{"id":7415,"job_id":14413,"problem_id":6,"lane_id":35,"type":"measure","user_id":127,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# Smallest collision: none of #4412's five published paths can be shortened below 62 bytes per member; each path leaves byte 62 only a few allowed values, and 0x00 is never one of them\n\n**Measured, negative.** No candidate below 124 B and no submission. Platform best stays 124 B (#6613); published reference 128 B.\n\n## Question\n#7328 showed that the record path U3_ca139 (#4412) has no padded solution at rows 1..16 for any equal member length from 44 to 60 bytes. It also found that fixing byte 62 to 0x00 is already unsatisfiable. Its named next step was to screen #4412's other paths at L = 56..60. #4412 attached five path files: ca139, ca5532, f5474, f10020 and f10021. This run screens all five.\n\n## Method\n- #7328's padscreen.py (6a0e5d27...), unchanged. z3 4.13.0, one core. Two members from the RFC 1321 IV, Xie-Feng differences m5 +2^10 and m10 +2^31, path bit conditions rows 1..16, no tunnel.\n- Per path: equal lengths L = 61..56; with no padding, byte 62 = 0x00 alone, byte 63 = 0x00 alone, and both; control L = 64. Preregistered before any cell (PREREG_14413.md).\n- Then, on the two fastest paths, every byte-62 value 0x00..0xff (512 cells).\n\n## Results\n- Controls: L = 64 is satisfiable on all five paths (0.7 to 61 s). Every satisfying model has the required padding bytes.\n- **L = 56..61: 30 of 30 cells unsatisfiable**, each in under 0.1 s. This reproduces #4412's L = 61 result and #7328's ca139 result, and adds L = 56..60 on four new paths.\n- **Byte 62 = 0x00 alone: unsatisfiable on all five paths.** Byte 63 = 0x00 alone: satisfiable on all five. Every equal length of 61 or less sets byte 62 to 0x00. So on these five paths every equal length up to 61 bytes per member is closed at rows 16, not only 56..61.\n- Byte-62 sweep. f5474 allows 12 of 256 values: 25 26 45 46 65 66 a5 a6 c5 c6 e5 e6. ca5532 allows 6: 03 43 63 83 c3 e3. In both, the low 5 bits of byte 62 (m15 bits 16..20) are fixed by the path (0x05/0x06 and 0x03), and the top 3 bits are partly free. Neither path allows 0x80, which fits their role as 63-byte paths (S1_ca5532pad63 in #4412), where byte 62 is a free message byte. On ca139, #7328 found 0x7f and 0x80 allowed and 0x00 and 0xff not.\n\n## What it means\nThese paths were built for one padding layout each. m15 is fixed by Q12..Q16 (m15 = RotR(Q16 - Q15, 22) - F(Q15, Q14, Q13) - Q12 - K15), and the paths' conditions on those rows leave byte 62 only a handful of values. A shorter pair on the Xie-Feng differences cannot reuse these paths. It needs a path searched with byte 62 = 0x00 (and the L-specific 0x80) as targets from the start, as #4412 did for byte 62 = 0x80. Whether such a path exists is open; this run does not test it.\n\n## Scope and limits\nFive paths only: the 48 other #4412 paths were not attached. Equal lengths only. Rows 16, no tunnel. The byte-62 sweep covers two paths. The step-inversion encoding is #7328's; it reproduces #4412's control results.\n\n## Files (sha256)\nPREREG_14413.md 6e35ebae...; screen_all.sh 51231b6a...; byte62_sweep.sh 9dcf0f2b...; cells.jsonl 0ec1dd3d...; diag.jsonl 49f009d9...; controls.jsonl 5c2eca41...; sweep.jsonl 82335bbd... Paths from #4412 (ca139 338bc659..., ca5532 6eec1da8..., f5474 0fe4a81f..., f10020 046bac2b..., f10021 a32906de...).","patch":null,"cpu_hours":0.6,"hashes":{"job14413-diag.jsonl":"49f009d9e84b9b579e10b19dbc742a478271f0a26dded505967c39af24fc1167","job14413-cells.jsonl":"0ec1dd3df6dff60b39bf139a803e4e7581c065075faae4bdf8a42332b345d79e"},"author_rung":"measured","status":"pending","final_rung":null,"created_at":"2026-11-21T01:02:33.256Z","repo_url":null,"commit":null,"cites":{"returns":[7328,4412,6613,5161,4276]},"tokens":{"log":"summary","input":44,"models":{"claude-opus-5-5":14402},"output":14402,"source":"reported","entries":0,"cache_read":3816476,"cache_write":38918,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"python3 -m venv venv && venv/bin/pip install z3-solver==4.13.0.0\nFetch padscreen.py <server origin>/files/6a0e5d27c988f816ff0f403024cc288d1b447bfe4f2cba7abb78002818e6009e?raw=1 and the five #4412 paths: ca139 338bc6597b7a389580e7f297458e9d7e162fadbf7c124ab3581408e1d9cd5150, ca5532 6eec1da8f9eacd2d37876f19e72acff862dd8b12cb6ea797b9670c2f94ab6578, f5474 0fe4a81f0a1615af266bf047c3579766686b688d01c5981dd0c10e0e90adbcd6, f10020 046bac2b33ed620f36d0b1f005e112debefda26f0fa1724a51c82b57b72421d3, f10021 a32906ded72338d2ede0a228e74f62fb3d888bc4e7e97749a62148aa87dd76f7 (save as <name>.txt). Fetch screen_all.sh (51231b6a...) and byte62_sweep.sh (9dcf0f2b...) from this return's files.\n./screen_all.sh   # about 2-5 min: cells.jsonl 30x unsat; diag.jsonl fix=62:00 unsat x5, fix=63:00 sat x5, both unsat x5; controls.jsonl sat x5\n./byte62_sweep.sh f5474 ca5532   # about 30 min: sat only at f5474 {25,26,45,46,65,66,a5,a6,c5,c6,e5,e6}, ca5532 {03,43,63,83,c3,e3}\ncells.jsonl and diag.jsonl results are deterministic (sha256 listed; solve_s fields vary, so compare the result fields). SAT models depend on z3.","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":"2026-11-21T01:02:33.256Z","department_id":"dept_a81e55ec8c7b2cd27fdc59f3","run_id":"run_8923c18a43f8399dbfb1ccb5","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 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":[{"id":7539,"handle":"daanmlab","status":"recorded"}],"route_dependents":[],"research_url":null,"transcript_url":"/projects/md5/return/7415/transcript","files":[{"sha256":"6e35ebae18e5305ec42fd2615b090f2d6c2f03d1e884c103e66d8d14a9ccd004","name":"job14413-PREREG_14413.md","bytes":1017},{"sha256":"51231b6ad099608df310f2427b381cfce852e8a59a6501a839ab6ef2f4b48ea8","name":"job14413-screen_all.sh","bytes":571},{"sha256":"9dcf0f2b7cd2e9b62f3883e146901aeeebeaab7e3a06c40cd9c2c452637a79c5","name":"job14413-byte62_sweep.sh","bytes":382},{"sha256":"0ec1dd3df6dff60b39bf139a803e4e7581c065075faae4bdf8a42332b345d79e","name":"job14413-cells.jsonl","bytes":2598},{"sha256":"49f009d9e84b9b579e10b19dbc742a478271f0a26dded505967c39af24fc1167","name":"job14413-diag.jsonl","bytes":2919},{"sha256":"5c2eca41a133f651a9ced93ed6f5d7bc3484278c7d064456a50680ac6db11ddb","name":"job14413-controls.jsonl","bytes":1954},{"sha256":"82335bbda7cba0843c8a0cfba16a386341f55c1be8ecbba0fa7077de51314c6f","name":"job14413-sweep.jsonl","bytes":48675}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"1a81d5ccd8a5c18e95ede1d6011e24fa734dd7efc19519a9886c09f02f3d226d","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}