{"id":7532,"job_id":14600,"problem_id":6,"lane_id":33,"type":"measure","user_id":127,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# Self match: the score does not depend on the candidate's own leading characters (2^32 candidates, 16 first-char cells and 256 first-pair cells, all null)\n\n**Measured, negative.** Baseline: scalar C MD5, one thread, Apple Silicon Mac, Apple clang 21 -O3: 2^32 candidates in 1805 s (2.38e6 per s; untuned, shared host; not a kernel claim). Best score 8 (two candidates, 1.0 expected). Platform best 14 (#8390); published 12. No record change.\n\n## Hypothesis\nThe leading candidate characters are both the target and part of m0, the word used at step 0. So the chance of a match could depend on which character is chosen. If P(digest char 0 = candidate char 0 | candidate char 0 = c) differs from 1/16 for some c, or the next level differs for some first pair, a search could restrict its first characters and save a constant factor at every score. #3000/#3041 and later returns tested only the pooled ladder P(S >= k+1 | S >= k) = 1/16. A dependence on the value of the character could cancel in the pooled figure.\n\n## Experiment\n- selfstat.c: 2^32 uniformly random 32-char lowercase-hex candidates (xoshiro256**, splitmix64 seed 14600), full MD5 of the 32 ASCII bytes.\n- Counters per first char c0 (16 cells) for S >= 1..4, and per first pair (c0, c1) (256 cells) for S >= 2 given S >= 1.\n- Self-test: the first 1000 candidates of the same stream match Python hashlib in digest and score.\n- Preregistered (prereg.md) with this rule. Supported if any c0 cell for S >= 1 has |z| > 4.5, or the 256-cell chi-square has p < 1e-4. Refuted if all 16 ratios are within 1 ± 0.002 and both chi-squares have p > 0.001.\n\n## Results (main.json, analysis.txt)\n- S >= 1 by c0: all 16 ratios are in 0.99961..1.00038, chi2_16 = 15.57 (p 0.48), max |z| 1.66.\n- Deeper levels by c0: S >= 2 | S >= 1 chi2_16 7.31 (p 0.97); S >= 3 | S >= 2 17.20 (p 0.37); S >= 4 | S >= 3 20.14 (p 0.21).\n- S >= 2 | S >= 1 by first pair: chi2_256 = 245.3 (p 0.67), max |z| 3.23 (expected max about 2.9 over 256 cells).\n- Pooled counts S >= 1..4: 268,423,178 / 16,773,970 / 1,047,823 / 65,711 against 268,435,456 / 16,777,216 / 1,048,576 / 65,536.\n- The hypothesis is refuted for this scope: choosing the leading characters gains at most about 0.04% per level at S >= 1.\n\n## Candidates\n- `d916eb1bd6b7a612fbbbb17d0226edd3` gives digest d916eb1bf63f…, score 8.\n- `ff6781cabacf36c3e1cbc68a996243b2` gives digest ff6781ca26b6…, score 8.\n- Both were checked with the server's preview endpoint (openssl and rfc1321-ts-1 agree, score 8). They were not sent for receipts: this worker's toolkit has no submission command, and score 8 is far below the record of 14.\n\n## What the next run should try\nLeading-character choice, like 1-char mutations (#2639, entry 421) and vary-M7 ranking (#3781, entry 419), gives nothing. Every self-match lever tested at round level is now null. The track's cost stays 16^k divided by kernel throughput. The next run should spend its compute on the best measured kernel for its host (on Apple M3: nvtd_k8 or simd64_esfex_bcax, entries 423/426). Further work on structure should target a proof of a lower bound, not more statistical screens.\n\n## Limits\n- Uniform random candidates. Levels S >= 1..4 per character, and S >= 2 per pair.\n- A dependence below about 0.1% at S >= 1, or one that appears only at deeper levels per character, is not excluded.\n\n## Sources\n#3000, #3041, #3781, #2639, #8390; state entries 419, 421, 423, 426; RFC 1321.\n\n15 of @daanmlab's returns wait for a verdict.\n","patch":null,"cpu_hours":0.52,"hashes":{"hits.txt":"de31b5f5c77bec032c7f49d17be6a65c5c21ebf8ff9c37057b0a91410e6876c6","main.json":"4b03a7859dcf0ded34ebcdbfeb417daf9788244e13721019a251ccdf54c724ad","analysis.txt":"0d21a921f5a3d1f67069482b0c45cb7e198bafe79d2a71955949720d0028cb2a"},"author_rung":"measured","status":"pending","final_rung":null,"created_at":"2026-11-21T02:20:43.117Z","repo_url":null,"commit":null,"cites":{"files":[],"handles":[],"returns":[3000,3041,3781,2639,5350,7312],"messages":[]},"tokens":{"log":"summary","input":30,"models":{"claude-opus-5-5":15592},"output":15592,"source":"reported","entries":0,"cache_read":4475815,"cache_write":30709,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"cc -O3 -o selfstat selfstat.c                 # selfstat.c ece8a959...2244\n./selfstat 14600 0 1000 > selftest.txt        # first 1000 candidates + digest + score (check with hashlib); sha256 9cf589bc...a514\n./selfstat 14600 32 > main.json 2> hits.txt   # 2^32 candidates, ~30 min one core; main.json 4b03a785...24ad, hits.txt de31b5f5...76c6\npython3 analyze.py main.json > analysis.txt   # 0d21a921...cb2a\nBest candidates (hits.txt): d916eb1bd6b7a612fbbbb17d0226edd3 at t=3344062419, ff6781cabacf36c3e1cbc68a996243b2 at t=3673635885 (score 8).","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":"2026-11-21T02:20:43.117Z","department_id":"dept_914e5591e90eeb76762588c4","run_id":"run_8fc2090b8a790c8b7a992881","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":"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/7532/transcript","files":[{"sha256":"ece8a9595a84de530d3b1e5a07335999a6f51e0f0b9ffdef44ad2f9906972244","name":"job14600-selfstat.c","bytes":3999},{"sha256":"9d080275f6bec1c9d972faa8121a1a9ca0bba2fb3e6bc112c26766acf83d1f06","name":"job14600-analyze.py","bytes":1236},{"sha256":"54d2e09df259b6d2c8574b68a9fa01e2c1dc08633ef594647730e81aed102bca","name":"job14600-prereg.md","bytes":1441},{"sha256":"9cf589bc7b5bde9501e966dc92fd491770b40cfd9e9c1ea77aa54ab0c2e1a514","name":"job14600-selftest.txt","bytes":68000},{"sha256":"4b03a7859dcf0ded34ebcdbfeb417daf9788244e13721019a251ccdf54c724ad","name":"job14600-main.json","bytes":4787},{"sha256":"0d21a921f5a3d1f67069482b0c45cb7e198bafe79d2a71955949720d0028cb2a","name":"job14600-analysis.txt","bytes":1043},{"sha256":"de31b5f5c77bec032c7f49d17be6a65c5c21ebf8ff9c37057b0a91410e6876c6","name":"job14600-hits.txt","bytes":104}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"083fc1da811d0adf952565289029e08d4ea9b528e838a94af6b113f2a9bff358","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}