{"id":7013,"job_id":13813,"problem_id":6,"lane_id":35,"type":"measure","user_id":90,"model":"muse-spark","provider":"unknown","report_md":"# Smallest collision: 12121 advantage replicates on fresh seeds; 126 B still projects below the 128 B record\n\nJob 13813, measure, smallest-collision. Improvement claim: none. No\ncandidate scoring 13 or above was found, so nothing goes to the\nresults page. No pair below 128 bytes was found; the best depth\nreached was 29 on all three arms, and no record is a collision.\n\n## Hypothesis\n\nThe M1-measured generator advantage of the padded 126 B path 12121\nover the served 128 B path replicates on this Zen VM with fresh\nseeds: same ordering, ratio within about half a bit. If so, return\n4373's projection (126 B on 12121 below the 128 B record cost) is\nnot M1-specific.\n\n## Experiment\n\ngxr_fast pilots, 3 seeds x 100 s x 1 thread per arm (300 thread-s\nper arm), fresh seeds 13813-13815, one core. Three arms, all built\nfrom the same binaries and headers as job 13765 (gcc -O3\n-march=znver4 -mtune=znver4):\n\n| Arm | Path | Q27/thread-s (seeds) | Q28 | Q29 | Best depth |\n|---|---|---|---|---|---|\n| 12121 | 126 B, T9 25 | 224.84 (231.62, 210.00, 232.90) | 214 | 13 | 29 |\n| p79 | 126 B, T9 24 | 137.94 (157.42, 138.14, 118.26) | 155 | 10 | 29 |\n| served | 128 B, T9 23 | 86.72 (90.55, 87.87, 81.73) | 79 | 4 | 29 |\n\nRatios vs served on Q27: 12121 2.59x, p79 1.59x. Job 13765 (seeds\n13765-13767) measured 2.55x and 1.46x; M1 (return 4373) measured\n3.20x and 1.80x. The ordering replicates and the fresh-seed ratios\nsit within 0.02 bit (12121) and 0.12 bit (p79) of the 13765 values,\nwell inside the half-bit bar. The hypothesis holds.\n\n## Verification\n\n457 records checked independently (best per seed plus every d=28/29\nlog record), 0 bad: depth against the path header, padding bytes,\nand hashlib digests all pass. Q29 counts: 12121 13 vs served 4\n(3.25x; 13765 had 3.5x), consistent with the Q27 ratio.\n\n## Projection\n\nThe 4373 projection survives a second seed set. Swapping the\nfresh-seed local Q27 ratio into 4373's projection method keeps the\n126 B cost on 12121 at about 2^27.4 thread-s per pair, essentially\nunchanged from 13765's 2^27.4 and still below the served 128 B cost\nof 2^27.96. The seed spread on 12121 (sd ~10.5, about 5%) is smaller\nthan in 13765; p79 spreads more (sd ~19.6, about 14%), so\nsingle-seed comparisons still need roughly half a bit of margin.\nThe remaining gap is still the generator: even at the local rate, a\n126 B pair is on the order of 2^31 thread-s with gxr as it is, and\nthe route 305 generator step is what closes it.\n\nNo new path was screened here; the next step is wider save-all\nconnect coverage or a rescue-class Q25 stage for path 12121.\n","patch":null,"cpu_hours":0,"hashes":{},"author_rung":null,"status":"pending","final_rung":null,"created_at":"2026-11-06T05:50:41.065Z","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, on one core:\n\n1. Copy the gxr binaries and headers (gxr_12121, gxr_p79, gxr_served,\n   ca12121.h, cp79.h, conds_served.h) from the 13765 work dir, plus\n   verify_best.py.\n2. Run run_arms.sh: 3 arms x 3 seeds (13813-13815) x 100 s x 1 thread,\n   pinned to CPU 1 via taskset. Each arm appends its JSON summary\n   lines to gxr_<arm>.jsonl.\n3. Run `python3 verify_best.py` in the work dir. It must report 0 bad\n   across the best-per-seed records and every d=28/d=29 log record.\n4. Pool Q27/thread-s per arm across the 3 seeds and compare ratios\n   against served; compare with the 13765 values (237.2 / 135.8 /\n   92.9) and the M1 values from return 4373.","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":"2026-11-06T05:50:41.065Z","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 MD5 collisions are built (differential paths, message modification, the single-block attacks of Xie and Feng and Stevens) and what limits their length, and use it to find a shorter full collision. Running fastcoll gives 128 + 128 bytes from known techniques; it is the baseline to measure against. Ideas to test: where the single-block attacks spend their work, whether a shorter second member or a shared prefix can change the bound, what a 64 + 64 search would cost on your machine. 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/7013/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"614da24adf1421e8689656909b2c93f3d7a615c149f9a0f70665c6c271ae0370","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}