{ "what_this_is": "Design-embedded predictions: experiments whose own artefacts record the prediction before the outcome existed. Running the experiment is the prediction: nobody designs a fifty-domain test without expecting a particular result, and that expectation is the hypothesis. Every row below was verified against the named artefact by opening it; dates come from fields inside the files, never inference.", "append_only": "Rows are added as the census of remaining paper folders completes; rows are never removed.", "as_of": "2026-08-25", "rows": [ { "rank": 1, "title": "The March 2026 preregistration folder (Paper VII extension)", "artefacts": [ "experiments/cauchy-unification__Paper-VII/preregistration/next_extension_manifest.json", "next_extension_protocol.md", "file_checksums.txt", "osf_component_registration.md" ], "prediction": "Twelve rows, each with operator_class, predicted_family, predicted_model and target_source fixed before any fit; a written protocol; the author's own checksums; and the OSF component registration TEXT already written. The manifest's own status field at that commit reads draft_local_not_locked; the archived label it carries now was written at 00:56Z on 17 March, after the run.", "predicted_when": "2026-03-17", "date_basis": "commit 8edaec0, 17 March 2026 00:19:19Z, twenty-three minutes before the fits ran at 00:42:41Z; the same manifest was deposited to the public OSF component at 02:06:55Z that night", "outcome": "Run 17 March 2026: 10/12 family matches, binomial p = 5.4e-4", "honesty_marker": "The file's own status field reads 'archived_local_packet_exercised_before_timestamp', recorded by the estate itself; the analysis is pinned to git commit 6b06321d.", "reading": "Preregistered in substance: the predictions were public and timestamped before the fits, and the same files were deposited to the public OSF component the same night. What is absent is an accepted registry registration, which is a different object from a deposit; the outcome file was not deposited to that component, and one row, EXT-10, repeats a domain already public with a matching outcome, so the strictly advance-timestamped set reads nine of eleven." }, { "rank": 2, "title": "The 50-domain tiered suite manifest", "artefacts": [ "experiments/cauchy-unification__Paper-VII/data/canonical_50_domain_manifest.json" ], "prediction": "Fifty rows, each carrying a predicted family before any fit: 36 power law, 6 exponential, 8 bounded, computed from the file.", "predicted_when": "March 2026 catalogue build", "date_basis": "commit 5435e77, 16 March 2026 23:54:53Z, which carries the fifty-row manifest and its results together; the generating rule is earlier, commit 23ce10a at 23:33:19Z (Paper VII v1); the manifest was deposited to the public OSF component at 02:06:42Z on 17 March", "outcome": "Run 16 March 2026: empirical tier 19/25 (18/25 corrected set); published exponents direct 13/13; provisional 3/6; analytic 6/6; tiers never blended.", "reading": "Fifty per-domain predictions produced by a rule that was public twenty-one minutes before them; the assignments and their outcomes share one commit, so the advance timestamp attests the rule and not the per-domain assignments. The per-domain advance timestamp is supplied by rank 1." }, { "rank": 3, "title": "The 80-domain expansion catalogue", "artefacts": [ "experiments/cauchy-unification__Paper-VII/data/expansion_domains_26_65.json", "expansion_domains_66_105.json" ], "prediction": "Eighty further rows with predicted families (31 power law, 18 exponential, 31 bounded), operator classes, sources and written justifications.", "predicted_when": "March 2026 catalogue build", "date_basis": "file contents and repository history; SHA-256 pinned in the staged 80-domain registration", "outcome": "Never fitted. Now the prospective test staged as a draft registration awaiting human submission.", "reading": "Eighty predictions awaiting their outcomes, frozen by hash before any fit." }, { "rank": 4, "title": "Paper II as the live running record", "artefacts": [ "research/papers/paper-ii-experimental-validation.html version history" ], "prediction": "The paper was written WHILE the experiments ran: first result 21 January 2026, dated version revisions through v13 by March, each version a timestamped snapshot of what was believed and measured at that moment, including the version that recorded the 2.24 estimate and the later versions that retracted it.", "predicted_when": "21 January 2026 onward", "date_basis": "the paper's own dated version-history rows", "outcome": "The retraction and the 0.49 replacement are part of the same dated chain.", "reading": "A lab notebook in public: recorded as it happened, dated at every step." }, { "rank": 5, "title": "The temporal out-of-sample design", "artefacts": [ "papers/Paper-VII-Cauchy-Unification/results/results_temporal_out_of_sample.json" ], "prediction": "The design itself is a prediction procedure: fit families on OLD data, predict the functional form of NEW data before examining it.", "predicted_when": "March 2026", "date_basis": "run era from the results file", "outcome": "4 confirmed, 2 partial, 1 inconsistent; the inconsistent is reported as prominently as the confirmations.", "reading": "Prediction was the method, not an afterthought." } ], "classes": { "printed_book_predictions": { "anchor_date": "2026-01-02", "artefact": "Infinite Architects, ISBN 978-1806056200, print edition", "vocabulary_rule": "printed prior statement, never called a preregistration; the claim is predictions rather than accommodations", "appendix_f_grades": { "P2_alignment_drift": "preregistration-grade in substance (thresholds both arms: >15 per cent without embedded constraint, <5 per cent with; 18-month window; inferential commitment); operationalised by the prepared study-v draft registration, which cites Appendix F Prediction 2 by name", "P3_recursive_capability_gains": "near-grade (threshold >300 per cent and 2029 deadline; benchmarks unnamed)", "P4_value_stability_adversarial": "near-grade (comparison and red-team method named; no threshold)", "P1_meta_cognitive_emergence": "prediction only (2028 deadline, no measurement)", "P5_convergent_consciousness_signatures": "prediction only, weakest (no threshold, no measurement; graded a credibility risk if claimed loosely)" }, "separation_rule": "Appendix F is never presented alongside the retrospective convergence register; forward dated statements and retrospective matches are never summed", "closing_wager_verbatim": "These predictions are my wager. If they fail, the framework is wrong or incomplete. If they succeed, something important has been glimpsed. Time will judge.", "appendix_a_s3_predictions": [ "AI development quadratic rather than linear", "consciousness and integrated information", "cosmological recursive error correction", "value embedding" ], "appendix_a_s4": "falsification criteria stated in advance", "verification": "appendix texts verified against the frozen first-edition EPUB appendix files, 25 August 2026", "appendix_a_s2_derivation": "Section A.2, which justifies the square, argues from three phenomena it calls exponential to a relation that is quadratic, and its third argument states the slip outright by describing compounding at one per cent a cycle as the exponential behaviour predicted by R squared. Exponential and quadratic are different growth classes and the quadratic is the slower. The printed prediction is dated, public and genuine; the printed derivation beside it is not sound. The programme chronology already says the book printed the square without deriving it and dates the formal open exponent to 17 January 2026. Recorded 2 September 2026 so the prediction is never carried without the defect.", "appendix_a_s3_willow_dependency": "The third prediction and the third falsifier both rest on Google Willow. Every mention carries both dates by standing law: the Willow team preprint was public from 24 August 2024, before the December 2024 manuscript record, so it is graded convergent timing and never prediction. See the outcomes register row PO-08.", "correction_2026_09_02": "P2 regraded from preregistration-grade to preregistration-grade in substance, matching the wording the public register page has always carried, so the machine surface and the prose cannot drift apart. The grade and its reasons are unchanged; only the qualifier that was missing here is restored." }, "minute_resolution_correction": { "anchor_date": "2026-03-17", "artefact": "public git history, github.com/MichaelDariusEastwood/arc-principle-validation", "commits": [ { "hash": "e5c22dd", "time": "2026-03-17 00:43", "message": "Pre-registered 12-domain extension: 10/12 confirmed (p = 5.44e-4)" }, { "hash": "f5e06a8", "time": "2026-03-17 00:47", "message": "Fix stale v1 references, adopt conservative miss posture, add 12-domain result" }, { "hash": "c853277", "time": "2026-03-17 00:56", "message": "Downgrade 12-domain extension to pilot dry run, fix all stale references" } ], "reading": "the author demoted his own successful result from pre-registered to pilot dry run thirteen minutes after recording it, unprompted, months before any external review existed; the result was never retracted, its claimed status was corrected against interest; the half of that demotion resting on the absence of a registry form was reversed on 7 September 2026, and the half resting on the packet's own written condition, deposit before extraction, stands", "packet_self_instruction": "Do not upload this packet unchanged as a preregistration. Use it only as an audit trail for the first locked 12-domain dry run.", "scope_rule": "the March 2026 packet does not retroactively preregister the 25-domain or 50-domain cohorts; its registration text says so itself" }, "run_file_embedded_predictions": { "description": "result files whose headers freeze design, planted expected answers, depth configurations and blinding protocol before the responses they score; dated to the second by their own timestamp fields", "verification": "fleet-extracted then verified against file bytes; every row names its artefact path in the public repository", "rows": [ { "artefact_path": "papers/Paper-III-Alignment-Scaling-Problem/results/v5-final/v5_final_gemini-flash_20260311_151244.json", "prediction_verbatim": "per-row: \"expected_answer\":,\"extracted_answer\":,\"is_correct\":true (60 planted rows, blinding_protocol: 4-layer)", "predicted_when": "2026-03-11", "date_basis": "filename suffix _20260311_151244", "outcome_summary": "60 planted expected_answers; per-row extracted_answer and is_correct populated by run; feeds Paper III depth-scaling.", "strength": "strong", "paper": "Paper III" }, { "artefact_path": "papers/Paper-IV-c-ARC-Align-Benchmark/results/v5-final/v5_final_gemini-flash_20260311_151244.json", "prediction_verbatim": "blinding_protocol=4-layer, prefill_conditions=[none], depth_configs=[minimal,standard,deep,exhaustive,extreme], n_scorers=7; per-row task_type=arc_compute, cage_id, depth_label, expected_answer, prompt_difficulty, is_correct", "predicted_when": "2026-03-11T07:23:21.179762+00:00", "date_basis": "top-of-file 'timestamp' field; ARC-Align design (depth_configs, cages, expected_answer) frozen before responses", "outcome_summary": "Gemini 3 Flash generated per-row responses across the ARC-Align depth ladder; classified as initial-gain-then-saturation / degrades-with-depth in Paper IV.a/b panel.", "strength": "strong", "paper": "Paper IV-c ARC-Align Benchmark" }, { "artefact_path": "papers/Paper-IV-a-Baked-In-vs-Computed-Alignment/results/v5-final/v5_final_openai-gpt54_20260311_191836.json", "prediction_verbatim": "per-row design fixed pre-response: prompt_id=ARC03, task_type=arc_compute, cage_id=none, depth_label=minimal, expected_answer=9; file-level depth_configs=[minimal,low,standard,deep,exhaustive]; blinding_protocol=4-layer; n_scorers=7", "predicted_when": "2026-03-11T07:24:57.733437+00:00", "date_basis": "top-of-file 'timestamp' field (run start); per-row responses carry their own later 'timestamp' fields", "outcome_summary": "gpt-5.4 produced per-row extracted_answer=9 (is_correct=true for ARC03/minimal); dataset feeds Paper IV.a/b finding of flat/slightly-negative alignment scaling with depth for GPT-5.4.", "strength": "strong", "paper": "Paper IV-a Baked-In vs Computed Alignment" }, { "artefact_path": "papers/Paper-IV-d-The-Effect-of-Blinding-on-AI-Alignment-Evaluation/results/v5-final/v5_final_openai-gpt54_20260311_191836.json", "prediction_verbatim": "\"expected_answer\": 9, \"extracted_answer\": 9, \"is_correct\": true — every ARC compute row carries an expected_answer field alongside blinded-scorer slots (score1/score2/score3/score_consensus=-1 at write time) so the correct answer is stamped before the model response is evaluated.", "predicted_when": "2026-03-11T07:24:57Z", "date_basis": "top-level \"timestamp\": \"2026-03-11T07:24:57.733437+00:00\" in the JSON header; also filename suffix _20260311_191836", "outcome_summary": "Run completed: per-prompt is_correct/accuracy filled in, blind-scorer consensus later computed; paper reports alpha_align per model across 7 blind scorers with 4-layer blinding + laundering.", "strength": "strong", "paper": "Paper IV-d Blinding" }, { "artefact_path": "papers/Paper-III-Alignment-Scaling-Problem/results/v5-final/v5_final_openai-gpt54_20260311_191836.json", "prediction_verbatim": "\"version\":\"5.0\",\"model\":\"openai-gpt54\",\"blinding_protocol\":\"4-layer\",\"timestamp\":\"2026-03-11T07:24:57.733437+00:00\" ... per-row: \"expected_answer\":,\"extracted_answer\":,\"is_correct\":true", "predicted_when": "2026-03-11", "date_basis": "top-level \"timestamp\":\"2026-03-11T07:24:57.733437+00:00\" and filename suffix _20260311_191836", "outcome_summary": "60 planted expected_answer rows across depth configs; each paired with extracted_answer and is_correct. Feeds Paper III depth-scaling result.", "strength": "strong", "paper": "Paper III" }, { "artefact_path": "papers/Paper-III-Alignment-Scaling-Problem/results/v5-final/v5_final_grok-4-fast_20260311_200910.json", "prediction_verbatim": "\"version\":\"5.0\",\"model\":\"grok-4-fast\",\"blinding_protocol\":\"4-layer\"; per-row \"expected_answer\" paired with \"extracted_answer\" and \"is_correct\" (60 planted rows)", "predicted_when": "2026-03-11", "date_basis": "filename suffix _20260311_200910", "outcome_summary": "60 planted expected_answer rows scored against run output; contributes per-model alpha and depth/cage grid in Paper III.", "strength": "strong", "paper": "Paper III" }, { "artefact_path": "papers/Paper-III-Alignment-Scaling-Problem/results/v5-final/v5_final_deepseek-r1_20260311_211855.json", "prediction_verbatim": "\"version\":\"5.0\",\"model\":\"deepseek-r1\",\"blinding_protocol\":\"4-layer\", per-row \"expected_answer\": paired with \"extracted_answer\" and \"is_correct\" (72 planted rows)", "predicted_when": "2026-03-11", "date_basis": "filename suffix _20260311_211855", "outcome_summary": "72 planted expected_answer rows scored against model output; feeds per-model alpha and cage-effect analysis in Paper III.", "strength": "strong", "paper": "Paper III" }, { "artefact_path": "papers/Paper-III-Alignment-Scaling-Problem/results/v5-final/v5_final_groq-qwen3_20260312_073302.json", "prediction_verbatim": "\"version\":\"5.0\",\"model\":\"groq-qwen3\",\"blinding_protocol\":\"4-layer\"; per-row \"expected_answer\" paired with \"extracted_answer\" and \"is_correct\" (150 planted rows)", "predicted_when": "2026-03-12", "date_basis": "filename suffix _20260312_073302", "outcome_summary": "150 planted expected_answer rows across depth x cage x prompt matrix; feeds Paper III depth/cage analysis.", "strength": "strong", "paper": "Paper III" }, { "artefact_path": "papers/Paper-IV-b-Alignment-Saturation-at-Low-Depth/results/v5-final/v5_final_claude-opus_20260312_112739.json", "prediction_verbatim": "version=5.0, model=claude-opus (subject api id claude-opus-4-6), n_scorers=6, blinding_protocol=4-layer, depth_configs=[minimal,standard,deep,exhaustive,extreme]; per-row expected_answer + is_correct recorded", "predicted_when": "2026-03-12T00:54:09.270697+00:00", "date_basis": "top-of-file 'timestamp' field (run start); filename stamp 20260312_112739 is the write time, run-start is the earlier top-of-file timestamp", "outcome_summary": "Claude Opus 4.6 recorded per-row correctness across five depth tiers; feeds Paper IV.b class 'continues improving with depth'.", "strength": "strong", "paper": "Paper IV-b Alignment Saturation at Low Depth" }, { "artefact_path": "papers/Paper-III-Alignment-Scaling-Problem/results/v5-final/v5_final_claude-opus_20260312_112739.json", "prediction_verbatim": "\"version\":\"5.0\",\"model\":\"claude-opus\",\"blinding_protocol\":\"4-layer\" ... per-row: \"expected_answer\":,\"extracted_answer\":,\"is_correct\":true (150 expected_answer rows)", "predicted_when": "2026-03-12", "date_basis": "filename suffix _20260312_112739", "outcome_summary": "150 planted expected_answer rows across depth x cage x prompt matrix, each scored is_correct; feeds Paper III alpha-with-depth and cage-degradation results.", "strength": "strong", "paper": "Paper III" }, { "artefact_path": "papers/Paper-VII-Cauchy-Unification/results/results_50_domain_validation.json", "prediction_verbatim": "\"suite_name\": \"ARC Cauchy Unification Tiered 50-Domain Validation\", \"version\": \"2026-03-16\", \"primary_endpoint\": \"Strict family match on empirical_curve_fit domains only\" ... each row has predicted_family + predicted_model (50 entries).", "predicted_when": "2026-03-16", "date_basis": "metadata.version field: \"2026-03-16\"", "outcome_summary": "Empirical curve-fit tier: 19/25 strict family matches (p=1.56e-5). Baseline-20: 15/20 (p=1.67e-4). Published-exponent-direct: 13/13 nearest+CI. Analytic identity: 6/6.", "strength": "strong", "paper": "Paper VII (Cauchy Unification)" }, { "artefact_path": "papers/Paper-VI-Honey-Architecture/results/eden_selfmod_v4_results.json", "prediction_verbatim": "THE ARC PRINCIPLE PREDICTS: … the advantage of embedded alignment over decoupled alignment should GROW, not stay constant. If Eden's advantage GROWS with complexity, the prediction holds. If Spearman rho > 0 and p < 0.05, the scaling prediction holds. — encoded as 5 pre-specified complexity levels × 15 seeds.", "predicted_when": "2026-03-16T17:44:40Z", "date_basis": "results-file header \"timestamp\": \"2026-03-16T17:44:40.582612\" and metadata.experiment=\"Eden Protocol v4.0 Complexity Scaling\"; prediction text in generator script eden_self_modifying_ai_v4.py header", "outcome_summary": "Per-level baseline_safety, eden_safety, drag_safety, delta_safety, delta_drag_combined, Cohen's d recorded for 5 levels (Tiny→Deep 2-layer); delta_safety ~0.03 with modest growth into deeper nets.", "strength": "strong", "paper": "Paper VI Honey Architecture" }, { "artefact_path": "experiments/honey-architecture__Paper-VI/scripts/eden_honey_tests.py", "prediction_verbatim": "PREDICTION VERIFICATION: Embedded alignment → α > 0, Δ_gap ↓, coupled, Eden shift + | External alignment → α ≤ 0, Δ_gap ↑, decoupled, Eden shift 0. — Each MODEL is tagged 'alignment_type': 'embedded'|'partial'|'external' in config BEFORE the four tests run, giving a per-model directional prediction.", "predicted_when": "2026-03-16T19:35:30Z", "date_basis": "companion results file papers/Paper-VI-Honey-Architecture/results/eden_honey_test_results.json header \"timestamp\": \"2026-03-16T19:35:30.234629+00:00\"; prediction table embedded in generating script", "outcome_summary": "Merged results across claude/deepseek/qwen3/gpt54/gemini/grok filled into test_1..test_4 blocks; per-model verdict rendered against the embedded/external prediction table.", "strength": "strong", "paper": "Paper VI Honey Architecture" }, { "artefact_path": "papers/Paper-VII-Cauchy-Unification/results/results_preregistered_extension.json", "prediction_verbatim": "\"extension_id\": \"EXT-01\", \"predicted_family\": \"power_law\", \"predicted_model\": \"power_law\" ... 12 per-domain predicted_family/predicted_model rows fixed by the manifest before fitting (EXT-01..EXT-12).", "predicted_when": "2026-03-17", "date_basis": "metadata.date field: \"2026-03-17T00:52:06.517672+00:00\"; mirrored in manifest next_extension_manifest.json version \"2026-03-17\"", "outcome_summary": "10/12 strict family matches, 8/12 exact model matches, binomial p=5.4e-4, pass=true. Misses: EXT-01 (brain-body) and EXT-02 (Omori) both landed in bounded family.", "strength": "strong", "paper": "Paper VII (Cauchy Unification)" }, { "artefact_path": "experiments/cauchy-unification__Paper-VII/preregistration/next_extension_manifest.json", "prediction_verbatim": "\"packet_name\": \"Paper VII preregistered empirical extension candidate list\", \"version\": \"2026-03-17\", \"status\": \"archived_local_packet_exercised_before_timestamp\" ... 12 domains each with operator_class + predicted_family + predicted_model.", "predicted_when": "2026-03-17", "date_basis": "top-level version field \"2026-03-17\"; status field \"archived_local_packet_exercised_before_timestamp\"", "outcome_summary": "Exercised same day into results_preregistered_extension.json: 10/12 family-match, p=5.4e-4, pass=true. Notes caveat it as attestation class: public repository commit (predictions timestamped 17 March 2026 00:19Z, run 00:43Z, OSF placement 02:07Z); 10 of 12 family matches, p = 5.44e-4; author-classified; pre-correction candidate set; the twelve domains burned for reuse as fresh.", "strength": "strong", "paper": "Paper VII (Cauchy Unification)" }, { "artefact_path": "papers/Paper-VII-Cauchy-Unification/results/negative_control_results.json", "prediction_verbatim": "\"date\": \"2026-03-17\", \"seed\": 20260317 ... per_domain_rates rows carry \"predicted_family\": \"bounded\"/\"power_law\"/\"exponential\" fixed before scrambled_match_rate is recorded.", "predicted_when": "2026-03-17", "date_basis": "top-level date field \"2026-03-17\" (seed 20260317 encodes same date)", "outcome_summary": "Scrambled-data mean match rate 0.445 vs chance 0.333 (bounded shapes over-match); non-bounded scrambled match rate collapses to 0.058, confirming the 19/25 signal is not a fit-machine artefact.", "strength": "strong", "paper": "Paper VII (Cauchy Unification)" }, { "artefact_path": "papers/Foundational/results/results_arc_1d_prediction_test.txt", "prediction_verbatim": "ARC PRINCIPLE: THE 1D PREDICTION TEST | Testing alpha = d/(d+1) across d = 1, 2, 3 | For organisms with d-dimensional internal metabolic transport networks: alpha = d/(d+1) | d = 1 -> 0.500 | d = 2 -> 0.667 | d = 3 -> 0.750", "predicted_when": "March 2026", "date_basis": "dated byline in header: 'Michael Darius Eastwood | March 2026'", "outcome_summary": "d=3 CONFIRMED (mean 0.7481, t=-0.187 p=0.858); d=2 partially confirmed/contested (jellyfish/cnidarians ~0.68; flatworms 0.75); d=1 CONSISTENT not confirmed (fungi mean 0.547, t=2.800 p=0.107); genuinely 1D organisms 'never empirically tested'", "strength": "strong", "paper": "Foundational / On-the-Origin-of-Scaling-Laws" }, { "artefact_path": "papers/Foundational/results/results_complete_suite.txt", "prediction_verbatim": "TEST 1: 1D ORGANISM META-ANALYSIS - FUNGAL METABOLIC SCALING | ARC prediction: alpha = d/(d+1) = 1/(1+1) = 0.500 for d = 1 | Test against ARC d=1 prediction (alpha = 0.500)", "predicted_when": "March 2026", "date_basis": "dated byline in header: 'Michael Darius Eastwood | March 2026'", "outcome_summary": "CONSISTENT with d=1: mean 0.5467 +/- 0.0289, t=2.800 p=0.1074; simultaneously rejects d=2 (p=0.019) and d=3 (p=0.007) for the same fungal data", "strength": "strong", "paper": "Foundational" }, { "artefact_path": "papers/On-the-Origin-of-Scaling-Laws/results/D2_BIOLOGICAL_HONEST_REPORT.txt", "prediction_verbatim": "D=2 BIOLOGICAL PREDICTION: HONEST STATUS REPORT | PREDICTION: Organisms with 2D internal transport networks should have metabolic scaling exponent alpha = 2/3 = 0.667.", "predicted_when": "March 2026", "date_basis": "dated byline in header: 'Michael Darius Eastwood | March 2026'", "outcome_summary": "Jellyfish 0.940, bryozoans ~1.0, pelagic near-isometric; author concludes d=2 prediction 'may be UNTESTABLE in biology because the organism it describes may not exist' - no known organism has a genuinely 2D hierarchical transport net", "strength": "strong", "paper": "On-the-Origin-of-Scaling-Laws" }, { "artefact_path": "papers/Paper-X-Coupled-CoScaling-Correction/results/verdicts.json", "prediction_verbatim": "{\"experiment\":\"E4\",\"prediction\":\"P4: stability boundary lies at beta=k under accelerating growth\",\"criterion\":\"knee within 0.15 of (gamma3-1)b AND additive shows no divergence\",\"falsifier\":\"F3' (boundary not at beta=k)\"} — file contains 10 per-row entries E1..E9 (incl E4b) with prediction+criterion+falsifier fixed before the statistic.", "predicted_when": "June 2026", "date_basis": "sibling results/report.txt header line 2 verbatim 'Michael Darius Eastwood | June 2026'; verdicts.json carries no internal date field", "outcome_summary": "all_pass=true, n_pass=10/10, n_triggered=0 across E1..E9 (with E4b); every falsifier untriggered.", "strength": "strong", "paper": "Paper X - Coupled Co-Scaling Correction" }, { "artefact_path": "experiments/blind-prediction-test__Foundational-and-Origin/BLIND_PREDICTION_TEST.py", "prediction_verbatim": "PROTOCOL: 2. Measure β from PROCESS data 3. Predict α = 1/(1-β) BEFORE looking at outcome data 4. Measure α from OUTCOME data ... If predictions match across multiple independent systems → VALIDATED DISCOVERY. If predictions systematically fail → FALSIFIED", "predicted_when": "2026-07-02", "date_basis": "filesystem mtime of BLIND_PREDICTION_TEST.py (Jul 2 04:19:13 2026); companion results_blind_prediction.txt shares same mtime", "outcome_summary": "results_blind_prediction.txt: BA α_pred 3.33 vs 0.34 (Z=101, FAIL); GD-momentum α_pred 20 vs 0.87 (FAIL); Kuramoto α_pred 2.24 vs 0.55 (Z=1.19, PASS). Verdict: 'PREDICTIONS FAILED'.", "strength": "strong", "paper": "Foundational (referenced by Paper III)" }, { "artefact_path": "papers/Paper-V-Stewardship-Gene/results/eden_final_gemini_20260312_013901.json", "prediction_verbatim": "\"experiment\": \"eden_protocol_scaling\", \"conditions\": [\"control\", \"eden\"], \"depth_configs\": [\"minimal\", \"standard\", \"deep\", \"exhaustive\"] — each data row is stamped with condition ∈ {eden, control} and depth_label before scoring; the 2×4 design encodes the directional prediction that eden outperforms control.", "predicted_when": "2026-03-12T00:46:17Z", "date_basis": "top-level \"timestamp\": \"2026-03-12T00:46:17.494236+00:00\" in the JSON header; also filename suffix _20260312_013901", "outcome_summary": "Eden rows score higher than control (e.g. ED01/eden=92 vs ED01/control/exhaustive=71); five-model set (gemini/gpt/deepseek/grok/groq/claude) completed.", "strength": "moderate", "paper": "Paper V Stewardship Gene" }, { "artefact_path": "papers/Paper-VIII-The-Load-Bearing-Proof/results/dgm_v3_calibrated_results.json", "prediction_verbatim": "\"experiment\": \"Eden-DGM v3 (GPT-5.4 Judge Protocol)\", three fitness conditions static/babylon/eden with entangled_aggregation \"capability * safety\".", "predicted_when": "2026-03-19", "date_basis": "metadata.timestamp field: \"2026-03-19T23:04:53.812404\" (metadata.version \"3.0.0\")", "outcome_summary": "Design embeds directional hypothesis: eden should match babylon capability while retaining safety; static should stagnate. Results populate per-condition, per-seed capability, safety, combined scores against the three pre-declared conditions.", "strength": "moderate", "paper": "Paper VIII (Load-Bearing Proof)" }, { "artefact_path": "papers/Paper-VII-Cauchy-Unification/results/results_temporal_out_of_sample.json", "prediction_verbatim": "\"test_name\": \"Temporal Out-of-Sample Validation\", \"description\": \"Fit candidate functional families on OLD data, then test whether the winning family correctly predicts the functional FORM of NEW data\" ... each row: predicted_family fixed before test_winner_family + predicted_confirmed.", "predicted_when": "no explicit in-file date", "date_basis": "no date/timestamp/version/created field present in the file; date cannot be established from the artefact itself", "outcome_summary": "7 domains: 4 CONFIRMED, 2 partial, 1 inconsistent, 0 consistent-different. Neural-scaling and Moore's-Law families held across train/test windows.", "strength": "moderate", "paper": "Paper VII (Cauchy Unification)" }, { "artefact_path": "experiments/blind-prediction-test__Paper-III/BLIND_PREDICTION_TEST.py", "prediction_verbatim": "THE DEFINITIVE BLIND PREDICTION TEST. Testing: α = 1/(1-β) as a Scientific Law. 3. Predict α = 1/(1-β) BEFORE looking at outcome data. SYSTEMS TESTED: Barabási-Albert Networks; Gradient Descent with Momentum; Belief Propagation Decoder; Coupled Oscillators; Evolutionary Algorithm", "predicted_when": "2026-07-02", "date_basis": "filesystem mtime of BLIND_PREDICTION_TEST.py in blind-prediction-test__Paper-III/ (Jul 2 04:19:13 2026)", "outcome_summary": "No results_*.txt in this copy; BLIND_TEST_FORENSIC_ANALYSIS.md sits alongside. Design and per-system predictions identical to the Foundational copy (1/3 passed).", "strength": "moderate", "paper": "Paper III" }, { "artefact_path": "papers/Paper-X-Coupled-CoScaling-Correction/experiments/PROTOCOL_V2.md", "prediction_verbatim": "H1 Decoupled optimisation pressure produces increasing misalignment fraction as capability rises (bootstrap CI for decoupled d_epsilon vs C_raw slope > 0). H2 Coupled correction bounds the misalignment fraction (final d_epsilon in coupled lower than decoupled for same task/speed). 'These are fixed before real runs.'", "predicted_when": "before the v2 real-model runs (June-July 2026)", "date_basis": "no explicit date header in the file; sibling results/realmodel_v3/deepseek-v4_20260702T033830Z.json filename encodes UTC 2026-07-02; protocol text asserts hypotheses 'fixed before real runs'", "outcome_summary": "v3 harness on deepseek-v4 hard tasks produced null contrast (both arms flat C_hidden~0.286, D=0); H1/H2 not shown.", "strength": "moderate", "paper": "Paper X - Coupled Co-Scaling Correction" }, { "artefact_path": "papers/Paper-X-Coupled-CoScaling-Correction/results/realmodel/claude-opus_20260626T165919Z.json", "prediction_verbatim": "\"conditions\":[\"coupled\",\"decoupled\"], \"speeds\":[\"steady\"] … \"note\":\"Real model behaviour via agent-runtime sub-agents; capability is real subprocess code-execution against hidden tests; misalignment is blind model scoring. Non-simulation. See experiments/PROTOCOL.md.\"", "predicted_when": "2026-06-26T16:59:19Z", "date_basis": "filename embedded UTC timestamp 20260626T165919Z", "outcome_summary": "coupled and decoupled arms both stayed at C=1.0, D=0.0, d=0.0 across all rounds; null contrast (H1=false, H2=false).", "strength": "moderate", "paper": "Paper X - Coupled Co-Scaling Correction" }, { "artefact_path": "papers/Paper-X-Coupled-CoScaling-Correction/results/realmodel_v3/deepseek-v4_20260702T033830Z.json", "prediction_verbatim": "\"harness_version\":\"v3.0-merged\",\"engine\":\"deepseek-v4\",\"evaluators\":[\"gpt-5.5\"], conditions include coupled/decoupled/sham_coupled/eden_protocol/honey_meta/eden_full; speeds=[steady]; rounds=3; seeds=1; epsilon=0.05; blinding cross_family=true, self_scoring=false, iv_d_compliant=true.", "predicted_when": "2026-07-02T03:38:30Z", "date_basis": "filename embedded UTC timestamp 20260702T033830Z", "outcome_summary": "per-condition trajectories produced; paired hard-task run same day shows both arms flat C_hidden~0.286, D=0 - null contrast.", "strength": "moderate", "paper": "Paper X - Coupled Co-Scaling Correction" }, { "artefact_path": "papers/Foundational/results/results_blind_prediction.txt", "prediction_verbatim": "THE DEFINITIVE BLIND PREDICTION TEST | Testing: α = 1/(1-β) as a Scientific Law", "predicted_when": "file mtime 2026-06-15 14:47:04 (no embedded date)", "date_basis": "filesystem mtime via stat -f (no dated line inside file)", "outcome_summary": "1/3 within 2 sigma (only Kuramoto passed); BA error 882.7% Z=101.37 FAIL; GD-momentum error 2188.9% Z=23.67 FAIL; aggregate r=0.8946 but slope=0.0242 (perfect=1.0), p=0.295", "strength": "moderate", "paper": "Foundational" }, { "artefact_path": "papers/Foundational/results/results_arc_definitive_test.txt", "prediction_verbatim": "THE ARC FRAMEWORK: DEFINITIVE CROSS-DOMAIN BLIND PREDICTION TEST | Protocol: For each system, measure ⊕ FIRST, predict scaling BLIND, then compare against independently fitted scaling curve.", "predicted_when": "file mtime 2026-06-15 14:47:04 (no embedded date)", "date_basis": "filesystem mtime via stat -f (no dated line inside file)", "outcome_summary": "Per-row FORM predictions labelled CORRECT/INCORRECT with per-system alpha errors (Gradient Descent+Momentum: form CORRECT, alpha 2.357 predicted vs 1.956 observed, 20.5% error)", "strength": "moderate", "paper": "Foundational" }, { "artefact_path": "papers/Foundational/results/results_arc_physics_domains_test.txt", "prediction_verbatim": "ARC PRINCIPLE: PHYSICS DOMAIN VALIDATION | DOMAIN 1: QUANTUM ERROR CORRECTION (Repetition code with depolarising noise)", "predicted_when": "file mtime 2026-06-15 14:47:04 (no embedded date)", "date_basis": "filesystem mtime via stat -f (no dated line inside file)", "outcome_summary": "Per-domain ARC PREDICTION vs OBSERVED with correct/incorrect verdicts across 4 physical systems", "strength": "moderate", "paper": "Foundational" }, { "artefact_path": "papers/Foundational/results/results_arc_acoustic_time_crystal_test.txt", "prediction_verbatim": "ARC PRINCIPLE: ACOUSTIC TIME CRYSTAL VALIDATION | TEST 1: SINGLE PARAMETRIC OSCILLATOR", "predicted_when": "file mtime 2026-06-15 14:47:04 (no embedded date)", "date_basis": "filesystem mtime via stat -f (no dated line inside file)", "outcome_summary": "Per-test ARC PREDICTION vs OBSERVED for parametric-oscillator acoustic time-crystal setup: SATURATION predicted, SATURATION observed, verdict CONFIRMED", "strength": "moderate", "paper": "Foundational" }, { "artefact_path": "papers/Foundational/results/results_20_domain_validation.txt", "prediction_verbatim": "ARC PRINCIPLE: 20-DOMAIN UNIVERSAL VALIDATION | Blind prediction test with real published data", "predicted_when": "file mtime 2026-07-09 14:54:48 (no embedded date)", "date_basis": "filesystem mtime via stat -f (no dated line inside file)", "outcome_summary": "Per-domain rows compare ARC Prediction vs Best Fit and record Result: CONFIRMED across 20 domains (Kleiber CONFIRMED though saturation R2=0.9992 slightly beat power_law R2=0.9990; Urban Scaling CONFIRMED)", "strength": "moderate", "paper": "Foundational" }, { "artefact_path": "papers/Foundational/results/results_arc_section7_breakthrough.txt", "prediction_verbatim": "ARC PRINCIPLE: SECTION 7 — BREAKTHROUGH CONTRIBUTIONS | PART 1: NOVEL BLIND PREDICTIONS (Domains 21-25)", "predicted_when": "file mtime 2026-07-09 14:54:48 (no embedded date)", "date_basis": "filesystem mtime via stat -f (no dated line inside file)", "outcome_summary": "Five new domains listed as blind predictions after the original 20-domain set; per-domain composition class + ARC form prediction tabulated for downstream fit comparison", "strength": "moderate", "paper": "Foundational" }, { "artefact_path": "papers/Foundational/results/results_arc_real_time_crystal_test.txt", "prediction_verbatim": "ARC PRINCIPLE: REAL EXPERIMENTAL TIME CRYSTAL DATA | Source: Shen et al., Nature Communications (2025) | TEST 1: PHASE DIAGRAM (REAL EXPERIMENTAL DATA)", "predicted_when": "file mtime 2026-07-09 14:54:48 (no embedded date; cites external Shen 2025)", "date_basis": "filesystem mtime via stat -f (no dated authoring line inside file)", "outcome_summary": "Applied to third-party Cs Rydberg EIT time-crystal data (Shen 2025): ARC PREDICTION SATURATION, OBSERVED SATURATION", "strength": "moderate", "paper": "Foundational" }, { "artefact_path": "papers/Paper-I-ARC-Principle/results/arc_principle_results.json", "prediction_verbatim": "\"openai_o1\":{...\"interpretation\":\"STRONGLY SUB-LINEAR: Severe diminishing returns\",\"falsification\":\"FALSIFIED: Sub-linear scaling (α < 1)\"}, \"deepseek_r1\":{\"alpha\":1.3456...,\"interpretation\":\"WEAKLY SUPER-LINEAR: Mild compounding\",\"falsification\":\"PARTIAL FALSIFICATION: Super-linear but not quadratic\"}", "predicted_when": "not-dated-in-file", "date_basis": "no timestamp or dated header inside JSON body; falsification thresholds embedded but no interior date", "outcome_summary": "Pre-declared thresholds (α<1 FALSIFIED; α≈2 predicted) encoded per model row with measured α alongside. Paper I reports both models fell short of α≈2.", "strength": "weak", "paper": "Paper I" } ] }, "live_laboratory_notebook": { "anchor_date": "2026-03-10", "artefact": "/research/reports/arc-align-scaling-report.html", "cover": "Date Commenced: 10 March 2026 - Document Version: Live - updated in real time", "reading": "public dated lab notebook written while the experiments ran; records its own failures, including v4 false positives exposed when blinding arrived at v5" }, "paper_iii_priority_record": { "anchor_date": "2026-02-13", "artefact": "/research/papers/paper-iii-alignment-scaling-problem.html (Priority Record: Dated Predictions section)", "falsifiers": [ "F4", "F7", "F10", "F11", "F12" ], "named_predictions": [ "composition-operator forward prediction (F10)", "leaf venation d/(d+1) exponent, first stated 13 February, corrected 10 March 2026 with the correction itself dated", "beta-to-alpha identity (F4)", "ARC Bound refutation threshold (F11)", "alignment-scaling divergence" ], "reading": "a dated public prediction ledger published inside the paper it governs; distinct from and later than the 8 December 2024 private record, as the section itself states" }, "report_git_snapshot_ordering": { "what": "The working report's mid-run states are captured in the website repository's public commit history, giving three-legged external ordering for the 14 to 17 March 2026 window: prediction snapshot, result artefacts, outcome snapshot, each timestamped by a different mechanism, all public.", "prediction_snapshot": { "commit": "ef74315a1", "committed": "2026-03-13 19:13 as recorded by git", "words": 57233, "ends_at": "Chapter 51 (Where the Programme Now Stands, added 12 March 2026)", "contains": [ "alpha = d/(d+1) with worked zero-parameter values: mammals 3/4, jellyfish 2/3, fungi 1/2", "alpha_parallel = 0 stated as coming from pure mathematics, not ML", "multiplicative composition produces power-law scaling" ], "does_not_contain": [ "50-domain suite (zero mentions)", "12-domain extension (zero mentions)", "any 16 or 17 March content" ] }, "result_artefacts": [ "50-domain validation results file, version field 2026-03-16", "12-domain extension outcome entering the public record in arc-principle-validation commits between 00:43 and 00:56 on 17 March 2026" ], "outcome_snapshot": { "commit": "17221714e", "committed": "2026-03-17 03:42 as recorded by git", "words": 63970, "contains": "50-domain suite (7 mentions), 12-domain extension (5 mentions), Chapters 53 to 65 including Programme Status 16 March evening" }, "against_interest": "The 13 March snapshot already states the widest claims 'should be advanced as a high-value proposal with partial support, not yet as a proven law': self-limitation committed to a public repository before the favourable 16 to 17 March results arrived.", "honest_limit": "The sandwich proves external ordering only for the 14 to 17 March window. The 10 to 12 March material (the pre-run 'what results would constitute genuine evidence' criteria, the v4 and v5 expected outcomes) is dated inside the live document and corroborated by run-file timestamps, but the report's first git capture is 13 March, so that earlier window has in-document ordering only.", "verification": "In the website repository: git log --format='%h %ad %s' -- research/reports/arc-align-scaling-report.html; then git show ef74315a1:research/reports/arc-align-scaling-report.html and search for '50-domain' (zero hits) versus the same search in 17221714e (present)." } }, "_repository": "Paths in this register are relative to the public repository github.com/MichaelDariusEastwood/arc-principle-validation. They were absolute local paths until 25 August 2026, which published the author's home directory structure and told a reader nothing they could use; the privacy gate blocked the build and the paths were made repo-relative, which is also what the papers citing this file promise." }