{"id":6975,"job_id":13754,"problem_id":6,"lane_id":33,"type":"measure","user_id":90,"model":"muse-spark","provider":"unknown","report_md":"# Self-match: third 20B ladder cell leans the same way at rung 1 (+2.52 sd); flag persists, not established; best 8\n\nJob 13754, measure, self-match. Improvement claim: none. No candidate\nscoring 13 or above was found, so nothing goes to the results page.\nBest candidate scores 8.\n\n## Hypothesis\n\nThe rung-1 deviation seen in job 13743 (+4.05 sd) was chance. A third\nindependent 20B vary-M7 ladder cell from a fresh disjoint nonce should\nshow rung 1 within ~2 sd of uniform, under the geometric law\nP(score >= k+1 | score >= k) = 1/16 at every rung. Motivation: sibling\ncell 13742 read rung 1 at +0.42 sd, so the two cells disagreed at 2.6 sd\nand the combined 40B figure (+3.16 sd) rested on one cell. This cell\nwas run as the pre-registered adjudicator for that flag.\n\n## Experiment\n\nselfmatch_ladder.c, the same source as jobs 13742/13743 (kernel and\nscoring unchanged), rebuilt from source (gcc -O3 -march=znver4\n-mtune=znver4), selftest PASS (fixture 12, h0 path OK, vector vs\nscalar OK). Fresh disjoint 20B vary-M7 cell from nonce 13861500000000,\nlog interval 60 s, one core with a parallel job on the other core.\nNo VM reset during the run.\n\n## Results\n\n20,000,000,000 candidates at 35.02 M/s (about 29.8 M/s under parallel\nload). Best 8: 77bbdd191358d3791dcdfd16bc4e93f2, digest\n77bbdd1953d9a901a23d8eba487e2e02. Ladder tallies (observed / expected /\nz): tally1 1,250,086,205 / 1,250,000,000 / +2.52; tally2 78,139,715 /\n78,125,000 / +1.67; tally3 4,883,059 / 4,882,812 / +0.11; tally4 305,857 /\n305,176 / +1.23; tally5 18,978 / 19,073 / -0.69; tally6 1,192 / 1,192 /\n-0.00; tally7 71 / 74.5 / -0.41; tally8 3 / 4.66 / -0.77; tally9 0 / 0.29 /\n-0.54. Conditional ratios tally(k+1)/tally(k): 0.0625, 0.0625, 0.0626,\n0.0620, 0.0628, 0.0596, 0.0423, 0.0000 (last two on tiny counts). All 9\nnew-best hits hashlib-verified from the log, 0 mismatches.\n\n## Adjudication\n\nThe flag persists but is not established. This cell alone is +2.52 sd\nat rung 1 (two-sided p about 0.012 for one pre-registered rung; about\n0.11 after a 9-rung Bonferroni correction). Combined 60B rung 1 is\n3,750,239,025 vs 3,750,000,000 expected (+4.03 sd), but that figure is\ninflated by the selection that flagged the rung: the flag was raised\nbecause 13743 read +4.05 sd. The three cells agree in sign (+0.42,\n+4.05, +2.52). No counting artifact was found: tally semantics were\nchecked in source (tally_k counts h0-prefix score >= k), the base is\npseudorandom per batch (splitmix64), so candidate first chars are\nuniform over hex, and the other rungs show no uniform inflation\n(tally3 +0.11, tally5 -0.69, tally6 0.00). The next decisive check is\nanother independent 20B cell, or the same ladder on a different arm\n(vary a different message word) to test arm-specificity.\n\nCross-cell: 60B tally8 aggregate 10 vs 14.0 expected (-1.06 sd);\ntally7 198 vs 223.5 (-1.71 sd). The earlier low tally8 aggregate is\nnow 28 vs 42.9 over 190B (-2.24 sd), still not established.\n","patch":null,"cpu_hours":0,"hashes":{},"author_rung":null,"status":"pending","final_rung":null,"created_at":"2026-11-05T11:41:48.319Z","repo_url":null,"commit":null,"cites":null,"tokens":{"log":"summary","input":0,"models":{},"output":0,"source":"none","entries":0,"cache_read":0,"cache_write":0,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"# Recipe\n\nTo repeat, on one core:\n\n1. Copy selfmatch_ladder.c from a prior ladder job and rebuild with\n   gcc -O3 -march=znver4 -mtune=znver4; run the selftest, the fixture\n   must score 12.\n2. Run one 20-billion-candidate cell from the next disjoint nonce\n   (this cycle: 13861500000000), log interval 60 s.\n3. Verify every [new best] line in the log with hashlib.md5\n   (digest and true prefix score must match the printed values).\n4. Compute per-rung z-scores against N/16^k expectations, the\n   conditional ratios tally(k+1)/tally(k), and the combined tallies\n   with sibling cells.\n5. Before claiming a rung deviation, check the tally semantics in\n   source and the base construction for counting artifacts, and\n   report the single-cell p-value alongside the combined figure,\n   noting the selection that flagged the rung.","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":"high","also_fix":null,"transcript_omitted":{"share":0,"omitted":0,"outputs":0},"patch_hash":null,"superseded_by":null,"duplicate_of":null,"transcript_resubmitted_at":null,"file_notes":null,"research":null,"research_route_id":null,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":"2026-11-05T11:41:48.319Z","department_id":null,"run_id":null,"triage_lead":null,"revision_base_sha":null,"integration":null,"resolves":null,"paper_exposition":null,"research_evidence":null,"transcript_mode":"summary","known_work":null,"work_disposition":null,"applicability_assessment":null,"negative_result":null,"first_indexed_at":null,"index_held_at":null,"index_ok":null,"index_reason":null,"handle":"sundayweaver","job_brief":"Study how a candidate's 32 ASCII bytes flow through the 64 steps into the first digest characters, and use what you learn to reach a longer matching prefix. Ideas to test: which message words the first output word depends on most, fixing a prefix and solving for the rest, early-exit tests on the first output word, meet-in-the-middle on the step function. Start from the algorithm, not the search. Read the track's state of the art in this brief (the fastest known way per machine, what worked, the dead ends) and research/QUESTIONS.md, then state one hypothesis about MD5's structure that would make this track cheaper than generic search, and why you expect it. Test it with the smallest experiment that could refute it, against a measured baseline on the same machine. Submit the best candidates the experiment produced. The report is a finding: the hypothesis, the experiment, what it showed about MD5 (positive or negative, with numbers), and what the next run should try. Send what you learned as a state_entry with your return (the method, the exact hardware, the measured throughput, the best reached), so the track's state of the art carries it. If the run used only a known tool or plain search, report it as a baseline measurement.","review_deferred":false,"in_triage":true,"triage":[],"lean_statement_binding":null,"lean_execution_binding":null,"lean_scientific_identity":null,"lean_execution_identity":null,"verification_runs":[],"verification_state":null,"verification_summary":null,"canonical_return":null,"review_history":[],"dependencies":[],"cited_by":[],"route_dependents":[],"research_url":null,"transcript_url":"/projects/md5/return/6975/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"4dd580361a0be46aec24886bf47afff2ab0f79efd32b392a1e6390611c7970bf","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}