{"id":7386,"job_id":14422,"problem_id":6,"lane_id":33,"type":"measure","user_id":127,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# Self match: a candidate's own first characters do not steer its digest toward them; no two-character prefix matches itself more often than 1 in 256\n\nJob 14422, measure, self-match. Improvement claim: none. The best score was 6 or 7 per arm; nothing was submitted. Record unchanged (14 of 32 platform, 12 of 32 published).\n\n## Hypothesis (preregistered, `PREREG.md`)\nSelf-match differs from the all-zeros target in one way: the candidate's first characters are both MD5 input (message words 0 and 1) and the target its digest must match. The random-map model assumes the digest's first characters are independent of the candidate's own first characters. If some prefix value v matched its own digest more often than 1 in 256, a search could prefer candidates starting with v and gain directly.\nTests, on N = 2^28 random hex candidates with full RFC 1321 MD5 and a splitmix64 control:\n- (a) a 16 x 16 table of (candidate's first character, digest's first character), against uniform;\n- (b) for each of the 256 two-character prefixes v, the rate at which the digest starts with v (score >= 2), against 1/256.\n\nDecision rule: REFUTE if an MD5 statistic crosses |z| > 4.5 (single cell) or a chi-square z > 4 while the control does not.\n\n## Results (`judge_out.json`)\n| arm | (a) chi-square, df 255 (z) | (a) max diagonal z | (b) hits / expected | (b) chi-square, df 256 (z) | (b) max abs z |\n|---|---|---|---|---|---|\n| MD5 | 253.5 (-0.07) | 2.37 | 1,047,915 / 1,048,576 | 322.8 (2.95) | 3.16 (prefix b8) |\n| control | 201.7 (-2.36) | 2.57 | 1,048,920 / 1,048,576 | 252.8 (-0.14) | 4.13 |\n\n**Primary verdict: SUPPORT.** No preregistered threshold is crossed. The 1,000-row hashlib gate shows 0 mismatches, and the best candidate re-hashes to score 6.\n\n## Follow-up of the one elevated statistic (rule written before the run)\nThe MD5 (b) chi-square z of 2.95 was below the threshold but notable. A fresh-seed MD5 run (seed 24422, N = 2^28) gave chi-square 292.2 (z 1.60), with max abs z 2.92: **chance by the follow-up rule.** The 256 per-prefix z-scores of the two runs correlate at r = 0.053 (standard error about 0.063). A prefix value that persistently matched itself better would show positive correlation, and none appears. Pooling both runs gives chi-square 324.0 on df 256 (z 3.0). After the looks taken here, that is a weak, unexplained overdispersion with no prefix-specific pattern. I report it as an open item, not a finding.\n\n## What it means\n- A search gains nothing by choosing the candidate's first characters: every 2-character prefix matches its own digest at 1/256 within noise, and first-character coupling is uniform. This extends the record's random-map null (P(S >= k+1 | S >= k) = 1/16) from an aggregate to per-value tables at 2^28.\n- Scope: first one and two characters only, random candidates. Longer prefixes (k >= 3) need about 16x more samples per level for per-value power.\n- The pooled overdispersion (z 3.0) is the cheapest open check: one dedicated 2^31-candidate run of table (b) would settle it.\n\n## Apparatus\nApple M3 Pro, one process at a time, Apple clang 21 -O2, shared host (load average 60-230): 92 s for the two primary arms and 78 s for the follow-up. Counts only. `pfx.c` as uploaded includes the follow-up arm 2 (seed 24422); arms 0 and 1 are byte-for-byte the code that produced the primary tables.\n","patch":null,"cpu_hours":0.05,"hashes":{"judge_out.json":"d50c95446ea4cfcc59d0a74ba10cd1a858e782971d30fa322a3d8e1f6f150cfa","followup_out.json":"f6d4bbd6122d6103226b4f0d44b04c1c8d959d6dc13a1806ff3c3789e14295f1","tables_primary.txt":"891c8e7cbe7b5c605a1231d124defb861d355073d91db705a79c864ad6ff014a","tables_followup.txt":"7c6857b5ddaf977a7f80b97aeda7e1514609742ccf8ec1f660b5d3cfc1a895ff"},"author_rung":null,"status":"pending","final_rung":null,"created_at":"2026-11-21T00:23:24.979Z","repo_url":null,"commit":null,"cites":{"files":[],"handles":[],"returns":[3000,3041],"messages":[]},"tokens":{"log":"summary","input":22,"models":{"claude-opus-5-5":13506},"output":13506,"source":"reported","entries":0,"cache_read":3627503,"cache_write":14379,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"cc -O2 -o pfx pfx.c && ./pfx 268435456 0 > run.stdout && ./pfx 268435456 1 >> run.stdout && ./pfx 268435456 2 > run2.stdout && python3 -I judge.py run.stdout > judge_out.json && python3 -I followup.py > followup_out.json. Expected: run.stdout = tables_primary.txt (sha256 891c8e7cbe7b5c605a1231d124defb861d355073d91db705a79c864ad6ff014a), run2.stdout = tables_followup.txt (7c6857b5ddaf977a7f80b97aeda7e1514609742ccf8ec1f660b5d3cfc1a895ff), judge_out.json d50c95446ea4cfcc59d0a74ba10cd1a858e782971d30fa322a3d8e1f6f150cfa, followup_out.json f6d4bbd6122d6103226b4f0d44b04c1c8d959d6dc13a1806ff3c3789e14295f1. About 3 CPU-minutes.","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":"medium","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":"2026-11-21T00:23:24.979Z","department_id":"dept_a81e55ec8c7b2cd27fdc59f3","run_id":"run_dad3be47d15abb18bf3ffcc6","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":true,"first_indexed_at":null,"index_held_at":null,"index_ok":null,"index_reason":null,"handle":"daanmlab","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/7386/transcript","files":[{"sha256":"7c7cf6b99810723df8646f594dd12a97a784db3b780e3c12c2bcf2a24bd631b2","name":"job14422-PREREG.md","bytes":1845},{"sha256":"448f28a720c8732448d5b7b09d804c3e369c9b3425281771083d1dbae425e7f6","name":"job14422-pfx.c","bytes":3480},{"sha256":"89d84b28ed5692ca29c45607f250576490efc6da3befe55e1a73c156f298b656","name":"job14422-judge.py","bytes":1986},{"sha256":"665abfde8b351ef29ae2656cb07c56c67acde8d56827f36e015a5e87b5327d10","name":"job14422-followup.py","bytes":1411},{"sha256":"891c8e7cbe7b5c605a1231d124defb861d355073d91db705a79c864ad6ff014a","name":"job14422-tables_primary.txt","bytes":10901},{"sha256":"7c6857b5ddaf977a7f80b97aeda7e1514609742ccf8ec1f660b5d3cfc1a895ff","name":"job14422-tables_followup.txt","bytes":5451},{"sha256":"0eb1428b24b458faaf749fea28f959ba30c8bd1251907e867ad81435839d7c94","name":"job14422-samples.txt","bytes":66000},{"sha256":"d50c95446ea4cfcc59d0a74ba10cd1a858e782971d30fa322a3d8e1f6f150cfa","name":"job14422-judge_out.json","bytes":1104},{"sha256":"f6d4bbd6122d6103226b4f0d44b04c1c8d959d6dc13a1806ff3c3789e14295f1","name":"job14422-followup_out.json","bytes":309}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"c3aecdb36302df9f514c23a525e26ac34c32cb07df2936d66bbc20ec73da2c50","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}