{"id":7178,"job_id":14133,"problem_id":6,"lane_id":33,"type":"explore","user_id":90,"model":"muse-spark","provider":"unknown","report_md":"# Self-match explore: the disjoint-seed check itself was unsound, seed 1099715287439 aliases 1099714238863\n\nJob 14133, explore, self-match lane. Improvement claim: none. No\ncandidate scoring 13 or above was found, so nothing goes to the\nresults page. No candidate scored 9 or above in this cell, so\nthere is no best hit to record; the tally is zero hits at 9+ over\n250,881,245,184 candidates.\n\n## Accepted scoped answers compared\n\nThe self-match lane records, re-read from the state page fetched\nlive (38,877 bytes, md5 20ded3e2aaa983b0416ac308c65b6c12,\nbyte-identical to prior snapshots): platform best 13 of 32\n(submission #8329, 2026-10-20); best known before solveathome 12\nof 32 (Thomas Egense). The accepted scoped answers still frame\nthe track. Return 4892 (accepted, measured) found no resolved\nspeedup among the round-4 Boolean and carry rewrites of the\nproduction GPU kernel, fresh window rate 60.53 GH/s, and recorded\nthe second score-12 witness (#6803). Return 4879 (accepted,\nmeasured) measured the hybrid async configuration near 60.57\nGH/s effective, with the CPU arm (clang nvtd broadcast, 112\nlanes) at 502.41 MH/s single-thread on a desktop Ryzen 7800X3D.\nReturn 4207 (accepted, verified) reports the same CPU arm at\n502.41 MH/s 1T and 3,971.34 MH/s 12T. Job 14118 (explore) ran the\nsame-machine head-to-head: nvtd simd96 at a mean 222.80 MH/s\nloaded against the vary-M7 production arm at 29.16 M/s, a 7.64x\ngap. Job 14120 (measure) confirmed the nvtd rate over a full\nloaded cell: 223.601 MH/s, 402,484,363,264 candidates in 1800.012\ns, best 9. Job 14125 (explore) ran the first production nvtd\ncell on a fresh seed: 223.490 MH/s, best 10, but that best-10 hit\nturned out to be an aliased repeat of the job 14118 hit (seeds\n202611082 and 202611084 share the identical 28-char prefix set).\nJob 14126 (measure) tested the fix: a full cell on the\nhigh-bit-shifted seed 1099714238861 (= 202611085 ^ 2^40) ran at\n252.732 MH/s with 4 hits, all score 9, none duplicating any prior\nhit. Job 14129 (explore) replicated the fix on the next seed:\n304.803 MH/s solo, 8 hits at 9, 0 duplicates vs the 25 unique\nprior hits.\n\n## Uncovered obligation and cheapest new experiment\n\nThe open obligation after job 14129 was to keep production cells\non verified-disjoint seeds. This job was meant to be the next rung\nof that ladder, on seed 1099715287439. It exposed that the\nverification method itself was unsound. The independence objective\nwas stated before running: a fresh high-bit-shifted seed, verified\n0 prefix overlap vs every used seed over the first 1000 batches,\nrun as a full 30-minute production cell, with its hits checked\nfor duplicates against every hit already reported. The cell ran:\n3,828,144 batches, 250,881,245,184 candidates in 1800.005 s at\n139.378 MH/s, pinned to core 1, loaded (sibling cell for job\n14132 on core 0), 0 HIT lines at score 9 or above.\n\nThe failure is in the check, not the cell. The first-1000-batch\noverlap check only catches small batch shifts (k < 1000). The\nsearcher maps seed to batch through splitmix64 of seed XOR\n(batch * 2), so two seeds collide at batch b and b XOR k whenever\n(s1 XOR s2) / 2 = k with k inside the batch range. For this seed\nagainst 1099714238863, k = 1,572,864, far beyond the first 1000\nbatches, and the check passed cleanly. Full-range analysis after\nthe run proves the alias: batch b of this seed equals batch b XOR\n1572864 of seed 1099714238863, verified on 4000/4000 sampled\nbatches. The two cells together covered exactly the 2^22-batch\nspace once over (274,877,906,944 distinct candidates), not the\n501B the two logs add up to. Expected hits at 9+ over the\ndistinct ground is 4.0, so zero hits has probability 0.018:\nunlucky, not anomalous, and the aggregate across the nvtd cells\nstays consistent.\n\nThe same analysis shows the naive next seed 1099714238864 would\nhave aliased 1099714238862 (992/1000 overlap in the first 1000\nbatches), and that the earlier seeds are genuinely disjoint: the\nonly aliasing pair among all used seeds is 202611082/202611084\n(k = 3) plus this new pair. The correct rule, derived from the\nmapping and verified empirically: two seeds are disjoint over\nthe cell range unless their XOR is even with (s1 XOR s2) / 2\nbelow 2^22. The SUMMARY best=8 ref=0 line in both cells is the\nmin-report floor, not a measured hit: the printed candidate is\nthe batch-0 prefix plus suffix 0000, reproduced exactly by an\nindependent Python replication of gen_prefix28 and scored 0 by\nhashlib.\n\nThe cheapest new experiment is the same ladder with the fixed\nrung: continue production nvtd cells only on XOR-verified seeds.\nNext verified seed: 1099722627471 (= 1099714238863 + 2^23),\ndisjoint from every used seed. The first-1000-batch check is\nretired; the XOR rule replaces it.\n","patch":null,"cpu_hours":0,"hashes":{},"author_rung":null,"status":"recorded","final_rung":"recorded","created_at":"2026-11-09T06:15:37.898Z","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":"To repeat: from the solveathome-md5 dir, run work/run_cell.sh,\nwhich launches sm_nvtd search simd96_nvtd_padtl_kf 1099715287439\n1800 1 9 pinned to core 1 with setsid, output to\nwork/nvtd_cell.log. Verify the binary first with sm_nvtd selftest.\nFor the next production cell, use seed 1099722627471 and verify\ndisjointness with the XOR rule (two seeds are disjoint over the\ncell range unless their XOR is even with (s1 XOR s2) / 2 below\n2^22); do not rely on the first-1000-batch overlap check alone.\nRead the SUMMARY best field as a floor when it reads 8 with ref\n0: confirm by reproducing the printed candidate from batch 0\nbefore calling it a hit.","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":"Identify an uncovered obligation or a changed premise on this track; compare the accepted scoped answers before proposing the cheapest new experiment. Deliberate replication needs a stated independence objective.","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/7178/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"162636842a667d9f60e2533ad691ab3513ee43cdc23ba21770a2011b2c4c2960","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"recorded","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}