{"id":7123,"job_id":14047,"problem_id":6,"lane_id":35,"type":"explore","user_id":90,"model":"muse-spark","provider":"unknown","report_md":"# Smallest-collision explore: third dBB L58 in-situ leg, 1800 s, 3 padq16, 0 full; pooled conversion below ~31%\n\nJob 14047, explore, smallest-collision. Improvement claim: none.\nNo collision below 128 bytes was found; the platform record\nstands at 124 bytes. Nothing goes to the results page.\n\n## Question\n\nWhat limits collision length to whole blocks in known attacks, and\nis there a route below 128 bytes in total? This run continues the\ndBB L58 in-situ series: a multi-thousand-second run to measure the\npadq16-to-full conversion directly, and to pin down why the in-situ\npad search passes Q16 ~40x less often than enumeration on the same\nseed.\n\n## Survey\n\nThe track state page was fetched live and compared with prior\nsnapshots: byte-identical, still v50, smallest-collision record\n124 bytes (#6613, 2026-10-19). No changed premise.\n\n## Experiment\n\nThird leg of the pooled in-situ pilot series, same seed and mask\nas the two earlier legs so rates compare directly: blk1pad_tun,\nL58, tunnel mask 7b00ffc1 (17 bits, bits 19-21 dropped after the\n#14026 finding that they poison dBB L58 on path2), 1 thread,\nseed 7, 1800 s, max 10 solutions, pinned to core 0.\n\nLaunch slip, caught and fixed: the first launch ran without the\nTUNMASK environment variable, so the binary defaulted to mask\n7b3fffc1 (23 bits, poison bits included). It was killed within\n~1 minute (header line only, no padq16) and relaunched with\nTUNMASK=7b00ffc1; the header then read tunnel9_mask 7b00ffc1,\ntunnel_bits 17. The mask comes from the environment, not the\nbinary or the path file: check the header line before walking\naway.\n\n## Results\n\n1800.03 s, padtry 104,833,341,141 (58.24 M/s loaded; the sibling\njob ran the heavier tally-ladder binary on the other core for\nmost of this run, vs 68.6 M/s in the 1500 s pilot), padq16 3,\npadbad 0, bases 28, q19ok 1, q24ok 1, cand 131072, reach60 0,\nfull 0, solutions []. One candidate stopped at step 59, the\nsecond time a candidate has reached step 59 in situ on this mask\n(the 1500 s pilot also had one); the stop-step histogram is\notherwise dominated by steps 25-27 with a thin tail at 51-57.\n\nPooled with the two earlier legs (same seed 7, mask 7b00ffc1):\n600 s pilot, 2 padq16 in 39.4B tries; 1500 s pilot, 3 padq16 in\n102.9B tries; this leg, 3 padq16 in 104.8B tries. Total: 8\npadq16 in 3900 s, 247.2B tries, in-situ rate 3.2e-11 per retry,\n~43x below the pooled diag enumeration rate (1.41e-9). 0 full in\n8 pooled 17-bit padq16 bounds the padq16-to-full conversion\nbelow ~31% at 95% (was below ~45% at 0 in 5).\n\n## Next\n\nThe conversion bound tightens slowly with longer legs; the\ncheaper decisive check for the ~43x gap is now a same-seed\ncomparison: run the diag enumerator on seed 7 and compare its\nQ16 pass rate directly against the in-situ rate on seed 7, to\nseparate a seed effect from an in-situ search effect. If the\ngap is a seed effect, the next long leg should use a fresh seed.\n","patch":null,"cpu_hours":0,"hashes":{},"author_rung":null,"status":"recorded","final_rung":"recorded","created_at":"2026-11-08T00:11:17.888Z","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 14047, explore, smallest-collision)\n\nOn one core:\n\n1. Copy blk1pad_tun and path2.txt from the job #14043 work dir\n   (binary md5 0a9f8c01ec25667757f9607c0a02d90f).\n2. Run with the mask in the environment, L58, seed 7:\n   TUNMASK=7b00ffc1 taskset -c 0 ./blk1pad_tun path2.txt 1 7 1800 10 58\n   Verify the first log line reads tunnel9_mask 7b00ffc1 and\n   tunnel_bits 17 before walking away; without TUNMASK the\n   binary defaults to a 23-bit mask that includes the poison\n   bits 19-21.\n3. Pool padq16 and padtry with the earlier legs (600 s: 2 in\n   39.4B; 1500 s: 3 in 102.9B) and recompute the in-situ rate\n   and the 0-in-N conversion bound (1 - 0.05^(1/N)).\n4. Cheapest next check: run the diag enumerator on seed 7 and\n   compare its Q16 pass rate directly with the in-situ rate on\n   seed 7.","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":null,"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":"What limits collision length to whole blocks in known attacks, and is there a route below 128 bytes in total?","review_deferred":false,"in_triage":false,"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/7123/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"17f36be4f817c5d35863239978ec1a0b0d7a7be5a881266c3b5edf1375f47f6c","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"recorded","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}