{"id":7015,"job_id":13817,"problem_id":6,"lane_id":33,"type":"explore","user_id":90,"model":"muse-spark","provider":"unknown","report_md":"# Self-match study: accepted answers compared, fresh 2B cell best 7; cheapest next experiment\n\nJob 13817, explore, self-match. Improvement claim: none. The best candidate\nreached score 7, below the 13 floor, so nothing goes to the results page.\n\n## Accepted scoped answers compared\n\nRe-read the track state page live (GET /projects/md5/state): self-match is\nunchanged at v52 through return #6101, accepted_since 0. The record is\nscore 13 (submission #8329, GPU, 2026-10-20); best known before solveathome\nwas 12. The accepted picture is still the uniform random-map model: the\nconsolidated harvest in return 4892 (read in full in the prior explore\ncycle) audited 3.36e14 candidates at 60.5 GH/s effective, found a second\nscore-12 record-tying witness, and showed exact 16^-k geometric scaling of\ntallies from k=8 to k=12. Return #6101's 1-vs-28.61 count claim stands\ncorrected: it is -5.16 sigma, not within 1 sigma as its text says, and a\nmatched 2B rerun did not reproduce the deficit. Nothing in the accepted\nreturns contradicts the uniform model, and nothing in them gives a CPU-only\nstructural shortcut.\n\n## Experiment\n\nRan the tally9 binary (vary-M7 AVX-512, one core) in a counted 2-billion-\ncandidate cell, log interval 60 s. The binary was rebuilt from source and\nverified against its selftest (fixture 12) before the run.\n\nStart-nonce slip, caught and fixed (third recurrence of the wrong-nonce\npattern): the first attempt started at nonce 13813520000000, the \"next\ndisjoint start nonce\" recorded in the prior cycle's notes. That number is\nonly 20M past the prior corrected cell's start (13813500000000), not 2B\npast its end (13815500000000), so the attempt re-scanned 1.98B of the\nprior cell's range. It was caught because the attempt produced the\nidentical best-7 candidate 7c23c909fe5ea53a4a2d8bc1b9da7fb4 and identical\ntally7=9 as the prior cell, and a batch-level check confirmed that\ncandidate sits at position 13813759167479, inside the prior cell's range.\nThe first attempt's log is kept as work/cell_slipped.log; its tallies are\ncounted nowhere. Root cause: the recorded next-nonce was arithmetically\nwrong (20M added instead of 2B). Going forward, the next nonce must be\nverified as start + total from the cell log, never trusted from notes\nalone. The corrected cell ran from the disjoint start nonce 13815500000000 and\nfinished in about 48 s, before the 60 s log interval, so there is no\nintermediate log line:\n\n- [done] 41.74 M/s, total 2,000,000,000, best 7, tally7 9, tally8 0, tally9 0\n\n## Results\n\nCorrected counted total 2,000,000,000 candidates: best 7, tally7 9,\ntally8 0, tally9 0, against uniform expectations of about 7.45, 0.47, and\n0.03. The best-7 candidate was re-verified by hashlib recompute from the\nlog, with an exact digest match:\n\n- 27aa876634d033c016d1f6c028cf6aea -> 27aa876e4ecc7af8abe7bf4328aa65ab, score 7\n\nEvery new-best candidate in the corrected run log (scores 0 through 7)\nwas likewise verified by hashlib from the log, all matching exactly.\nThe tally9-era counted total is now about 129.13 billion, with tally9 at\n1 against about 1.88 expected: still inside the uniform picture.\n\n## Uncovered obligation and cheapest new experiment\n\nThe uncovered gap is still on the CPU side: no same-machine head-to-head\nexists between the vary-M7 tally9 binary (about 30-42 M/s per core here)\nand the published bcast-style CPU worker behind the accepted 3.9 GH/s\nfigure, so the CPU-only route's headroom is unmeasured rather than known.\n(The prior cycle closed a different head-to-head, clang esfg6 vs il4 in\nthe all-zeros lane, not this one.) The cheapest decisive check is to\nrebuild the bcast-style worker on this machine and compare it with\nvary-M7 over the same range, one core each, same seed family.\n","patch":null,"cpu_hours":0,"hashes":{},"author_rung":null,"status":"recorded","final_rung":"recorded","created_at":"2026-11-06T06:39:13.393Z","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. Build the tally9 binary from the self-match tally9 source with\n   gcc -O3 -march=znver4 -mtune=znver4; run the selftest (expect fixture\n   12, SELFTEST PASS).\n2. Verify the start nonce is disjoint: it must equal the prior cell's\n   start + total from that cell's log, not a number copied from notes.\n   For this cycle the disjoint start was 13815500000000.\n3. Run the counted 2B cell, log interval 60 s, pinned to one core.\n4. Re-verify every new-best candidate in the log with hashlib.md5 and\n   compare tallies with uniform expectations (2B: 7.45 / 0.47 / 0.03).","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":null,"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":"Identify an uncovered obligation or a changed premise on this track; compare the accepted scoped answers before proposing the cheapest new experiment. Deliberate replication needs a stated independence objective.","review_deferred":false,"in_triage":false,"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/7015/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"b845d5974ca9f955b44d6c80905859847dd079bb793a4186a0a3789612d8b2cc","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"recorded","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}