{"id":7035,"job_id":13846,"problem_id":6,"lane_id":34,"type":"measure","user_id":90,"model":"muse-spark","provider":"unknown","report_md":"# All-zeros measure: long esfg6 harvest holds 1.44x over il4 under load; both arms uniform, best 9\n\nJob 13846, measure, all-zeros. Improvement claim: none. The best\ncandidate reached score 9, below the 13 floor, so nothing goes to the\nresults page.\n\n## Hypothesis\n\nThe clang esfg6 arm sustains its ~345 M/s solo rate over a long\nfresh-seed run and its rung-9+ yield stays uniform, and the il4\ncontrol cell stays uniform; the CS25 lead over il4 holds under load.\nA 96-base esfg6 harvest against a 200B il4 tally cell, one core each,\ntests both: loaded rates near 0.78x solo on both arms (symmetric core\ncontention) and tallies within about 2 sd of uniform at every rung.\n\n## What was run\n\n- esfg6 arm: surviving clang-built binary (selftest PASS again this\n  cycle), fresh seed 13846, 96 bases, 1 thread pinned to core 1.\n  412,316,860,416 trials (96 x 2^32 exactly), wall 1529.91 s, rate\n  269.50 M/s loaded.\n- il4 arm: production tally binary, fresh disjoint nonce\n  12489840000000 (verified disjoint: prior cell ran 12429840000000 +\n  60B, done line confirmed), 200,000,000,000 candidates, log interval\n  120 s, 1 thread pinned to core 0. 186.57 M/s loaded, [done] line\n  total=200000000000, exit 0.\n\n## Results\n\n- Rate: loaded ratio esfg6/il4 is 269.50/186.57 = 1.444x, matching\n  the solo ratio 344.75/240.64 = 1.433x. Load factor vs solo: esfg6\n  0.782x, il4 0.775x, symmetric core contention, not an arm effect.\n  The CS25 lead holds under load.\n- esfg6 yields (N=412.32B): score >= 6: 24,642 observed vs 24,576.0\n  expected (z +0.42); >= 7: 1,474 vs 1,536.0 (z -1.58); >= 8: 93 vs\n  96.0 (z -0.31); >= 9: 4 vs 6.0 (z -0.82). Best 9, no 10.\n- il4 yields (N=200B): score >= 2: 781,219,190 vs 781,250,000\n  (z -1.10); >= 4: 3,053,714 vs 3,051,758 (z +1.12); >= 6: 11,896 vs\n  11,921 (z -0.23); >= 7: 702 vs 745.1 (z -1.58); >= 8: 44 vs 46.6\n  (z -0.38). Best 9, no 10.\n- Pooled esfg6 score >= 9: this run (4 vs 6.0) plus #13833 (1 vs\n  3.0) = 5 vs 9.0 expected over 618B trials (z -1.33). Still too thin\n  for a yield ratio; the shared-range comparison stays open (see job\n  #13845 this cycle).\n- Verification: all 24,642 esfg6 HIT lines hashlib-verified from the\n  log (0 digest mismatches, 0 score mismatches); all 10 il4\n  best-progression hits hashlib-verified (0 mismatches). The esfg6\n  binary logs only score >= 6 hits, so the summary's observed counts\n  below rung 6 repeat the >= 6 count and are not used here. The il4\n  binary reports no >= 9 tally count; best progression only.\n- Best candidates, all score 9 (sub-13, tallied only):\n  - il4: efc0efbd887aa80b1985fccf05fead5cb13adfc265997d5768e276e408d97d9b2ce9e9ca6e04863c293ed10c3d0fee106742942f, digest 000000000d8df34350a3393ff57b9102\n  - esfg6 base 14: d9a3085f392847f36c2f0ccd87636c3ca84e62eb026690ef8106ccc6f5e88a49f2075b2db50be6d39a5572704f4c828e1bc1d1c3, digest 00000000039a275f305bc92dc37a827d\n  - esfg6 base 49: daa4174f094049048f6b88299b8e3df0b84d53ebf1f23c5f00e937e5a1e606bdd41ecc70d85c7657ab566bc04f4c828e210e0ba0, digest 000000000b15d2199539331cb4b5c4a9\n  - esfg6 base 59: 1b83c203033524f01f4521d42f8cc448e58d2b5f50672d32f14d4ccb2d7662165277b6b588490c144f0ef7c58d6a58bfc41373f1, digest 000000000d2509069133aa3b836794e0\n  - esfg6 base 88: 8bc3ca03fc1084fcd2a90e9cae5adc6e618d2b54a8e7e99d1fd2fd4a756ce100537fab27f4104b034c0e78dd8d6a58bfa97e13b0, digest 0000000006a9cff9f132fd3be5786154\n\n## Counted totals\n\nPer arm, as of this cycle: il4 tally cells about 1280B after this\ncell (1080B + 200B); next disjoint all-zeros start nonce\n12689840000000. Esfg6 trials 206B (#13833) + 412B (this run) = 618B.\nThe two arms keep separate totals; they cover different input\nfamilies.\n\n## What this showed\n\nPer the brief, this run is reported as a baseline measurement: it used\nonly known tools (the production il4 tally binary and the surviving\nclang esfg6 binary) and found no structural shortcut. The esfg6 arm is\na real 1.44x throughput win over il4 under load with uniform yield at\nevery rung measured, so volume on the faster arm remains the cheapest\nroute to the record, and the remaining open number is the rung-9+\nyield ratio at matched N.\n\n## Next run\n\nOne more 96-base esfg6 harvest from a fresh seed, pooled to 240 bases\n(about 1.03T trials, expected score >= 9 about 15.0) against the\npooled il4 arm, per job #13845's cheapest experiment. Keep the il4\ntally cell as a periodic uniformity control from the next disjoint\nnonce 12689840000000.\n","patch":null,"cpu_hours":0,"hashes":{},"author_rung":null,"status":"pending","final_rung":null,"created_at":"2026-11-06T16:07:39.113Z","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, one core per arm:\n\n1. Verify the next disjoint il4 start nonce from the prior cell's done\n   line: start + total, never the notes number alone. Next disjoint\n   start after this cycle is 12689840000000.\n2. esfg6 arm: run the surviving clang-built binary with selftest first\n   (`cs25_clang test`), then `taskset -c 1 cs25_clang search2esfg6\n   <fresh seed> 96 1`. One base is 2^32 trials.\n3. il4 arm: `taskset -c 0 md5tun_tally <start_nonce> 200000000000 120`.\n4. Verify hits with runs/job-13846/work/verify_zeros.py on both logs:\n   it checks every esfg6 HIT line and every il4 [new best] line with\n   hashlib.md5 and the true leading-zero score.\n5. Compare tallies against N x 16^-k expectations per rung; compare\n   the loaded esfg6/il4 rate ratio against the solo ratio.","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":"2026-11-06T16:07:39.113Z","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 what makes the first output word of MD5 small, and use it to reach more leading zeros than generic search would with the same compute. Ideas to test: freedom from extra message blocks, neutral bits and message modification from collision attacks applied to the output instead of a difference, early abort on the final additions. 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/7035/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"a60c2ae1e0403ae17cb8bb622b8b01147f0bee881c766f09f888cd13fd762f63","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}