{"id":3003,"job_id":6311,"problem_id":6,"lane_id":null,"type":"measure","user_id":73,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# All zeros: new platform record, 12 leading zero hex digits (submission #173, verified). It took 1.75 h of the v4 kernel from #2999 at 64.1 G/s, and the hit counts match 16^-k through k = 12.\n\n**Result.** A 52-byte input hashes to `0000000000005bd5b78bc62c81ce8d76`:\n\n`8e2bb35054c90189e8f2fa1039c94eb9313c681a1df153fdabcc9991f8fb5a058a74c61a28884bf9d165da607f5c9f14f76664d0`\n\n- The platform best moves from 11 to **12** (track page: submission 173). The published best remains 14 (Beneri #209).\n- Verification:\n  - the server's verifier marked it verified at 2026-10-11T14:39:00Z;\n  - the kernel's host check re-ran our RFC 1321 code against the device digest;\n  - `submit.py` re-hashed it with Python hashlib before sending;\n  - a fresh hashlib re-hash agrees.\n\n**Method, with no new claims about odds.**\n- **Kernel.** The v4 kernel from #2999: the q9_52 Q9-tunnel search of #2622 and #2887 (52-byte single block, Q1..Q24 fixed per base, x = Q9 derives m8, m9, m12, steps 24..60 per candidate), with K + m pre-added, FMA-pipe offload and two candidates per thread. It was validated with 0 mismatches in #2999.\n- **Run.** `md5q9 search 6311 q9_52 6300 10`, in the GPU sandbox (exit 0, 6,310 s wall, no survivors). It covered 93,962 tunnel bases and **4.036e14 candidates** at **64.1 G/s**, with 0 verification mismatches over 352 recorded hits.\n- **Submission and liveness.** A watcher submitted any hit with a score of at least 12 within 60 s; the 12 arrived about one hour in. It also sent an authenticated request every 30 min, so the assignment was never silent.\n\n**Counts against 16^-k** (cumulative, 4.036e14 candidates):\n\n| k | observed >= k | expected | ratio |\n|---|---|---|---|\n| 10 | 352 | 367.1 | 0.96 |\n| 11 | 26 | 22.9 | 1.13 |\n| 12 | 1 | 1.43 | 0.70 |\n\nAll are within Poisson noise. Pooled with #2906 and #2940 (2.3e14 candidates), this run takes the tunnel's measured odds to k = 12 over 6.3e14 candidates. That is consistent with #2950 and #2996: the record is a throughput race, and the tunnel changes cost, not odds.\n\n**Context.**\n- **Expectation.** At 64.1 G/s, a 12 is expected every 1.22 GPU-hours, so finding one within 1.75 h had probability about 0.76.\n- **The next digit.** A 13 needs 16x more work, about 19.5 GPU-hours on this card. A 14 (matching the published record) needs about 13 GPU-days.\n- **Where the speed came from.** The research behind this kernel (#2997, #2999) cut the expected time to a 12 from 1.77 h to 1.22 h. The structural returns (#2993, #2996) showed that no computation-sharing family beats the Q9 tunnel, so this run was the remaining option rather than a substitute for a better method.\n\n**Files.**\n- hits_ge11.tsv: every hit >= 11, with base seed, index, input, digest, score and the host-check flag.\n- search_progress.jsonl: the kernel's progress lines.\n- execution.json: the sandbox receipt.\n- submission_173.json: the server receipt.\n- watch.py, run_record.sh, submit.py: the run harness.\n","patch":null,"cpu_hours":1.75,"hashes":{"watch.py":"6b36138179dec12a3ca2a4fc4cebf79bdca59517c80fb8bc996a8782f62b3bd4","submit.py":"d1ca9e0189e938b5bd1a10729db0dbb6c1b706efb9dd0f21703c699e94e05a62","hits_ge11.tsv":"d4876dab462e6959c6da8d1ea300194e500ee53844c2a8ac2bc89c173c7f4b7e","run_record.sh":"68580eca2f9fa68b71c136f8c849973834dbc94aa4b9a839bc5e48dee6d7223f","execution.json":"ea153d86319e9500bb0f17f5e16201e6da0969ed6b9c721386138fbea553874b","submission_173.json":"e6eed2209921d69a73ba1b7edbd03886ad9d023163995a3e2edbd1db2a2da6e2","search_progress.jsonl":"a9667f148fb23257be8b41d3e745480497e2eeadffb16320b7ea6590563317ad"},"author_rung":"verified","status":"pending","final_rung":null,"created_at":"2026-10-11T15:24:40.866Z","repo_url":null,"commit":null,"cites":{"files":[],"handles":[],"returns":[2999,2997,2887,2906,2940,2622,2950,2996],"messages":[]},"tokens":{"log":"summary","input":22,"models":{"claude-opus-5-5":9709},"output":9709,"source":"reported","entries":0,"cache_read":3967980,"cache_write":430449,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"Check the record: `python3 -c \"import hashlib;print(hashlib.md5(bytes.fromhex('8e2bb35054c90189e8f2fa1039c94eb9313c681a1df153fdabcc9991f8fb5a058a74c61a28884bf9d165da607f5c9f14f76664d0')).hexdigest())\"` prints 0000000000005bd5b78bc62c81ce8d76 (12 leading zeros, 52 bytes).\nRerun: build v4 from #2999 (`nvcc -O3 -arch=sm_75 -std=c++17 -DVARIANT=4 -o md5q9_v4 md5q9opt.cu`) and run `./md5q9_v4 search 6311 q9_52 6300 10 hits.tsv`. Base seeds are deterministic, so the 12 is base seed 6311088475, index 2631453780; rows in hits_ge11.tsv can be rebuilt from those two values.","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":"2026-10-11T15:24:40.866Z","department_id":"dept_ef09d64fbbd7ddb34ab67f81","run_id":"run_c2ccb63b450f473296a41c24","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,"handle":"danieljmt","job_brief":"Using the validated v4 Q9-tunnel kernel (#2999: 62.4 G/s, 0 mismatches), does a 1.75 h search find a 52-byte input whose MD5 has >= 12 leading zero hex digits (a new platform record)? Do the hit counts stay at 16^-k through k = 12 at about 3.9e14 more candidates?\n\nWhy this step: The structural levers are closed (#2993, #2996), and kernel engineering is within about 20% of the card's issue and ALU limits at its power cap (#2997, #2999). With no research lead left, the person's direction allows record search. At 62.4 G/s a 12 is expected every 1.25 h, so P(>= 1 in 1.75 h) is about 75%.\n\nStop when: 1.75 h of search wall time has elapsed, or the run crashes. Every hit >= 10 is re-hashed on the host; any >= 12 is submitted immediately and verified.","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/3003/transcript","files":[{"sha256":"68580eca2f9fa68b71c136f8c849973834dbc94aa4b9a839bc5e48dee6d7223f","name":"run_record.sh","bytes":423},{"sha256":"6b36138179dec12a3ca2a4fc4cebf79bdca59517c80fb8bc996a8782f62b3bd4","name":"watch.py","bytes":2248},{"sha256":"a9667f148fb23257be8b41d3e745480497e2eeadffb16320b7ea6590563317ad","name":"search_progress.jsonl","bytes":125851},{"sha256":"d1ca9e0189e938b5bd1a10729db0dbb6c1b706efb9dd0f21703c699e94e05a62","name":"submit.py","bytes":3069},{"sha256":"d4876dab462e6959c6da8d1ea300194e500ee53844c2a8ac2bc89c173c7f4b7e","name":"hits_ge11.tsv","bytes":4480},{"sha256":"e6eed2209921d69a73ba1b7edbd03886ad9d023163995a3e2edbd1db2a2da6e2","name":"submission_173.json","bytes":1385},{"sha256":"ea153d86319e9500bb0f17f5e16201e6da0969ed6b9c721386138fbea553874b","name":"execution.json","bytes":429}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"490433da1d5d6316fbb97496aa36d898255e223af8c5a9ff16f1b679aaa907f6","research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}