{"id":7107,"job_id":14021,"problem_id":6,"lane_id":34,"type":"measure","user_id":90,"model":"muse-spark","provider":"unknown","report_md":"# All-zeros measure: 400B il8-vs-il4 head-to-head is uniform; il8 1.13x faster loaded, best 9\n\nJob 14021, measure, all-zeros. Improvement claim: none. No\nlonger leading-zero prefix was found, and none is claimed.\nNothing goes to the results page. Best score this run is 9.\n\n## Hypothesis\n\nThe il8 tunnel screens candidates on the first output word h0\nbefore running the full MD5, and its message construction (a\nfixed 64-byte block with a varying nonce region) could in\nprinciple bias h0. If it biased h0, a fresh cell's score\ntallies would deviate from uniform random search; if it holds,\nil8 is a pure throughput play with no distributional edge, and\nits rate advantage over il4 is the whole story. The decisive\ntest is fresh counted coverage with tallies compared against\nuniform expectations, plus a same-window il4 control cell on\nthe same machine under the same load.\n\n## Experiment\n\nTwo fresh disjoint 200B cells, one pinned core each, same\nwindow:\n\n- il8, core 0, range [13789900000000, 13989900000000): binary\n  md5tun_il8, md5 d85e47f75ab73cec72788064df75d154, identical\n  to the job 13994 build; self-test PASS this cycle (matches\n  scalar and il4 on 100,096 candidates). Result: 204.99 M/s\n  loaded, total 200,000,000,000, exit via [done] line, best 9.\n- il4, core 1, range [13989900000000, 14189900000000):\n  binary md5tun_tally, md5 d9dd2aa189ddcf40876a8dbcf7123d77;\n  self-test PASS this cycle (matches scalar and il2 on\n  100,000 candidates). Result: 180.79 M/s loaded, total\n  200,000,000,000, exit via [done] line, best 8.\n\nLoaded head-to-head: il8 is 1.134x il4 on this machine in\nthis window (204.99 vs 180.79 M/s), narrower than the solo\nratio (321.48 vs ~240 M/s, 1.34x) because two-core contention\ncompresses both. Both rates are loaded rates, not solo peaks.\n\n## Results\n\nTallies against uniform expectations for 200B candidates:\n\n- il8: tally2 781,267,125 vs 781,250,000 (z +0.61), tally4\n  3,053,439 vs 3,051,758 (z +0.96), tally6 11,789 vs 11,921\n  (z -1.21), tally7 717 vs 745 (z -1.03), tally8 51 vs 46.6\n  (z +0.65), tally9 3 vs 2.91. Uniform at every rung.\n- il4: tally2 781,333,322 vs 781,250,000 (z +2.99), tally4\n  3,050,802 vs 3,051,758 (z -0.55), tally6 11,859 vs 11,921\n  (z -0.57), tally7 721 vs 745 (z -0.88), tally8 44 vs 46.6\n  (z -0.38). The il4 tally2 is the only rung beyond 2 sigma\n  in 10 rung-cells; the il8 cell's tally2 is +0.61 sigma on\n  the adjacent range, so it is not a shared nonce-region\n  effect, and one ~3 sigma draw among 10 correlated rung\n  counts is not evidence of bias. No tally9 is reported for\n  the il4 cell: this build of the tally binary prints no\n  tally9 field, so none is claimed.\n\nBest candidates, all hashlib-verified over the raw candidate\nbytes, 0 mismatches (9 new-best lines per cell):\n\n- il8 best 9: digest 000000000f12e9ee0b8ec03902c49201, input\n  1e8141f6e074baac35340cf3e5e4945a4ba1ab2671d2d0922aa9d286cc6985a5897c90b0394ef63085bbf78d0b732b48cbd27d35.\n  Two further score-9 hits in the same cell, also verified:\n  digests 000000000ea869b2f6e387bf55c4dc68 and\n  0000000002de22919d0b4d2868113a8a.\n- il4 best 8: digest 0000000054eca5b809752d3648218a96, input\n  9968a06d2c5f3139449044b933534d84ae89580947b3e3145e3e6a14f0f511ac5b53fdac4b5d57886b225b0f6c362d53bde48bc8.\n\nNext disjoint all-zeros start nonce: 14189900000000.\nFresh counted coverage this cycle: 400B. Nothing scored 13\nor above; nothing to the results page. Sub-13 hits are\ntallied in the zeros tally only.\n","patch":null,"cpu_hours":0,"hashes":{},"author_rung":null,"status":"pending","final_rung":null,"created_at":"2026-11-07T16:55:10.309Z","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 14021, measure, all-zeros)\n\nCopy md5tun_il8 from the job 13994 work dir (md5\nd85e47f75ab73cec72788064df75d154) and md5tun_tally from the\njob 13721 work dir (md5 d9dd2aa189ddcf40876a8dbcf7123d77).\nSelf-test both before launch. Run one pinned core each:\ntaskset -c 0 ./md5tun_il8 13789900000000 200000000000 120 >\nil8_cell.log 2>&1; taskset -c 1 ./md5tun_tally\n13989900000000 200000000000 120 > il4_cell.log 2>&1.\nVerify with hashlib over the raw candidate bytes from the\n[new best] lines (the binaries hash raw 64-byte candidates,\nnot ASCII hex). Next disjoint start nonce: 14189900000000.","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":"2026-11-07T16:55:10.309Z","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/7107/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"313cbb1791d29ad640b5f0d246ca893e8001cfb4fb925d13a596b8cb320714bb","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}