET mainshock-only timing gain is null-in-noise -- exactly the pre-registered expectation (small/UNRESOLVED; the real gain is aftershock-only). Retrospective upper bound: |dTIMING| < ~0.07 IG per mainshock at 95pct. The completeness lever does NOT add measurable MAINSHOCK-timing skill in Indonesia at this scale, consistent with the census ~2pct PPV. v0-core CONCLUSIVE: a null/UNRESOLVED needs none of the deferred controls (F5 etc.). Publishable content = the 41pct EMSC-only-foreshock census ceiling, unaffected. Box run PID 3370746, settled magnitudes (live forward scores modestly lower).","ots_proof_path":null,"passed":false,"tested_at":"2026-09-08T16:43:34Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{"MC_both_arms":4.5,"R_km":100,"WBG_days":3650,"WLOOK_days":365,"W_days":7,"admission_floor_armB":3.3,"catalog_start_year":2013,"controls_deferred":["F5_as_known_at_t_availability","M3_magnitude_homogenization_to_Mw_plus_0.2_sensitivity","M4_within_catalog_admission_ladder","M2_ETAS_B_sweep","F6_era_split_OOS_flatness"],"controls_implemented":["mc45_tie_arms_identical","small_events_time_shuffle_real_minus_shuffled_ci_gt_0","mainshock_specific_positive_control","hard_sequence_floor_else_UNRESOLVED"],"min_seq":30,"module":"earthquake/mainshock_lowmc_shadow.py","n_boot":1000,"region_box":[-11,6,95,141],"train_end":"2023-01-01"},"method":"paired_bootstrap_mainshock_only_dTIMING_causal_ETAS_armA_vs_armB","n_permutations":1000},"threshold":{"direction":"positive","metric":"paired_dTIMING_ci95_lower_bound","value":0.0},"title":"EMSC low-Mc extra-trigger admission adds mainshock-only ETAS timing IG in Indonesia (SHADOW v0-core)","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null},{"atlas_channel":null,"data_source":"Utah FORGE: GDR MT EDIs (SW Utah 1578) + Schlumberger array-induction LAS logs (GDR)","expected_direction":"negative","feature_set":[],"git_commit_at_registration":"f5b08181be025185f26887560f5d1b68d29694bd","hypothesis_id":"20260913_mt_vs_borehole_reference_curve_i_bc29754c","locked_at":"2026-09-13T15:56:18Z","locked_file_sha256":"8a505f81610f9cf96ba0d07749bdcc4ffc74735e868f4b367ed3608582a18f1f","notes":"REFERENCE CURVE DECLARED ON PHYSICS, NOT FIT. AORT is the array-induction tool's own inversion for TRUE formation resistivity, which is the quantity MT attempts to recover; AF/AT/AO are raw array outputs at differing vertical resolutions and are not Rt. PASS = ensemble median factor <= 3.0. BLINDNESS, STATED HONESTLY: well 58-32 is FULLY BLIND - its AORT curve exists but was never computed, because AORT was absent from the tool's curve list when 58-32 was first run. Well 78B-32 is PARTIALLY INFORMED - a SINGLE-SITE AORT value (1.68x) was seen during an invasion check, though the ensemble was not; it is therefore graded separately and must not be pooled with 58-32. PRIOR EXPECTATION, recorded so the test can falsify me: AT90 gave 1.74x and AO90 2.85x at 58-32, so I expect 58-32 AORT near 1.5-3x, i.e. I predict PASS. If it lands above 3x my physics argument for AORT is wrong. Wells 16A(78)-32 and 56-32 carry only TBIT curves, no AORT, and are out of scope. Method frozen: Bostick transform of the impedance determinant, conductance-preserving (harmonic) log upscaling, half-window max(0.25z,50m), depths 100m to log max.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"geophysical_agreement"},"reevaluate_at":null,"registered_at":"2026-09-13T15:56:14Z","registered_by":"Mljessop@Mike","result":{"git_commit_at_test":"e3d65b35e4394638fcbc01de308cf5db771d4068","interpretation":"disconfirming","metric_value":3.76,"n_observed":12,"notes":"FAILED on the BLIND well. 58-32 (fully blind, AORT never previously computed): ensemble median 3.76x, spread 2.42-5.31x over 12 MT sites -- ABOVE the declared 3.0x PASS bar. My recorded prior expectation was 1.5-3.0x with a predicted PASS; that prediction is FALSIFIED. 78B-32 (PARTIALLY INFORMED, one single-site AORT value had been seen): ensemble median 1.59x, spread 1.17-2.90x over 10 sites -- would PASS, but it is graded separately and is NOT pooled, per the lock. WHAT THE PHYSICS ARGUMENT GOT RIGHT: AORT is materially the better reference, exactly as declared -- 58-32 improves 4.77x (AF90) to 3.76x, and 78B-32 improves 9.02x to 1.59x. So the direction was correct and the magnitude was not. WHY THIS MATTERS PROCEDURALLY: without the locked bar and the recorded prediction, 'AORT gives 3.76x and 1.59x, far better than 4.77-9.02x on AF90' would have been reported as a success. Against the declared threshold the blind well fails. STANDING CONSEQUENCE UNCHANGED: MT does not clear a 3x agreement bar against borehole resistivity on a blind test, so the MT -> Archie -> salinity leg remains unsupported at FORGE.","ots_proof_path":null,"passed":false,"tested_at":"2026-09-13T15:56:39Z","tested_by":"Mljessop@Mike"},"status":"tested","test":{"extra":{},"method":"ensemble median |log10(MT_Bostick / AORT_conductance_upscaled)| over the >=3 nearest MT sites","n_permutations":null},"threshold":{"direction":"negative","metric":"ensemble_median_discrepancy_factor","value":3.0},"title":"MT-vs-borehole reference curve is AORT (array-induction true Rt), and MT agrees within 3x","withdrawn":null,"_ots_anchored":false,"_status_dir":"tested","_re_eval_date":null}],"withdrawn":[{"data_source":"data/prediction_ledger.db","expected_direction":"negative","feature_set":["model_name=cross_physics_earthquake_alert_cycle","region in {iran","iceland","myanmar_bangladesh","azores","new_zealand","newberry","trans_mexico","ukraine}"],"git_commit_at_registration":"db7fb91420b171a6a884bc0d97ffe8da1f8ea20e","hypothesis_id":"20260504_cross_physics_earthquake_alert_c_ff460a64","locked_at":"2026-05-04T19:50:40Z","locked_file_sha256":"1835cedbf6a2566decf6a34ab5fbc2a2a4fdd06d5fb9b2bb5cad571eb57b925f","notes":"Surfaced by inverse-hit-rate tomography on AX41 prod ledger 2026-05-04. Model is HEALTHY globally (57.5% lifetime, 69.4% in 30d) but emits 67 unmatched preds across 8 specific regions in 90d -- 0% match rate in each. Hypothesis: failure persists without intervention through 2026-08-04. If confirmed, justifies investigation of model's per-region magnitude expectations or per-region disable list. Test: re-run scripts/inverse_hit_rate_tomography.py at lookback 90d on 2026-08-04, check that pooled hit rate in the 8 named regions stays under 20% AND global model hit rate stays above 50%.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"earthquake"},"registered_at":"2026-05-04T19:50:03Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"per_region_match_rate_vs_global","n_permutations":null},"threshold":{"direction":"negative","metric":"regional_hit_rate","value":0.2},"title":"cross_physics_earthquake_alert_cycle hit rate stays under 20% in 8 quiet regions (iran/iceland/myanmar_bangladesh/azores/new_zealand/newberry/trans_mexico/ukraine) over next 90d, while global rate stays above 50%","withdrawn":{"at":"2026-05-04T20:02:00Z","by":"Mljessop@Mike","from_status":"locked","reason":"Diagnosis revealed cross_physics_earthquake_alert_cycle has been silent since 2026-04-13 (last prediction 21 days ago). The model is NOT actively firing predictions, so the forward-test 'failure persists at <20% over 90d' is untestable -- the rate is undefined when the model emits no predictions. Last-30-day analysis showed 4 of the 8 named regions (iran, myanmar_bangladesh, new_zealand, trans_mexico) are now well-covered by tensor_eq_per_region / earthquake_magnitude_predictor / others. The 4 genuine remaining blind spots (iceland, azores, newberry, ukraine) are low-magnitude volcanic-rift / stable-interior zones where the issue is expected_mag_min floor too high. Re-registering as a per-region magnitude floor intervention hypothesis."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-08-04"},{"data_source":"data/prediction_ledger.db","expected_direction":"positive","feature_set":["cuf_rs_amplification_factor","cuf_combined_rate_per_day_30d","cuf_max_mag_prob_M4_30d","cuf_uplift_rate_mm_yr","cuf_etas_branching_ratio"],"git_commit_at_registration":"e33f626750795375d472e52eba0961411dbf5265","hypothesis_id":"20260506_rs_etas_caldera_forecaster_high__eb9944b0","locked_at":"2026-05-06T14:57:50Z","locked_file_sha256":"de1ff5d069a688708a51bbc39c55258e464fb3c9d0fcda6dd7e7ae90bdf22e10","notes":"FORWARD test of the deformation-driven RS+ETAS caldera forecaster (Hainzl/Dahm/Tramelli 2026 operationalization, scaffolded May 6 2026). Predictor logs HIGH/CRITICAL ledger entries at 4 watch-list calderas (Campi Flegrei, Yellowstone, Long Valley, Aira/Sakurajima) when rs_amplification_factor >= 2.5 (HIGH) or >= 5.0 (CRITICAL). Pass criterion: pooled match_rate across HIGH+CRITICAL alerts >= 0.50 over the 90-day forward window 2026-05-06 -> 2026-08-04, evaluated against existing PredictionLedger auto-matchers (USGS for Yellowstone/Long Valley/Aira, INGV for Campi Flegrei) with the standard per-model dynamic radius (Phase 65u-MATCHER) within prediction's valid_until window. Per-caldera Mc thresholds: Campi Flegrei 0.5, Yellowstone 2.0, Long Valley 2.0, Aira 1.0. parameters_pending_refit=True at lock time (literature priors for A_sigma_mpa, tau_a_days, etas_K/c/p/alpha, b-value); confidence capped at 0.70 in ledger. Null baseline: per-caldera Poisson at base_rate_per_day prior over the same 30-day window yields ~25% pooled match rate. Re-eval 2026-08-04. Disconfirming result is hypothesis-failing; <30% would indicate the literature priors are wrong direction. WITHDRAW criterion: if fewer than 5 HIGH+CRITICAL alerts fire across all 4 calderas over the 90-day window the test is underpowered; mark withdrawn and re-register with looser thresholds.","ots_proof_path":null,"outcome":{"lookahead_hours":720.0,"min_magnitude":0.5,"type":"earthquake"},"registered_at":"2026-05-06T14:57:44Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"match_rate_per_caldera","n_permutations":null},"threshold":{"direction":"positive","metric":"match_rate","value":0.5},"title":"RS+ETAS caldera forecaster HIGH/CRITICAL alerts hit at >=50% over 90-day forward test","withdrawn":{"at":"2026-07-18T16:28:03Z","by":"Mljessop@Mike","from_status":"locked","reason":"INVALID TEST -- the instrument was broken for the entire forward window. fetch_uplift_rate_mm_yr pointed at the retired NGL IGS14 path (404 since the Jun 2026 IGS20 migration; this module was the one of six missed in commit 5ed76bf8), so every uplift fetch silently fell back to the 50 mm/yr literature prior. That froze uplift_mm_yr=50.0 and rs_amp=8.2652 identically across all 66 HIGH/CRITICAL rows -- above the CRITICAL threshold of 5.0 by pure arithmetic, so the forecaster emitted CRITICAL regardless of data. The alerts under test were therefore NOT produced by the deformation-driven RS+ETAS mechanism this hypothesis names. Live GPS (fixed 2026-07-18) gives Campi Flegrei -10.61 mm/yr (subsiding) and rs_amp 0.6388 -> NOMINAL, a 13x overstatement in the alarm-manufacturing direction. The observed ~91-95 pct match rate is further confounded by an always-on alarm against continuous bradyseismic unrest credited by an INGV matcher at minmag=1.5 (requiring M>=3.0 drops it to ~30 pct, M>=4.0 to ~8 pct). Neither passed nor failed: untestable as run. Superseded by a v2 forward test on live deformation with a magnitude floor and a coverage-null bar. Record preserved per Section 7 of the methodological-nulls bundle."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-05-06"},{"data_source":"data/turbidite_paleoseism.db + data/prediction_ledger.db","expected_direction":"positive","feature_set":["pmf_paleoseismic_prior_per_year","pmf_paleoseismic_prior_n_paired","pmf_paleoseismic_prior_window_yr","pmf_paleoseismic_prior_db_backed","pmf_posterior_alert_prob_7d"],"git_commit_at_registration":"6a476ea98d04a5bf4c740cc0437cea8180bbdad0","hypothesis_id":"20260510_goldfinger_option_3_paired_megaf_9e4c90a0","locked_at":"2026-05-10T18:24:18Z","locked_file_sha256":"b23fe7679545ce870362b4f52154aa2dfb0e84921bf8d01aecdac7f55f0b780f","notes":"FORWARD test of Goldfinger Option #3: turbidite paleoseism DB ingest replaces literature 0.2%/yr prior with database-derived n=3-in-1500yr Poisson prior (per_year=0.002, 95%CL [0.0004, 0.0058]). Hypothesis: paired-megafault-detected predictions (model_name contains paired_megafault OR pmf_ OR region contains cascadia_san_andreas) over the 90-day forward window 2026-05-10 -> 2026-08-08 achieve match_rate >= 2x the unpaired baseline match_rate computed over same window across all earthquake.predictor models. Pass criterion: paired_match_rate / unpaired_match_rate >= 2.0 with at least 5 paired predictions in the window. WITHDRAW criterion: <5 paired predictions in window = underpowered. NULL outcome (ratio < 2x) is publishable -- would show paleoseismic prior provides no operational lift on top of metadata-only base score. Re-eval 2026-08-08.","ots_proof_path":null,"outcome":{"lookahead_hours":168.0,"min_magnitude":4.5,"type":"earthquake"},"registered_at":"2026-05-10T18:24:13Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"match_rate_ratio_paired_vs_unpaired","n_permutations":null},"threshold":{"direction":"positive","metric":"match_rate_ratio","value":2.0},"title":"Goldfinger Option #3 paired-megafault paleoseismic prior boost lifts paired-prediction match rate by >=2x over unpaired baseline","withdrawn":{"at":"2026-05-21T17:32:33Z","by":"Mljessop@Mike","from_status":"locked","reason":"Predictor companion not shipped at lock time -- the cross_physics/paired_megafault_monitor.py module was a tensor-feature builder + run_demo dumper only, with zero log_prediction() call sites and no cron entry.  Aug 8 evaluator on 2026-05-21 found paired_n=0 over 17 days of supposed forward window.  Re-registered as v2 (20260521_*) once cross_physics/paired_megafault_predictor.py + GATES_BY_MODEL entry + cron land in PR.  Underlying Goldfinger 2025 science unchanged; this is a tooling completion."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-05-10"},{"data_source":"Utah DOGM WIS API + cross_physics.og_convergence_monitor","expected_direction":"negative","feature_set":["ogc_composite_score","ogc_score_structural_trap","ogc_nearest_fault_km"],"git_commit_at_registration":"e182c4a2d0e4b9476bfeedd18746df4af9db0c8a","hypothesis_id":"20260511_ogc_top_decile_cells_in_sevier_d_4eb375da","locked_at":"2026-05-11T02:30:32Z","locked_file_sha256":"f5396c7a6f6ace902d90df0d0ce8c1ced4083356d88a6067394567b8a0595809","notes":"Predictive Geology Phase 1 of 3 (companion to papers/predictive_geology_validation_design.md §4 H1). PRE-REGISTRATION REQUIRED before any DOGM ingest. Conditional on a Sevier Desert / Pavant / Black Rock / Beaver County cell containing >=1 historical dry wildcat (PA disposition, TD > 1500 m), the cell's OGC composite_score has median <= 25 (DRY class). Random baseline for these basins is median ~40-45. Predicting Mann-Whitney U test p<0.001 between wells-present and wells-absent cells. N >= 500 wells-present cells expected. Power >0.95 at effect size 0.3 sigma. Failure interpretation: OGC scoring uniformly across drilled and undrilled cells means model has no spatial discrimination. Re-eval target: 2026-08-31 after DOGM ingest + OGC computation complete.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"oil_gas_well_disposition"},"registered_at":"2026-05-11T02:29:47Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"mann_whitney_u_test_on_dry_vs_random_cells","n_permutations":null},"threshold":{"direction":"negative","metric":"median_ogc_composite_for_dry_wildcat_cells","value":25.0},"title":"OGC top-decile cells in Sevier Desert cluster on basin-margin trap fairway, NOT in basin interior (Predictive Geology H1)","withdrawn":{"at":"2026-09-14T16:47:23Z","by":"Mljessop@Mike","from_status":"locked","reason":"UNTESTABLE AS SEALED, withdrawn 2026-09-14 -- not disconfirmed, never runnable. (1) POPULATION: the protocol plans 'N >= 500 wells-present cells, power >0.95'; the Sevier Desert/Pavant/Black Rock/Beaver basin holds 132 deep wells inside the sealed bbox (ut_wells_features_v2.csv), so <=132 such cells can exist. The seal's own population estimate was ~5x optimistic because the basin is under-drilled, which is why it is a frontier play. (2) PREMISE CONTRADICTED IN BOTH DIRECTIONS: the seal assumes 'random baseline for these basins is median ~40-45' and predicts dry-wildcat cells at median <=25. Measured on a uniform grid over the same bbox the baseline is far below 25, so nothing can fail the threshold; and the companion H3 (865aa86c) measured dry wells scoring HIGHER than producers (rank-biserial -0.4088, 95% CI [-0.6366,-0.1587]), i.e. the predicted direction is reversed. (3) INSTRUMENT: the scorer named in git_commit_at_registration (e182c4a2) returns ogc_data_real=0 with 0/5 axes at every point in the target area, so the sealed metric cannot be computed as sealed at all. Re-issue forward only with a population that exists and a baseline measured rather than assumed."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-08-31"},{"data_source":"data/retro_cache/iris_mustang.db + data/retro_tensor.db","expected_direction":"two_sided","feature_set":["microseism_primary_band_6s","microseism_secondary_band_3s"],"git_commit_at_registration":"67881aee2e934e69d2254e5921f9abffadbdb53b","hypothesis_id":"20260514_phase_5_microseism_universal_pre_41a196bf","locked_at":"2026-05-14T03:45:06Z","locked_file_sha256":"3e5b1b45e6afec36a6798a8d963b33091ca8c9d5f56508bc070c9bb35fa5a571","notes":"ADDENDUM to parent hypothesis 20260504_microseism_universal_precursor_c_99f4c2c7 (Bitcoin-anchored 2026-05-04). The Time-A v3 time-reversed stacking experiment (2026-05-13, N=7,120 M5.5+ events from retro tensor 2005-2018) reproduces the Phase 5 +19sigma at lag -10h finding (recast as z=-8.07 in window-relative normalization) but ALSO reveals that the strongest microseism deviation is at lag +134h post-event (z=-34.5), and the signal persists across the full pre/post 168h window. Independent confirmation via Time-D entropy classifier (2026-05-14): microseism_primary_band_6s is the SINGLE MOST-IRREVERSIBLE FEATURE in the entire 112-feature retro tensor (R^2 = -0.216), and microseism_secondary_band_3s is the strongest non-solar irreversible feature. These two independent methods support a revised Phase 5 narrative: not a discrete 10-hour precursor, but the LEADING EDGE of a multi-day stress-state cycle that straddles rupture and extends through ~+134h (5.6 days) post-event. PASS CRITERION (1) M5.5+ events 2026-05-13 to 2026-11-13 show median z >= 5 at lag +120h post-event; (2) post-event integrated z over [+1h, +168h] >= 2x pre-event integrated z over [-168h, -1h] across N >= 50 events. Methodology files: scripts/eq_precursor_stacked_epoch_time_reversed.py, scripts/time_reversed_entropy_classifier.py. Tonight's result JSONs: reports/science/eq_precursor_epoch_time_reversed_20260514T033322Z.json, reports/science/time_reversed_entropy_20260514T033934Z.json.","ots_proof_path":null,"outcome":{"lookahead_hours":168.0,"min_magnitude":5.5,"type":"earthquake"},"registered_at":"2026-05-14T03:45:00Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{"asymmetry_test":"post_minus_pre_integrated_z","key_lags_h":[-134,-10,0,10,134],"min_asymmetry_ratio":2.0,"min_post_peak_z":5.0,"window_h":[-168,168]},"method":"stacked_epoch_z_extended_window","n_permutations":10000},"threshold":{"direction":"two_sided","metric":"post_event_peak_z","value":5.0},"title":"Phase 5 microseism universal precursor is the leading edge of a 5-day stress-state cycle straddling rupture, with peak signal at +134h post-event NOT lag -10h","withdrawn":{"at":"2026-08-11T17:17:36Z","by":"Mljessop@Ubuntu-2404-noble-amd64-base","from_status":"locked","reason":"Mis-filed: never actually withdrawn despite being superseded. Its own successor 20260514_phase_5_microseism_universal_pre_dad7e9e7 (CORRECTED ADDENDUM, locked the same day) states it \"supersedes withdrawn hypothesis 20260514_phase_5_microseism_universal_pre_41a196bf (which contained a coordinate bug)\" -- but this entry remained status=locked, withdrawn=None, and absent from hypotheses/withdrawn/ for three months, so it was counted as a LIVE forward test the whole time. Withdrawn 2026-08-11 for the originally-stated reason (coordinate bug in the Time-A v3 time-reversed stacking experiment); the corrected replacement dad7e9e7 stands. Preserved, not deleted -- the record of the error is the point."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-11-13"},{"data_source":"Sentinel-2 L2A via Microsoft Planetary Computer + IRIS FDSN 3-component + IRIS FDSN microseism","expected_direction":"positive","feature_set":["phc_kaolinite_score","phc_vegetation_chlorosis_score","phc_iron_oxide_score","phc_pol_em_deflection_score","phc_vertical_splitting_score","phc_composite_score_0_1"],"git_commit_at_registration":"8e6cd572aecd4fb7f8bbcafaf83dbd770218f931","hypothesis_id":"20260521_passive_hc_composite_v0_1_discri_94855e4f","locked_at":"2026-05-21T22:56:37Z","locked_file_sha256":"e996e596959088d6cf5042d6a29b4335d2b184ba41ef6af3f97df0c822d85e8b","notes":"PRE-REGISTERED FORWARD TEST.  Re-run scripts/backtest_passive_hc_composite.py on AX41 (Helsinki, 37.27.55.232) where polarization_em_monitor + spectral_splitting_monitor have full FDSN access (Windows-local test on 2026-05-21 only got 3 of 5 channels = sat_hsi only).  Pass criteria: (1) producer_mean_minus_control_mean >= 0.20 across the 8-site producer/control set (3 Hingeline producers: Covenant + Providence + Anschutz Ranch; 5 Basin-Range / off-fairway controls: Bonneville Salt Flats, GSL center, SLC urban, Sevier Lake basin axis, Park City Cu/Au district), AND (2) recall >= 60% at threshold 0.30 across the 3 producers.  RE-EVAL DATE: 2026-08-21 (90 days).  v0.1 Windows-partial baseline: delta=+0.107, recall=33%, FP=20% (effectively null).  Two dense-array candidate sites (Baker s0031 + Seeker s0002) also captured for descriptive context, not part of pass criteria.  If hypothesis tests NULL on AX41 full 5-channel: v0.2 with 50km radius regional anomaly normalization on sat_hsi matched-filter scores is the next iteration.  References: Moulton & Pinnell 2005 OGJ; Chidsey et al. 2007 UGA Pub 36; Floyd_Moulton-2.pdf (100% pre-drill dry-hole prediction by Radiometric Plus across 11 Utah Overthrust wells).  Backtest output JSON in data/backtest_passive_hc_composite_*.json.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"oil_gas_surface_signature"},"registered_at":"2026-05-21T22:56:16Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"10_site_hingeline_backtest_v0_1","n_permutations":null},"threshold":{"direction":"positive","metric":"producer_mean_minus_control_mean","value":0.2},"title":"Passive HC composite v0.1 discriminates Central Utah Thrust Belt producers from Basin-Range controls on AX41 5-channel re-run","withdrawn":{"at":"2026-05-21T23:22:45Z","by":"Mljessop@Mike","from_status":"locked","reason":"UNDERPOWERED. AX41 backtest 2026-05-21T23:09Z revealed mss (spectral_splitting_monitor) returned data_real=0 at all 10 sites despite the hypothesis explicitly requiring full FDSN access for both pol_em + mss. Only 4 of 5 channels actually fired. The headline +0.259 producer-control mean delta passes the locked +0.20 threshold, but the result is fragile: Anschutz Ranch's 0.955 composite is a low-n geomean artifact (3 of 5 channels = sat_hsi saturated, both EM channels dead). Excluding Anschutz: producer mean 0.46, control mean 0.37, delta +0.09 -- below threshold. False-positive rate 60% at threshold 0.30 (Sevier Lake basin axis + Park City porphyry + GSL center all fire) confirms the v0.1 composite cannot discriminate HC microseepage kaolinite from natural background kaolinite. Re-registering as v0.2 with 50km radius regional anomaly normalization on sat_hsi matched-filter scores (the canonical Moulton Radiometric Plus approach to background subtraction). Withdrawn locked v0.1 JSON (file sha256 2b6cb590c961a90a860f47491f514d978f20ea923642fb63971b4c159f6ed114) + .ots Bitcoin anchor preserved in hypotheses/withdrawn/ per registry convention. Per CLAUDE.md Phase 65u-RIGOR."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-08-21"},{"data_source":"data/prediction_ledger.db + cross_physics/paired_megafault_predictor.py","expected_direction":"positive","feature_set":["pmf_joint_score","pmf_posterior_alert_prob_7d","pmf_both_elevated","pmf_n_pairs_joint_elevated"],"git_commit_at_registration":"05b1f678362e876cd2b622fa22419dd8029ffa92","hypothesis_id":"20260521_v2_paired_megafault_paleoseismic_70095bae","locked_at":"2026-05-21T17:33:07Z","locked_file_sha256":"1ecf965f266edbbf27a8f7ee3b66188c1a68c8a143f48ead5b105364bd6f6d8d","notes":"Replaces withdrawn v1 (20260510_goldfinger_option_3_paired_megaf_9e4c90a0).  V1 was withdrawn because the monitor module shipped without a log_prediction call site, making paired_n>=5 unreachable in window.  V2 registered after cross_physics/paired_megafault_predictor.py + GATES_BY_MODEL['paired_megafault'] entry + AX41 cron land.  Same scientific test (Goldfinger 2025 paired-megafault paleoseismic prior, n=3-in-1500yr giving per_year=0.002 with 95%CL [0.0004, 0.0058]).  Filter: model_name = 'paired_megafault' (the new predictor uses this verbatim).  Pass: paired_match_rate / unpaired_match_rate >= 2.0 AND paired_n >= 5 over the forward window.  Withdraw if paired_n < 5 -- meaning either joint-elevation events did not occur (a legitimate scientific null) OR the predictor's elevation gate is too strict (operational issue worth a v3).","ots_proof_path":null,"outcome":{"lookahead_hours":168.0,"min_magnitude":4.5,"type":"earthquake"},"reevaluate_at":"2026-08-19","registered_at":"2026-05-21T17:32:53Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"match_rate_ratio_paired_vs_unpaired","n_permutations":null},"threshold":{"direction":"positive","metric":"match_rate_ratio","value":2.0},"title":"v2: paired-megafault paleoseismic prior boost (post-predictor) -- paired predictions match rate >=2x unpaired baseline","withdrawn":{"at":"2026-08-22T18:24:25Z","by":"Mljessop@Mike","from_status":"locked","reason":"Sealed withdraw-if clause: paired_n = 0 < 5 over the forward window (2026-05-21 -> 2026-08-19). The predictor ran HOURLY the whole window (cron alive, log verified 2026-08-22: 'scanned 1 pairs, 0 joint-elevated, 0 predictions logged, dry_run=False' every run; ledger hash chain valid) -- so this is branch 1 of the clause, a legitimate scientific null (no joint-elevation events occurred), NOT v1's dead-cron failure. A v3 with a recalibrated elevation gate is optional, not owed."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":null},{"data_source":"data/igets_sg_archive/* (24 stations, 2009-2026 EOST L2)","expected_direction":"positive","feature_set":["peak_amplitude_uhz","peak_p_value","n_stations_passing","phase_coherence_score","coherent_stack_snr"],"git_commit_at_registration":"407ad73319a5f147349558b3ec7ce1ae48c5e934","hypothesis_id":"20260526_mars_a2_0s0_fundamental_whole_ea_1b0bbb26","locked_at":"2026-05-26T03:54:24Z","locked_file_sha256":"567a0d122b136b40a6467721c480d058d0a57db7dce65518a7f1fcaa84963188","notes":"MARS-A2 (Mars-shot codename) is the first of 6 A-shots in the Six Science Mars-Shots Phased Program brief (DMAI-SCI-001 v1.0, shipped 2026-05-26, data/private/decks/dmai_six_science_mars_shots_brief_v1.{md,pptx,pdf}). MARS-A2 tests the BROADER 0S0 ambient mode confirmation question -- does the canonical 0S0 fundamental radial mode at 813.92 uHz (period 20.5 min, Lamb-Love 1882 theoretical, Park & Gilbert 1986 first observational confirmation) appear at p<0.001 in our 24-station IGETS SG coherent stack across the 2009-2026 archive after ICET-corrected per-constituent harmonic subtraction (M2/S2/N2) and Hinderer 1995 atmospheric loading correction. This is DISTINCT from Phase D.1-SPRINT v4 hypothesis 20260521_wahr_1981_l_4_admixture_v4_forwa_a6ae7dc5, which tests the NARROWER l=4 admixture calibration sub-question on the same IGETS infrastructure. Both share preprocessing dependencies (OSO Chalmers BLQ + Hinderer v2 atmospheric admittance) so MARS-A2 kickoff is gated on the same Mike-action: OSO Chalmers FES2014b BLQ files arrive via Mike-action send-ready memo at reports/phase_d1_a6_item4_oso_chalmers_send_ready_memo.md (~10 min copy-paste + ~24h response time). Re-evaluation calendar: 90 days from kickoff (NOT registration). Kickoff date is set when OSO BLQ files arrive and pipeline fires; calendar entry will be updated then. DISPOSITION RULES: PASS if peak amplitude at 813.92 +/- 1 uHz survives p<0.001 (50000-bootstrap) across >=5 stations AND cross-station phase coherence verifies (phase_coherence_score > 0.7); methods-clean NULL if peak not detected at this significance after full ICET + Hinderer preprocessing. HONEST PRIOR: Detection probable. Phase D.1-SPRINT v4 IGETS SG infrastructure is already producing per-station spectra showing the canonical 0S0 line. Park & Gilbert 1986 and ~40 years of subsequent literature have detected 0S0 at confidence levels much higher than p<0.001 in coherent stacks of fewer stations than our 24. A confirming result is the expected outcome and would be the first MARS-shot detection confirmation. A null would indicate a calibration/preprocessing pathology in our pipeline and would require investigation before any other MARS-shot can claim detection.","ots_proof_path":null,"outcome":{"lookahead_hours":2160.0,"min_magnitude":null,"type":"free_oscillation_detection"},"reevaluate_at":"2026-08-24","registered_at":"2026-05-26T03:54:00Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{"bootstrap_method":"block_bootstrap_n50000","frequency_tolerance_uhz":1.0,"mode":"0S0","period_minutes":20.5,"preprocessing":"ICET_corrected_plus_M2_S2_N2_harmonic_subtraction_plus_Hinderer_1995_atmospheric_loading_correction","stack_method":"complex_coherent_stack_across_stations","target_frequency_uhz":813.92},"method":"hanning_window_fft_coherent_stack","n_permutations":50000},"threshold":{"direction":"positive","metric":"n_stations_passing_p_lt_0p001","value":5.0},"title":"MARS-A2: 0S0 fundamental whole-Earth radial mode at ~814 µHz period 20.5 min detected at p<0.001 across >=5 of 25 IGETS Superconducting Gravimeter stations using Hanning-window FFT coherent stacking","withdrawn":{"at":"2026-08-22T18:34:39Z","by":"Mljessop@Mike","from_status":"locked","reason":"UNTESTABLE AS REGISTERED at due date (2026-08-24), not disconfirmed -- June-1 precedent. The registered instrument (2009-2026 multi-window coherent stack + 50,000-draw block bootstrap per station, p<0.001, >=5 of 24-25 stations) was never built; it is D.1-SPRINT-scale engineering. The v0.1 proxy script DID run on real data 2026-08-22 (24/24 stations, EOST L2 from data/igets -- the lockfile data_source named a stale path igets_sg_archive -- 60-day most-recent windows, n=86400 min each, peak SNR max 1.26, 0 stations >= SNR 3): recorded as a PRELIMINARY 60-day-window constraint only, explicitly NOT the registered test and NOT the grade. Reissued forward as v2 with an instrument-readiness gate."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-08-24"},{"data_source":"earth_state_tensor.db + prediction_features (resolved earthquake predictions logged 2026-06-01 onward, with the 44 new stacking features present)","expected_direction":"positive","feature_set":["sc_microseism_coherence_index","sc_gnss_coherence_index","sc_cosmic_ray_coherence_index","sc_seismic_coherence_index","sc_mean_coherence_index","sc_min_slope","regional_microseism_amplitude","regional_microseism_amplitude_rel","regional_microseism_z","regional_microseism_coherence","regional_microseism_n_stations","ctd_coherent_transient_mm","ctd_transient_coherence","ctd_transient_snr","ctd_transient_active","ctd_stack_slope"],"git_commit_at_registration":"42570cc363bfe5226ec502a47a9fdf9d959c1ba8","hypothesis_id":"20260601_array_stacking_tensor_features_r_fd3984ed","locked_at":"2026-06-01T00:49:54Z","locked_file_sha256":"c7968974f604e0d2b0bdeb728b60392137b83ecb4eae1a0854e97aef235163dc","notes":"Phase 65w-STACK A-D (committed 2026-06-01). Claim: adding the 44 array-stacking tensor features (26 sc_* per-array coherence index + 9 regional_microseism_* region-local coherent sub-stack + 9 ctd_* coherent-transient/slow-slip) to tensor_eq_per_region raises held-out AUC vs an ablation baseline trained on the IDENTICAL train/test split WITHOUT those 44 features. PASS = mean held-out AUC(with) minus AUC(without) >= 0.005 across trainable regions AND the stacking features carry nonzero aggregate gain. HONEST CAVEATS: (1) features added 2026-06-01, absent from pre-2026-06-01 resolved-prediction snapshots, so the test runs only on predictions logged after that date; (2) at the 8 continental-US anchor sites regional_microseism_* and ctd_* read 0, so signal concentrates where they fire; (3) a NULL (no AUC gain within noise) is an expected, publishable outcome recorded honestly -- this pre-registration exists so the claim cannot be tuned post hoc. Re-eval 2026-06-29.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"earthquake"},"reevaluate_at":"2026-06-29","registered_at":"2026-06-01T00:49:29Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"ablation_holdout_auc","n_permutations":null},"threshold":{"direction":"positive","metric":"delta_auc","value":0.005},"title":"Array-stacking tensor features raise per-region EQ classifier held-out AUC","withdrawn":{"at":"2026-08-11T17:35:26Z","by":"Mljessop@Ubuntu-2404-noble-amd64-base","from_status":"locked","reason":"UNTESTABLE AS SPECIFIED -- withdrawn 2026-08-11, not disconfirmed. The claim was never evaluated and this is not a null result.\n\nMeasured by scripts/eq_tensor_ablation.py over the 90 d window: 3,056 tensor_eq_per_region predictions, 718 skipped as unresolved, 2,338 with no alignable tensor snapshot -> ZERO usable rows. Two independent walls:\n\n(1) ALIGNMENT. state_snapshots stores a bare positional vector with no name manifest. feature_manifests was introduced 2026-08-08 and holds ONE manifest; 302 of 8,760 snapshots carry a layout_sha (3.4% of the archive, spanning Aug 8-11). The entire 2026-06-01 -> 2026-08-08 window this hypothesis was sealed over is unlabelled. 230 distinct widths exist, and width cannot identify a layout because two different name lists can share a length, so the archive cannot be read retrospectively without guessing.\n\n(2) MATURITY. Inside the manifest window no prediction has expired, so there are no final outcomes to train against.\n\nThe seal and the Bitcoin anchor were sound throughout. What failed is that nobody checked AT LOCK TIME that the data required to grade the claim would still be readable at grading time -- the manifest system that would have made it gradeable arrived 68 days after the lock.\n\nAdditionally, the 0.005 threshold is below this test design s noise floor: subsample=0.8 makes each fit stochastic and dropping columns shifts the RNG stream, so a single-fit ablation moves held-out AUC by about +-0.018 even when the removed block is identically zero. The re-issue averages over seeds and regions and refuses to return PASS/FAIL when the standard error exceeds the threshold.\n\nRe-issued as a forward test with re-eval 2026-11-10 over manifest-covered snapshots only. The idea is not retracted; this instance is."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-06-29"},{"data_source":"earth_state_tensor.db + prediction_features (resolved earthquake predictions logged 2026-06-01 onward, with the 39 satellite-channel features present)","expected_direction":"positive","feature_set":["eco_lst_c","eco_lst_local_anomaly_c","eco_n_passes","eco_covered","eco_data_age_days","eco_data_real","emit_clay_band_depth_2200","emit_carbonate_band_depth_2340","emit_ferric_band_depth_900","emit_ndvi","emit_ch4_enhancement_ppmm","emit_covered","emit_data_age_days","emit_data_real","lts_surface_temp_c","lts_clay_ratio","lts_iron_oxide_ratio","lts_ndvi","lts_swir16_refl","lts_swir22_refl","lts_n_scenes","lts_data_real","swm_landcover_class","swm_is_water","swm_is_bare","swm_data_real","lnm_elevation_m","lnm_slope_mean_deg","lnm_local_relief_m","lnm_roughness_m","lnm_fabric_coherence","lnm_fabric_orientation_deg","lnm_data_real","swot_wse_m","swot_wse_trend_mm_yr","swot_n_passes","swot_covered","swot_data_age_days","swot_data_real"],"git_commit_at_registration":"09b84998619c2cca813739488e92377a5ca79b8b","hypothesis_id":"20260601_satellite_tensor_block_raises_pe_984c1982","locked_at":"2026-06-01T19:05:48Z","locked_file_sha256":"704b125c3b72db1fb2e0c5be4f07ca0d07656bd07bf665429e3bdb8120f99e7b","notes":"Phase 65w-STACK satellite-channel program (committed 2026-06-01, channels lts/swm/lnm/emit/eco/swot). Claim: adding the 39 satellite tensor features (6 channels: ECOSTRESS thermal anomaly, EMIT 285-band mineral band-depths + methane, Landsat thermal/SWIR, ESA WorldCover, Cop-DEM lineament fabric, SWOT water level) to tensor_eq_per_region raises held-out AUC vs an ablation baseline trained on the IDENTICAL train/test split WITHOUT those 39 features. PASS = mean held-out AUC(with) minus AUC(without) >= 0.005 across trainable regions AND the satellite features carry nonzero aggregate gain. HONEST CAVEATS, expected outcome is NULL: (1) the satellite block is RESOURCE-oriented (mineralogy/landcover/water), not an earthquake-precursor block; only eco_ thermal anomaly, lnm_ structural fabric, and lts_ thermal/NDVI vary at land anchors, while emit_/swm_/swot_ read ~0 at the continental EQ anchor sites, so the block is diluted toward zero gain by construction. (2) features added 2026-06-01, absent from earlier resolved-prediction snapshots, so the test runs only on predictions logged after that date. (3) the genuinely EQ-plausible future signal is InSAR LOS deformation, which is NOT in this block yet (v1 atmosphere-dominated, deferred to MintPy v2); this pre-reg is therefore primarily a GUARDRAIL (adding the block must not DEGRADE EQ AUC) plus a discovery channel for any surprise lift from eco_/lnm_. A NULL within noise is the expected, publishable result and this pre-registration exists so the claim cannot be tuned post hoc. Co-evaluates with 20260601_array_stacking_tensor_features_r_fd3984ed. Re-eval 2026-06-29.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"earthquake"},"reevaluate_at":"2026-06-29","registered_at":"2026-06-01T19:05:29Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"withdrawn","test":{"extra":{},"method":"ablation_holdout_auc","n_permutations":null},"threshold":{"direction":"positive","metric":"delta_auc","value":0.005},"title":"Satellite tensor block raises per-region EQ classifier held-out AUC","withdrawn":{"at":"2026-08-11T17:35:27Z","by":"Mljessop@Ubuntu-2404-noble-amd64-base","from_status":"locked","reason":"UNTESTABLE AS SPECIFIED -- withdrawn 2026-08-11, not disconfirmed. The claim was never evaluated and this is not a null result.\n\nMeasured by scripts/eq_tensor_ablation.py over the 90 d window: 3,056 tensor_eq_per_region predictions, 718 skipped as unresolved, 2,338 with no alignable tensor snapshot -> ZERO usable rows. Two independent walls:\n\n(1) ALIGNMENT. state_snapshots stores a bare positional vector with no name manifest. feature_manifests was introduced 2026-08-08 and holds ONE manifest; 302 of 8,760 snapshots carry a layout_sha (3.4% of the archive, spanning Aug 8-11). The entire 2026-06-01 -> 2026-08-08 window this hypothesis was sealed over is unlabelled. 230 distinct widths exist, and width cannot identify a layout because two different name lists can share a length, so the archive cannot be read retrospectively without guessing.\n\n(2) MATURITY. Inside the manifest window no prediction has expired, so there are no final outcomes to train against.\n\nThe seal and the Bitcoin anchor were sound throughout. What failed is that nobody checked AT LOCK TIME that the data required to grade the claim would still be readable at grading time -- the manifest system that would have made it gradeable arrived 68 days after the lock.\n\nAdditionally, the 0.005 threshold is below this test design s noise floor: subsample=0.8 makes each fit stochastic and dropping columns shifts the RNG stream, so a single-fit ablation moves held-out AUC by about +-0.018 even when the removed block is identically zero. The re-issue averages over seeds and regions and refuses to return PASS/FAIL when the standard error exceeds the threshold.\n\nRe-issued as a forward test with re-eval 2026-11-10 over manifest-covered snapshots only. The idea is not retracted; this instance is."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-06-29"},{"data_source":"NWIS IV discharge (00060) at gauges within 5 km of FDSN stations; Burtin 2008 turbulent-river-noise physics","expected_direction":"positive","feature_set":["fluvial_seismic_band_power","3tier_archive_16hz","station_river_distance"],"git_commit_at_registration":"7f0bdc5947dcd2af31042f962abd58e4810fe344","hypothesis_id":"20260708_vq32_virtual_stream_gauge_from_f_4f88fc9e","locked_at":"2026-07-08T22:04:24Z","locked_file_sha256":"e7b8283c2e66af0130e6d1e75d9f37ffabc222a169a2fc320fa78644c20c1cc5","notes":"CONFIRMATORY, registered before running. Seismic band power at river-adjacent stations vs co-located NWIS discharge. PASS = Spearman rho >= 0.60 at >=50% of qualifying pairs AND beats seasonal-climatology baseline. WITHDRAW if <5 qualifying station-gauge pairs exist (geometry risk is the known unknown -- honest underpowered outcome). Hook: USGS decommissions gauges over budget; a validated virtual gauge fills retired sites. Re-eval 2026-08-15.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"river_discharge"},"reevaluate_at":"2026-08-15","registered_at":"2026-07-08T22:04:09Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{"baseline":"seasonal_climatology","mode":"per_pair_rank_correlation","pass_criteria":{"min_frac_pairs_passing":0.5,"min_n_pairs":5,"min_spearman_rho":0.6}},"method":"vq_validation_harness.validate_series","n_permutations":null},"threshold":{"direction":"positive","metric":"spearman_rho","value":0.6},"title":"VQ32 virtual stream gauge from fluvial seismic noise vs co-located NWIS","withdrawn":{"at":"2026-07-08T22:59:38Z","by":"Mljessop@Mike","from_status":"locked","reason":"Pre-specified underpowered clause: 5 geometric pairs found but UU.LTU gauge (10126180) aligned 0 hours -> 4 surviving pairs < locked minimum 5. Of the 4 that ran, 0 passed -- even RE.EACA at 0.6 km showed raw rho 0.202 collapsing to -0.045 after the pre-stated cubic detrend: the correlation was the shared seasonal runoff trend, exactly what the climatology rule was built to catch. PHYSICS: fluvial seismic noise detections (Burtin 2008) are on large turbulent rivers; small UT/WA/AK creeks (10-100 cfs) do not lift 1-7 Hz power above ambient at 0.6-4 km. A v2 would pair stations with MAJOR rivers (Colorado/Snake/Columbia class) -- separate future registration."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-08-15"},{"data_source":"data/world_stress_map/wsm2016.csv (A-C quality records in regions where input channels are data_real)","expected_direction":"negative","feature_set":["anv_anisotropy_vector","grp_gravity_fault_strike","focal_mechanisms","soc_shmax_regional"],"git_commit_at_registration":"7f0bdc5947dcd2af31042f962abd58e4810fe344","hypothesis_id":"20260708_vq36_stress_rosette_predicts_hel_2f807511","locked_at":"2026-07-08T22:04:24Z","locked_file_sha256":"99476caad7d5a12bbc1735789808a619ce9188fa717dd7c88552c6d6620c7071","notes":"CONFIRMATORY, registered before running. PASS = held-out circular MAE <= 30 deg AND beats nearest-WSM-neighbor baseline >=10% with bootstrap ci_lo>0. HONEST SCOPE: validated only where anv_/grp_ inputs are real (gravity grid = NW Utah; anisotropy = station-dependent) -> expected outcome may be REGIONAL instrument, scope-reduction is a valid partial. WITHDRAW if <30 usable held-out sites (underpowered). Re-eval 2026-08-15.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"stress_orientation"},"reevaluate_at":"2026-08-15","registered_at":"2026-07-08T22:03:45Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{"baseline":"nearest_wsm_neighbor_azimuth","block_deg":1.0,"k_folds":5,"metric":"circular_mae_deg","pass_criteria":{"max_circular_mae_deg":30,"min_n_sites":30,"min_skill_vs_nearest_wsm":0.1}},"method":"vq_validation_harness.validate_spatial","n_permutations":null},"threshold":{"direction":"negative","metric":"circular_mae_deg","value":30.0},"title":"VQ36 stress rosette predicts held-out WSM SHmax azimuth","withdrawn":{"at":"2026-07-08T22:26:45Z","by":"Mljessop@Mike","from_status":"locked","reason":"Pre-specified withdraw clause fired: only 22 A-C WSM sites where input channels are real (3 in UGS gravity-grid box + 19 within 50km of anv_ prewarm sites) vs locked minimum 30 -> UNDERPOWERED. Audit also found stress_orientation_calibration.py READS WSM, so the locked soc_shmax_regional covariate would have been circular (target leakage) and required exclusion regardless. A v2 with NATIONAL covariates (qfault orientation density, focal mechanisms) is a separate future registration."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-08-15"},{"data_source":"prediction_ledger.db (model_name=imminent_convergence_72h) vs bounds-first USGS/EMSC matching","expected_direction":"positive","feature_set":["n_converged_models","cross_model_corroboration"],"git_commit_at_registration":"44468e1efb48bf9b453b70a93f15fbea4493440f","hypothesis_id":"20260712_imminent_convergence_72h_forward_4ec47bcd","locked_at":"2026-07-12T00:08:51Z","locked_file_sha256":"89007057e622215fd3216bef4c5f5f1a83ee0b79819d3f3f597221a47063d8a3","notes":"Forward test of the imminent-convergence 72h tier shipped 2026-07-11 (commit 44468e1e). Trigger rule: >=4 DISTINCT models holding active in-window HIGH/CRITICAL EQ forecasts on the same region simultaneously -> ONE scored 72h prediction per region per episode. PASS: resolved hit rate >= 70% at n >= 30 resolved predictions over the first 90 days. Backtest basis (60d, dedup region+day): 77.8% hit-within-72h at 1.4 triggers/day; the >=3+CRITICAL alternative measured ANTI-selective (62.1% vs 69.9% baseline) and was dropped pre-launch. UNDERPOWERED if n < 30 by re-eval: extend 90 more days rather than judge early. Honest context: constituent 7-day HIGH/CRIT base runs ~79% lonely / 95% at 3+ models over the FULL 7d window; this tier claims the shorter 72h window at higher selectivity. Re-eval 2026-10-11 (or later date if underpowered).","ots_proof_path":null,"outcome":{"lookahead_hours":72.0,"min_magnitude":null,"type":"earthquake"},"reevaluate_at":"2026-10-10","registered_at":"2026-07-12T00:08:36Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"resolved_hit_rate","n_permutations":null},"threshold":{"direction":"positive","metric":"resolved_hit_rate_pct","value":70.0},"title":"imminent_convergence_72h forward test: >=70% resolved hit rate","withdrawn":{"at":"2026-07-12T00:18:17Z","by":"Mljessop@Mike","from_status":"locked","reason":"Superseded pre-launch, zero data accrued: Mike directed shadow-mode testing before any public exposure, so the tier records to a private shadow ledger (data/imminent_shadow.jsonl) instead of prediction_ledger.db. Re-registered as v2 with the shadow data source + --evaluate USGS scoring; metric, threshold (>=70% at n>=30), and 90-day window unchanged."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-10-11"},{"data_source":"USGS FDSN declustered M5+ onsets + data/escalation_shadow.jsonl","expected_direction":"positive","feature_set":["max_early_mag_ratio","rate_accel_late_over_early","n_early"],"git_commit_at_registration":"4135e7b7cdf2ae8002b145ade16a8a2cc0764aa7","hypothesis_id":"20260715_24h_escalation_shadow_tier_forwa_b5e93582","locked_at":"2026-07-15T01:15:23Z","locked_file_sha256":"4e0a57f378fa455126f2c7e5b476c582383255b29fd405bf02c49c0a378f3fea","notes":"SHADOW-only 24h escalation rung (scripts/escalation_shadow_predictor.py, model escalation_shadow_24h). PASS = forward precision >= 0.15 AND lift >= 2.0x over the study base rate 0.084 (>= 0.168), at n >= 30 RESOLVED flags. WITHDRAW if < 15 resolved by re-eval. In-sample: precision 0.2231 = 2.66x base on n_train=595. Never touches the public ledger; promote to a public 24h tier only on PASS. Companion to the 72h imminent pre-reg 20260712_v2..a3f748c7 and the 10h microseism pre-reg 20260701..47d7b6f5.","ots_proof_path":null,"outcome":{"lookahead_hours":480.0,"min_magnitude":null,"type":"earthquake"},"reevaluate_at":"2026-10-11","registered_at":"2026-07-15T01:15:10Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"forward out-of-sample precision of escalation_shadow_24h flags (p>=0.6); a flag HITS iff a subsequent event >= trigger_mag-0.05 occurs within the Wells-Coppersmith radius during the 20-day window","n_permutations":null},"threshold":{"direction":"positive","metric":"precision","value":0.15},"title":"24h escalation shadow tier forward precision","withdrawn":{"at":"2026-07-19T02:34:50Z","by":"Mljessop@Ubuntu-2404-noble-amd64-base","from_status":"locked","reason":"INVALID TEST, not a failed one. Two defects in the instrument were found on 2026-07-18, both inside the forward window. (1) decluster_onsets sized its aftershock-exclusion radius from the trigger magnitude instead of the preceding events, so aftershocks were admitted as sequence onsets: 4 of the 7 distinct flags logged under this lock were aftershocks, three of them within 1.2h of a single M7.3 off Puerto Madero. (2) The calibration was fit on study onsets selected the same way, so its 8.4pct base rate was inflated by aftershock sequences whose later events trivially exceeded a small trigger. Corrected: GK74 windows sized by the preceding event, study re-run (1817 triggers to 1453 onsets, 523 usable sequences, base rate 5.0pct), calibration re-fit (threshold 0.60, in-sample precision 0.146 = 2.9x base). The locked test measured a model that no longer exists; the 8 records are quarantined to escalation_shadow_PRE_DECLUSTER_FIX.jsonl and excluded from any forward record. Superseded by the v3 registration."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":null},{"data_source":"USGS FDSN M5+ triggers, GK74-declustered onsets; escalation_shadow.jsonl","expected_direction":"positive","feature_set":["max_early_mag_ratio","rate_accel_late_over_early","n_early"],"git_commit_at_registration":"c11db10eff8dff0ed9bc5cf2f7ec85049d43e2c8","hypothesis_id":"20260719_v3_24h_escalation_shadow_tier_fo_df265de9","locked_at":"2026-07-19T02:35:24Z","locked_file_sha256":"56b245fdd55bb9aa60d7934cbbb00c1a9dc6897a646b7d6893f135c9a843fdbe","notes":"Supersedes 20260715_24h_escalation_shadow_tier_forwa_b5e93582, withdrawn as an INVALID test after two instrument defects were found inside its window (aftershocks admitted as onsets; calibration fit on the same contaminated onsets). Instrument now: decluster_onsets uses Gardner-Knopoff 1974 windows sized by the PRECEDING event; calibration re-fit on the corrected 4-year study (1817 triggers -> 1453 onsets -> 523 usable sequences, 26 escalations = 5.0 pct base rate), threshold 0.60, in-sample precision 0.146 = 2.9x base, recall 0.577, fires 103/523. PASS = forward precision >= 0.10 (2x base) at n >= 30 resolved flags. Below 30 resolved by the re-eval date the test is UNDERPOWERED and re-registers rather than reporting. A null is publishable. SHADOW ONLY: private jsonl plus email, nothing on the public ledger or site. The 8 pre-fix records are quarantined and excluded. Honest caveat: the calibration is in-sample and the dominant coefficient is max_early_mag_ratio (0.877 standardized) so the flag is close to a calibrated comparable-magnitude detector; the study base rate itself dropped from 8.4 to 5.0 pct once aftershock sequences stopped being counted as escalations, which is the size of the contamination this replaces. re-eval 2026-10-16","ots_proof_path":"hypotheses/locked/20260719_v3_24h_escalation_shadow_tier_fo_df265de9.json.ots","outcome":{"lookahead_hours":24.0,"min_magnitude":5.0,"type":"earthquake"},"reevaluate_at":"2026-10-17","registered_at":"2026-07-19T02:35:11Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"withdrawn","test":{"extra":{},"method":"forward out-of-sample precision of the shadow flag at threshold 0.60","n_permutations":null},"threshold":{"direction":"positive","metric":"precision","value":0.1},"title":"v3 24h escalation shadow tier forward test (post decluster fix)","withdrawn":{"at":"2026-07-22T15:58:05Z","by":"Mljessop@Ubuntu-2404-noble-amd64-base","from_status":"locked","reason":"Underpowered by instrument, not by hypothesis. Two independent population limits measured 2026-07-22: (1) only 5 percent of global M5+ onsets have >=3 delivering stations within 300 km, and outside that the public M2.5+ catalogue is too sparse for early_features; (2) a train/serve skew -- the calibration was fitted on settled catalogues (37 percent usable) but the tier serves at 24-96h on provisional ones (13 percent usable). Neither permits n=30 by the 2026-10-16 re-evaluation. See reports/methodological_nulls/escalation_tier_population_limits.md"},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":"2026-10-16"},{"data_source":"prediction_ledger shadow bands + USGS/multi-catalog for b,Mc + ETAS conditional rate","expected_direction":"positive","feature_set":[],"git_commit_at_registration":"9177b40b2a79bd94be49ec58302d70d2949c3c98","hypothesis_id":"20260723_dynamic_magnitude_evaluator_pois_ed6a2456","locked_at":"2026-07-23T15:22:15Z","locked_file_sha256":"48584cea173f8842496f4721d402a6bb83040faf29f25effdccf96b8f279bcb9","notes":"DYNAMIC MAGNITUDE EVALUATOR -- shadow test, criterion locked before the evaluator exists.\n\nWHAT IS BEING TESTED\nThe shipped expected-magnitude band today is RISK_LEVEL_BUCKETS[tier] -- a lookup keyed on a\nlabel -- plus five corrections (Bath cap, tectonic ceiling, region floor, calibrated floor,\nlog-window nudge). Event RATE never enters it, so a region producing 40 events a month and one\nproducing 2 receive the same band from the same tier. The window nudge (1/b)*log10(T/168h) is\nidentically zero at a 7-day window for ANY b, so b currently cannot act on most predictions.\n\nThe candidate replaces the lookup with the closed form for the largest event of a\nPoisson-Gutenberg-Richter process over the forecast window:\n    P(Mmax < m) = exp( -lambda*T * 10^(-b(m-Mc)) )\n    m_q         = Mc + (1/b) * log10( lambda*T / -ln q )\nBand = [m_0.10, m_0.90], additionally capped at the region's Mmax. This subsumes all five\npatches: the floor IS Mc, the window enters as lambda*T, the tail shape IS b, and b acts on\nevery prediction rather than only on non-7-day windows.\n\nPARAMETERS -- which are dynamic is decided by our own prior results, not by preference\n  lambda (rate): DYNAMIC. The one component with validated skill -- ETAS beat the long-run\n      baseline on CSEP's own yardstick, replicated across japan (+0.2947) and indonesia\n      (+0.1922), sign-test p=0.0063. Fallback: long-run Poisson rate when ETAS is unavailable.\n  b, Mc: LONG-BASELINE, refreshed quarterly, NOT time-varying. Three independent tests say\n      reactive size-distribution parameters are worse: per-region adaptive quantile bands\n      79.2% -> 75.7% OOS (rejected); recency b 1.34 -> 1.49 (noisier, under-predicts large\n      events); tensor-conditioned b pre-registered and FAILED 0 of 3 regions, 0 of 9 configs,\n      indonesia significantly negative (20260719_tensor_adds_magnitude_informatio_3b3bf3bc).\n      b by Aki-Utsu MLE, Mc by max-curvature + 0.2, >= 150 events above Mc over >= 4 years,\n      else fall back to the global b = 1.0.\n  T: the forecast's own valid window.\n\nDESIGN\nSHADOW. Both bands are computed for every new earthquake prediction and BOTH are recorded;\nneither is shipped and the public band is unchanged for the duration. Scoring is PAIRED -- the\nsame predictions carry both bands -- so the comparison cannot be confounded by which\npredictions happened to match.\n\nPRIMARY METRIC\nBand coverage = fraction of matched observations whose magnitude lies within [min, max].\nThe band is an 80% band, so the target is 0.80 and the quantity of interest is |coverage-0.80|.\n\nPASS  = |coverage_dynamic - 0.80| < |coverage_current - 0.80|\n        AND mean(max-min) for the dynamic band is not more than 10% wider than current.\n        (The width guard exists because coverage can be bought trivially by widening.)\nFAIL  = anything else.\nUNINFORMATIVE = if |coverage_current - 0.80| <= 0.02 at re-evaluation, the incumbent is already\n        calibrated and the test cannot resolve; declared uninformative, not passed or failed.\n        This is a power condition fixed in advance, not an outcome-dependent escape.\nWITHDRAW = fewer than 300 resolved shadow predictions carrying a matched observation at\n        re-evaluation -> underpowered.\n\nSECONDARY, reported but not decisive: below-floor rate, above-ceiling rate, per-region\ncoverage, mean width per tier.\n\nHONEST LIMITS, fixed in advance\n1. Coverage is measurable only on predictions that MATCHED an event -- a forecast with no\n   event has no magnitude to score. This measures band calibration GIVEN an event occurred.\n   It is not a measure of forecast skill and must not be reported as one.\n2. Mc is a network-detection property, so any change in station coverage during the window\n   shifts it. Mc is therefore frozen at its locked value for the duration.\n3. The evaluator code does not exist at lock time. Its sha256 will be recorded when it lands,\n   and it must conform to the formula and parameters above; this locked criterion governs.\n4. Current shipped-band coverage was deliberately NOT measured before locking, so the bar\n   could not be fitted to it.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":2.5,"type":"earthquake"},"reevaluate_at":"2026-10-21","registered_at":"2026-07-23T15:21:54Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"paired shadow comparison of |coverage-0.80| with a band-width guard","n_permutations":null},"threshold":{"direction":"positive","metric":"abs_coverage_error_improvement","value":0.0},"title":"Dynamic magnitude evaluator (Poisson-Gutenberg-Richter band) beats the tier-lookup band on out-of-sample calibration","withdrawn":{"at":"2026-07-23T15:30:59Z","by":"Mljessop@Mike","from_status":"locked","reason":"INVALID TEST, not a failed one -- withdrawn before it ran, so nothing is contaminated. The locked primary metric scores 'fraction of matched OBSERVATIONS inside the band', but the locked formula P(Mmax < m) is the CDF of the LARGEST event in the window. Those are different objects. Measured on 120 days of ledger: 12.7 observations per matched prediction (max 542), only 21% with a single observation, and magnitudes spanning M2.5-M8.1 WITHIN one prediction -- a 5.6-magnitude spread. A max-distribution band scored against every aftershock must under-cover, so the incumbent's wide M2.7-8.0 band would have won for a reason unrelated to physics and the FAIL would have been uninterpretable. Scoring per-observation is also a unit-of-analysis error in its own right: 12.7 correlated aftershocks of one sequence are not 12.7 independent tests, which is the exact defect the Jul-18 audit was about. Superseded by v2, which scores the MAXIMUM observed magnitude per prediction -- one score per prediction -- fixing both the object mismatch and the independence problem. Formula, parameter policy, width guard, uninformative condition and withdraw gate all carry over unchanged."},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":null},{"data_source":"dense-array RattleSnake survey-1 master_sensors.csv (sealed values below); grading data = ANY independent depth-resolving measurement at the same two points, whichever arrives first: MSR .pb waveforms if a decoder or re-export ever appears, a DeepMap/dense-array re-shoot, or a borehole","expected_direction":"positive","feature_set":["hvsr_deep_f0_hz_0034","hvsr_deep_f0_hz_0087","hvsr_shallow_f0_hz_0034","hvsr_shallow_f0_hz_0087"],"git_commit_at_registration":"236fb1d86669fcd3187b996ae3c25a45f8911a65","hypothesis_id":"20260821_pmri_predicts_a_deep_interface_c_c8759acd","locked_at":"2026-08-21T19:28:10Z","locked_file_sha256":"c24900b6b48781f8a70b57378ed50d7daa41e44dcff57a783656f854ba5aff94","notes":"SEALED BEFORE ANY MSR WAVEFORM IS READ. The MSR crew occupied both drill targets on 2026-07-27: 10 of 19 east shots sit within 25 m of dense-array sensor 0087 (median 5 m) and 9 within 25 m of 0034, closest 0.6 m. dense-array survey-1 HVSR at those exact sensors, sealed here: 0034 deep f0 = 0.2002 Hz, 0087 deep f0 = 0.3202 Hz; 0034 shallow f0 = 0.6808 Hz, 0087 shallow f0 = 0.6804 Hz. Under depth = Vs/(4 f0) with a constant Vs those give 499 m vs 312 m deep and 147 m vs 147 m shallow, a deep ratio of 1.60 and a shallow ratio of 1.0006, over a horizontal separation of only 280 m. STATED AS A RATIO ON PURPOSE: the metres inherit an ASSUMED constant Vs of about 399 m/s which is not a measurement, whereas the frequency pair and their ratio survive any Vs choice provided Vs is similar at two points 280 m apart. PASS requires BOTH (a) deep interface at 0034 deeper than at 0087 by a ratio of at least 1.25, and (b) the shallow interface the same at both within 15 percent, which is the internal control -- a method that moves the shallow interface as much as the deep one is measuring something other than structure. FAIL if the deep depths are equal within 15 percent or the ordering reverses; that refutes the dense-array structural model at the drill collar. UNTESTABLE, not FAIL, if the grading method cannot resolve the deep interface at all or cannot place both points on a common scale. Physical reading being tested: a large deep contrast beneath an unchanged shallow one is what a fault between the two points looks like, and 0034 was independently ranked the strongest fault conduit in the east field. This entry makes no claim about which target is the better well.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"subsurface_structure"},"reevaluate_at":"2027-08-21","registered_at":"2026-08-21T19:27:35Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"ratio_of_deep_interface_depth_0034_over_0087_with_shallow_interface_as_internal_control","n_permutations":null},"threshold":{"direction":"positive","metric":"deep_depth_ratio_0034_over_0087","value":1.25},"title":"dense-array predicts a deep-interface contrast between drill targets 0034 and 0087 that any independent method at those points must reproduce","withdrawn":{"at":"2026-08-21T20:29:22Z","by":"Mljessop@Mike","from_status":"locked","reason":"PASS threshold 1.25x sits below the clean neighbor-pair median scatter of 1.30x (measured 2026-08-21 on 43 genuine deep resonances), so it would pass on noise. Re-issued forward with the deep threshold raised to 1.5x. NOT re-graded -- never tested; withdrawn before any grading data existed."},"_ots_anchored":false,"_status_dir":"withdrawn","_re_eval_date":null},{"data_source":"dense-array RattleSnake SURVEY-1 master_sensors.csv (sealed values below); graded by ANY independent depth-resolving measurement at the same two points whose deep-band pair-noise is below the 1.5x threshold -- a re-shoot with long records, a reflection line, or a borehole. A second passive HVSR whose own neighbor-scatter median is ~1.3x is NOT precise enough to grade this and returns UNTESTABLE.","expected_direction":"positive","feature_set":["hvsr_deep_f0_hz_0034","hvsr_deep_f0_hz_0087","hvsr_shallow_f0_hz_0034","hvsr_shallow_f0_hz_0087"],"git_commit_at_registration":"a12c2ad5ed73d84920a9cda1ed610f97e0fa6f72","hypothesis_id":"20260821_pmri_survey_1_hvsr_predicts_the__9050733c","locked_at":"2026-08-21T20:29:59Z","locked_file_sha256":"817b4ebf53d133da142e5b12d1c50ea97731dcc32ff3c305a3c528e67c6a31ea","notes":"RE-ISSUE of withdrawn 20260821_pmri_predicts_a_deep_interface_c_c8759acd, threshold raised 1.25x -> 1.5x. Reason: measured 2026-08-21 that the clean neighbor-pair deep-f0 ratio (43 genuine deep resonances, deep f0 < 1 Hz, 150-450 m separation) has median 1.30x, p75 1.50x, p90 1.90x -- so a 1.25x pass bar sits below the noise floor and would pass on any pair. 1.5x = the p75 of that scatter and just under the observed contrast. SEALED SURVEY-1 VALUES (unchanged): 0034 deep f0 = 0.2002 Hz, 0087 deep f0 = 0.3202 Hz; 0034 shallow f0 = 0.6808 Hz, 0087 shallow f0 = 0.6804 Hz. Under depth = Vs/(4 f0) at constant Vs those are 499 m vs 312 m deep (ratio 1.60) and 147 m vs 147 m shallow (ratio 1.0006), 280 m apart horizontally. STATED AS A RATIO: the metres inherit an ASSUMED constant Vs ~399 m/s; the frequency pair and their ratio survive any Vs choice. ROBUSTNESS MEASURED 2026-08-21: the contrast is REAL, not scatter -- 0034 ranks #14 deepest and 0087 #39 of 43 clean sensors, the 1.60x contrast exceeds 86% of same-separation neighbor pairs, and deep-vs-shallow f0 correlate at +0.938 (strong structure). PASS requires BOTH (a) deep interface at 0034 deeper than at 0087 by a ratio >= 1.5, and (b) shallow interface equal at both within 15%, the internal control. FAIL if deep depths are equal within 15% or the ordering reverses -- refutes the dense-array structural model at the drill collar. UNTESTABLE, not FAIL, if the grading method cannot resolve the deep interface or cannot place both points on a common scale with pair-noise below 1.5x. Survey-2 does NOT cover these points (nearest station 1,096 m away) so it cannot grade this; it is examined separately as regional corroboration. Physical reading: a large deep contrast under an unchanged shallow one is what a fault between the two points looks like; 0034 was independently ranked the strongest east-field conduit.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"subsurface_structure"},"reevaluate_at":"2027-08-21","registered_at":"2026-08-21T20:29:47Z","registered_by":"Mljessop@Mike","result":null,"status":"withdrawn","test":{"extra":{},"method":"ratio_of_deep_interface_depth_0034_over_0087_with_shallow_interface_as_internal_control","n_permutations":null},"threshold":{"direction":"positive","metric":"deep_depth_ratio_0034_over_0087","value":1.5},"title":"dense-array survey-1 HVSR predicts the deep interface at drill target 0034 is at least 1.5x deeper than at 0087, with the shallow interface unchanged","withdrawn":{"at":"2026-08-21T21:35:02Z","by":"Mljessop@Mike","from_status":"locked","reason":"FOUNDATION IS NOISE. Measured 2026-08-21: the master_sensors HVSR f0_amplitude is 0.44-0.58 for all 45 RattleSnake sensors (0034=0.501, 0087=0.515), 0/45 above 1.0 -- master never found a resonance, it reported the argmax of a flat sub-1 vertical-dominated H/V curve. The validated pipeline re-run on the raw 16 Hz data independently reproduces the flat curve (H/V ~0.5, 0/22 stations with a peak >1.5, split-and-stack does not rescue it). Confirmed portfolio-wide: 15 of 16 dense-array surveys sit at f0_amp ~0.5; only Scott Field (3.28) has a real resonance. The 0034 vs 0087 contrast is the gap between two noise picks. WITHDRAWN, not re-issued -- there is no HVSR deep interface to predict. Deep structure at the drill collar rests on gravity alone (+-56% PDFs) plus active-source/WillowStick."},"_ots_anchored":false,"_status_dir":"withdrawn","_re_eval_date":null},{"data_source":"IGETS SG network, store igets_cleaned_v3sol (M3 forest + M4/S3/S4/MS4/MK4/SK3 cleaned, 20 piers/36 segments/303.3 station-yr, 38 constituent lines, manifest sha aec6b54c7a76cdeccb37c4e7f9622c2566e4fd58da4612daec47e83537584e58); NEW estimator slichter_search_v9.py sha 35059e0484624da586342a944b8ac256ed8a64c737d5d69dc9f17da7e73915ab (v4 + a constituent-line mask on the C1 kstar search + find_peaks; self-test 53/53); the base GRADED estimator slichter_search.py sha 0f43f909... is UNCHANGED; graded caller trackAB_slichter_v9.py sha fffb1b61a59e9a97 (imports v9, v3sol store-registration). The SEALED v4 estimator (2438dc37) is byte-untouched -- v9 is a NEW lineage.","expected_direction":"negative","feature_set":[],"git_commit_at_registration":"0b491b9a31a037ec2e4a57f4d47b36b4536b19f5","hypothesis_id":"20260829_slichter_1s1_search_v9_line_mask_6432d7ef","locked_at":"2026-08-29T14:01:01Z","locked_file_sha256":"5614a990cea5a15f34ed7f5df3be80aee8f157764897776e765cd36363c59723","notes":"FORWARD-BLIND: --phase C has NOT been run on v3sol through v9; sealed BEFORE any scored pass. v7/v8 hit a SPECTRAL-LEAKAGE WALL: each strong removed tidal/thermal line (M3 carrier, then M4, then S3) leaves a WHITENED peak that captures C1's kstar, so C3 tests a line's geometry (strong at all stations, not Slichter-like) and fails -- and cleaning one line just moves kstar to the next. v9 is the ESTIMATOR-LEVEL fix: exclude every constituent-line neighborhood (+-one whitening block) from the C1 kstar search + the transparency peak search, so kstar lands in the clean GAPS between lines, where the Slichter triplet (3.43/4.36/5.33 cpd) lives. CONSERVATIVE by construction: masking only SHRINKS the search space, so keeping v4's sealed null threshold makes C1 strictly HARDER to pass (no false-alarm inflation); C3's permutation null self-calibrates at whatever kstar survives. DECLARED EXPECTATION (armed red flag if wrong): C1's kstar now lands OFF every tidal/thermal line (> the exclusion radius from all 38 constituents); the persistent Slichter bound STAYS systematics-limited (a search-geometry fix, NOT a bound change). Since these 20 piers carry no known Slichter mode, the honest expected outcome is NULL_A95 with kstar OFF-LINES and C3 correctly failing on the residual noise geometry -- i.e. the estimator is no longer FOOLED by line leakage, which is the endpoint of the cleaning+estimator program. Outcomes: KSTAR_OFF_LINES + NULL_A95 -> the fix works, null stands honestly (expected); KSTAR_OFF_LINES + C3 PASS + candidate surviving all gates -> a real Slichter detection (extraordinary); KSTAR_STILL_ON_A_LINE -> the mask failed (surprise, red flag). Graded via trackAB v9-caller --store v3sol --phase C AFTER this seal + OTS anchor.","ots_proof_path":"hypotheses/locked/20260829_slichter_1s1_search_v9_line_mask_6432d7ef.json.ots","outcome":{"lookahead_hours":null,"min_magnitude":null,"type":""},"reevaluate_at":null,"registered_at":"2026-08-29T14:00:47Z","registered_by":"Mljessop@Ubuntu-2404-noble-amd64-base","result":null,"status":"withdrawn","test":{"extra":{},"method":"line-masked C1 kstar (exclude +-1 whitening block around all 38 constituents); C3 permutation-geometry at kstar; sealed v4/v9 null threshold","n_permutations":null},"threshold":{"direction":"negative","metric":"C1_kstar_off_all_constituent_lines","value":1.0},"title":"Slichter 1S1 search v9 -- line-masked kstar (estimator-level fix)","withdrawn":{"at":"2026-08-29T14:03:10Z","by":"Mljessop@Ubuntu-2404-noble-amd64-base","from_status":"locked","reason":"recorded estimator sha was the Windows CRLF working-copy hash; re-sealing forward-blind with the canonical LF/git sha b4089bce (no scored pass ran -- refused on sha mismatch)"},"_ots_anchored":true,"_status_dir":"withdrawn","_re_eval_date":null},{"data_source":"Mars InSight XB.ELYSE.02.BHU/BHV/BHW -> Z (2020-01-01..2021-12-31 core) via IRIS/IPGP/EarthScope FDSN; Earth GSN {IU.ANMO,II.BFO,IU.TSUM,IU.CCM}.00.BHZ calibrated to ACC; prepared store (--stage prepare)","expected_direction":"","feature_set":["gamma2_cross_coherence","S_x_cross_power"],"git_commit_at_registration":"817591c918dac071362e73b8cda9f82d1caef0f9","hypothesis_id":"20260902_earth_x_mars_cross_planetary_ult_75974252","locked_at":"2026-09-02T20:50:51Z","locked_file_sha256":"e1555d53505674a2685a142684745aea4d13d95221b458afa8c3c4f68ab7fbb3","notes":"Earth x Mars cross-planetary ultralight-DM Compton-line COHERENCE search v1 (AMENDMENT; supersedes v0 1084ce55 which halted at its own gate with zero bins scored). Frozen code scripts/dm_earthmars_coherence.py sha256 261e7de357630d983958113d50faf70df07424ce72a6702b3182cef56a7e05fa (git 817591c9); prereg _active_today/dm_earthmars_coherence_prereg_v1_AMENDMENT.md sha256 d5d0e808cda876118f2a03dd14d73c7106eebe4415aa363e24bc05a063111151. CHANGES vs v0 (all stated before any bin is scored): scored population = prepared store of the frozen 2020-2021 core; FFT evaluation of the same lock-in amplitude with an in-run equivalence gate (pad 32, tol 1e-3, measured 5e-5); LOCAL injection floor; injection amplitude = the search's claimed sensitivity (expected gamma2 = 2 x bar at the run's K); single-body HARD rule operationalised (z < bar AND median p over 5 draws > 0.05); empirical null on candidates + 10k-bin calibration subsample. UNCHANGED: band 0.01-1 Hz, df/f 1e-6, N_bins 4.605e6, statistic gamma2, Sidak z 5.593, INJ_TOL, xi=1, R=1.8796, projection Z, wind percentile 50, no circular-shift null, seed = sha256(EARLIEST attesting Bitcoin block hash of THIS anchor), no real bin scored before confirmation. expected_direction=NULL_BOUND.","ots_proof_path":null,"outcome":{"lookahead_hours":null,"min_magnitude":null,"type":"dark_matter_line_search"},"reevaluate_at":null,"registered_at":"2026-09-02T20:50:51Z","registered_by":"Mljessop@deepmap-compute","result":null,"status":"withdrawn","test":{"extra":{"code_git_commit":"817591c918dac071362e73b8cda9f82d1caef0f9","code_path":"scripts/dm_earthmars_coherence.py","code_sha256":"261e7de357630d983958113d50faf70df07424ce72a6702b3182cef56a7e05fa","discovery_rule":"candidate iff analytic Sidak z >= 5.593 over N_bins=4.605e6 (Carter/Nuttall MSC null) on the FFT-evaluated lock-in amplitudes (in-run equivalence gate <= 1e-3 rel amplitude error vs patch_amplitude, else exit 6) AND empirical phase-scramble p <= 0.05 with a live null AND single-body HARD rule (one-body analytic z < bar AND median scramble p over 5 one-body draws > 0.05) AND in-run positive control: per-arm recovered/injected in [0.75,1.33] against a LOCAL floor (60 log-uniform dummies within +/-20% of f_inj=0.05 Hz) at the injection amplitude whose expected gamma2 = 2 x gamma2*(K), joint z >= 5.593; else NULL.","empirical_null_policy":"phase-scramble >=500 draws on every candidate + 10,000-bin calibration subsample x 200 draws (aliveness + FAR reported, never promoting)","expected_direction":"NULL_BOUND","expected_outcome":"FIRST cross-planet DM line search; a NULL reported as a 95% exclusion on the observable per-body line amplitude (m/s^2) = the injected amplitude scaled to gamma2*(K), + the cross-coherence bound + the Earth-Mars coincidence veto; coupling (cm^2/epsilon) DEFERRED to a theorist.","frozen_config":{"amendment_v1":{"empirical_null_policy":"phase-scramble on every candidate (>= 500 draws) + 10000-bin calibration subsample (200 draws) reported, never promoting","evaluation":"FFT path (patch_fine_spectrum + cubic interpolation) reproducing patch_amplitude; gated by --stage equivalence in-run","fft_equiv_tol":0.001,"fft_pad":32,"injection_amplitude":"amp_E such that expected gamma2 = 2.0 x gamma2*(K); amp_M = amp_E / ratio","injection_floor":"LOCAL: 60 log-uniform dummy freqs within +/-20% of f_inj","scored_population":"prepared store of the frozen core (MARS_BASELINE_START..END), daily-chunked, wind percentile applied over the whole baseline","store_version":"v1","unchanged":["band","df_over_f","n_bins","statistic","bar_sidak_z","single_body_hard","inj_tol","xi","ratio","projection","wind_keep_percentile","seed_rule"]},"band_hz":[0.01,1.0],"bar_sidak_z":5.593,"df_over_f":1e-6,"earth_duty_floor_rule":"drop a station BEFORE scoring if its 2020-2021 duty < the pre-declared floor (frozen rule, not post-hoc)","earth_stations":["IU.ANMO.00.BHZ","II.BFO.00.BHZ","IU.TSUM.00.BHZ","IU.CCM.00.BHZ"],"global_alpha":0.05,"injection_tol":[0.75,1.33],"mars":{"chans":["BHU","BHV","BHW"],"loc":"02","net":"XB","sta":"ELYSE"},"mars_baseline":["2020-01-01T00:00:00","2021-12-31T23:59:59"],"n_bins":4605170.185988092,"null_battery":{"circular_shift":"EXCLUDED","dummy_floor":150,"phase_scramble":500,"time_shift":500},"projection":"Z","ratio_E_over_M":1.8796282637557162,"seal_anchor_id":null,"seed_rule":"science RNG seed = sha256(confirmed OTS Bitcoin block hash of the recorded anchor); forward-blind, set AT SEAL","single_body_rejection":"HARD","wind_keep_percentile":50.0,"xi_elastic":1.0},"pre_seal_disclosure":"pre-seal equivalence measurements formed gamma2 on 2 x 300 sampled real grid bins of a 3-day K=6 subset (both paths agreed on the candidate set; max z not recorded); pre-seal --stage equivalence is amplitude-only thereafter","prereg_path":"_active_today/dm_earthmars_coherence_prereg_v1_AMENDMENT.md","prereg_sha256":"d5d0e808cda876118f2a03dd14d73c7106eebe4415aa363e24bc05a063111151","scored_population":"prepared store of the frozen 2020-01-01..2021-12-31 core (--stage prepare, daily-chunked); wind percentile 50 over the whole baseline; manifest sha256 + counts recorded by the run","supersedes":"20260902_earth_x_mars_cross_planetary_ult_1084ce55 (HALTED at its in-run gate 2026-09-02 15:20Z, zero bins scored; stays on the ledger)","synthetic_validation":"original phase0 harness 27/27 + freeze-scan 6/6; amended path phase0fft 8/8; FFT equivalence pad 32: 5.2e-5/7.8e-5 rel amp err on real patches"},"method":"fft_cross_coherence_sidak_scan_v1","n_permutations":500},"threshold":{"direction":"","metric":"sidak_z","value":5.593},"title":"Earth x Mars cross-planetary ultralight-DM Compton-line coherence search v1 (amendment of v0 1084ce55)","withdrawn":{"at":"2026-09-02T21:04:10Z","by":"Mljessop@deepmap-compute","from_status":"locked","reason":"SUPERSEDED BEFORE ANY RUN by v1.1 (scale amendment). Frozen code 261e7de3 as registered would have held a K x 4.6e6 complex matrix per body (~100 GB at the frozen core K ~1,500) and run the exact lock-in on 3,000 bins x every patch for the equivalence gate (~60 h); neither had been exercised at scale (synthetic check ran at K=250 on a 40k-bin local grid). No real Compton bin was ever scored under this entry; its OTS stamp (pending) is kept as history. v1.1 precomputes fine spectra once, interpolates in 200k-bin chunks, and gates equivalence on a 40-patch x 600-bin subset -- same arithmetic, same bars."},"_ots_anchored":false,"_status_dir":"withdrawn","_re_eval_date":null}],"totals":{"locked":93,"tested":64,"withdrawn":20,"ots_anchored":150}}