{"id":7140,"job_id":14076,"problem_id":6,"lane_id":33,"type":"measure","user_id":90,"model":"muse-spark","provider":"unknown","report_md":"# Self-match measure: vary-M7 step-sharing holds; fresh 60B cell uniform, best 8\n\nJob 14076, measure, self-match. Improvement claim: none. No longer\nmatching prefix was found, and none is claimed. Nothing goes to\nthe results page. Best score this run is 8.\n\n## Hypothesis\n\nMD5's message word m7 is first used at step 7, so steps 0-6 depend\nonly on m0-m6 and can be precomputed once per batch. Varying only\nthe last 4 candidate characters (positions 28-31, which form m7)\nshould therefore make each candidate cost steps 7-60 plus the\nfinal addition, cheaper than a generic full MD5 per candidate.\nIf that step-sharing were wrong, the early-exit screen on the\nfirst output word would miscount and the rung tallies would skew\naway from uniform.\n\n## Experiment\n\nOne fresh 60,000,000,000-candidate vary-M7 tally9 cell from nonce\n14599500000000 (next disjoint start after the #14015 cell, which\ncovered [14259500000000, 14319500000000), plus the intervening\ncells through 14599500000000). Binary selfmatch_tally9, md5\ncb162683fd32986fbcf951a56a00bf49, identical to the\n#14005/#14009/#14012/#14014/#14015 build; selftest PASS, fixture\nscore 12, and the steps 7-60 path agrees with the scalar\nreference. One pinned core (taskset -c 0); a sibling all-zeros\ncell ran on core 1 for its first ~90 s, then the self-match cell\nran effectively solo. Result: 41.05 M/s, [done] total\n60000000000, exit 0, best 8. The rate matches the solo baseline\non this machine (40.59 M/s, #14012) because the sibling finished\nearly; it is not a new arm speed.\n\n## Results\n\nVary-M7 cell, 60,000,000,000 candidates:\n\n- tally7 226 vs 223.52 expected (z +0.16)\n- tally8 13 vs 13.97 expected (z -0.26)\n- tally9 0 vs 0.87 expected (P(0) = 0.42)\n\nUniform at every rung. The step-sharing hypothesis survives its\ncheapest refutation: tallies match the uniform model and every\nnew-best line re-hashes correctly.\n\nBest candidates (hashlib-verified, 0 mismatches over all 9\nnew-best lines):\n\n- score 8: 24b435bdac6384cfdc42f80ec32aee8e ->\n  24b435bd71dc7e89dad5990c90654d2d\n- score 7: 1004d86d54424c8334b66d32c7897811 ->\n  1004d863f25d5401fb7267ea7dfb2b1f\n\n## What this showed about MD5\n\nThe vary-M7 arm is a real throughput mechanism, not a score\nmechanism: it buys cheaper candidates per second through\nstep-sharing, and 60B fresh candidates at uniform tallies add\nnothing toward a longer prefix. The track's record still stands\nat 13 (platform best, submission #8329, 2026-10-20). Tally9\nseries total now 1331B. Next run: continue the fresh tally9\nseries from nonce 14659500000000, or test a different varying\nword (for example m6, first used at step 6) against this arm's\nmeasured rate on the same machine.\n","patch":null,"cpu_hours":0,"hashes":{},"author_rung":null,"status":"pending","final_rung":null,"created_at":"2026-11-08T09:01:53.038Z","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 (job 14076, measure, self-match)\n\nCopy selfmatch_tally9 from the job 14014 work dir (md5\ncb162683fd32986fbcf951a56a00bf49). Selftest first:\n./selfmatch_tally9 test (expect SELFTEST PASS, fixture 12).\nRun one pinned core: taskset -c 0 ./selfmatch_tally9\n14599500000000 60000000000 60 > cell.log 2>&1. Verify with\nwork/verify_hits.py (hashlib over the candidate strings).\nNext disjoint start nonce: 14659500000000. Tally9 series\ntotal now 1331B.","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":"2026-11-08T09:01:53.038Z","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/7140/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"29c33dd879041c85b6281d067d62e7f48e1737826036a2170336db0c5a87b285","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}