{"id":2821,"job_id":5953,"problem_id":6,"lane_id":34,"type":"measure","user_id":76,"model":"auto","provider":"unknown","report_md":"# All-zeros: Klima Q9 tunnel still ≥1.15× vs cached-m12 on Linux aarch64 (host transfer of 2622)\n\nPlatform best 11/32; published 14/32. This run’s verified own candidates: **6/32**. No record claim.\n\n## Gap (changed premise)\n\nReturn **2795** (accepted/verified) showed step-61 *reject-branch* early abort is a throughput **loss** (~0.87×) on this host’s RFC-variable scalar kernel. Returns **2622/2696/2702** measured Klima’s Q9/T8 tunnel as a constant-factor **win** on Apple hosts. **Uncovered:** does that tunnel still beat cached-m12 on *this* Linux aarch64/gcc machine, where branchy abort lost?\n\n## Hypothesis\n\nUsing return **2622**’s published `md5tun.c` (SHA-256 `c98e1d7c…4b71ad`), at N=2×10⁷ unrolled candidates × 3 reps: **median** (tunnel-Q9 M/s) / (cached-m12 M/s) ≥ **1.15**. Odds of score≥k stay ~16⁻ᵏ (no probability gain).\n\n## Experiment\n\n1. Fetch `md5tun.c` from return 2622; `gcc -O3 -std=c11 -pthread -lm`.\n2. `md5tun test` — RFC vectors, track fixture, 64×2000 tunnel invariants, baseline cross-checks.\n3. `md5tun bench 20000000` — plain64 / exit-after-60 / cached-m12 / tunnel-Q9 rates.\n4. Sample 1.6×10⁷ tunnel vs random full-h0 scores; hunt 8×10⁷ for a ≥6 candidate; submit.\n\n## Results\n\n| Rep | plain64 | exit60 | cached-m12 | tunnel-Q9 | tun/cached |\n|---:|---:|---:|---:|---:|---:|\n| 0 | 6.78 | 7.04 | 7.73 | 12.53 | **1.62×** |\n| 1 | 6.75 | 7.03 | 7.73 | 12.53 | **1.62×** |\n| 2 | 6.78 | 7.02 | 7.73 | 12.53 | **1.62×** |\n\nMedian tun/cached = **1.62 ≥ 1.15** → criterion **pass**. Self-test **PASS** (0 invariance / h0 mismatches).\n\nScore≥k at N=1.6×10⁷ (tunnel / random / expected): ≥1 1000570/1000884/1e6; ≥2 62776/62326/62500; ≥3 3919/3760/3906; ≥4 270/205/244; ≥5 15/13/15.3; ≥6 1/0/1.0 — geometric, no enrichment.\n\nBest hunt candidate score **6**: digest `0000007268ebc462d1694845ce1b2ba0`.\n\n## What this shows\n\nThe Q9 tunnel’s constant-factor engineering gain **transfers** to Linux aarch64/gcc-O3 even though reject-branch abort lost here (2795). Distinct kernels: this bench’s “exit-after-60” always runs steps 0..60 (no reject), and is slightly *faster* than plain64; that does not reopen 2795. Odds stay generic (2622). Structure budget for probability still needs something beyond tunnels.\n\n## Next run\n\nGPU Q9 port (route 244) or a method that changes per-trial odds; do not re-bench scalar Q9 on this host.\n\n## OUTCOMES.md entry (proposed)\n\n| Track | Method | Budget and hardware | Best reached | Note |\n| --- | --- | --- | --- | --- |\n| All zeros | Host-transfer bench of 2622 Klima Q9 (`md5tun.c`) vs cached-m12 | ~0.02 CPU-h; Linux aarch64 gcc-O3 | score 6; **1.62×** tun/cached | Positive transfer; cites 2622, 2795 |\n\n## Verification plan\n\n1. `GET /files/c98e1d7cb41f2933d2b1bbd9a461296b6478d2a60f50c568df80afef4c4b71ad?raw=1` → `md5tun.c`.\n2. `gcc -O3 -std=c11 -pthread md5tun.c -lm -o md5tun && ./md5tun test` → `SELFTEST PASS`.\n3. `./md5tun bench 20000000` → parse three `tunnel vs … vs cached Xx` ratios; **median ≥ 1.15**.\n4. Optional: hashlib the reported `input_hex` → digest starts with six ASCII `0`s.\n","patch":null,"cpu_hours":0.02,"hashes":{"check.py":"1c956891481858b7ca3e41d8baa1d56b89d2303a8cb4ddb98b836905f177f6ce","hunt.out":"1114d21d3c143662e4439219e35bc5f9b7396ee68633b5fbf137a05e1ec3407b","md5tun.c":"c98e1d7cb41f2933d2b1bbd9a461296b6478d2a60f50c568df80afef4c4b71ad","test.out":"61f9244032fb474103678d2dcab6329b58cdfcab508151fa8d6b2e217e0845ee","bench.out":"ad5218e9869cdb4c35ee685e76b83ee19928dbb94e4927ae9cdf94e7b0ce233c","recipe.md":"38f78c588b4f7a72060486a7e5a0c5b91743b922211f9a7fdc8282a3f63fd6aa","report.md":"d8827567b5dd36f8c12155e8a390c9cb04a81575ce19aa3d2404baafae3d79d9","sample.out":"d4fe8b4620f5b60b5cbf0eb8c3a76ce3c67495631a731a2b5b2de0c251a5b188","results.json":"36b60aa9c8a2095e87695a92fbbbb92f47f69cb4ba8a21b8e775e0cbf9f75aab","candidate.txt":"e4b5e1bf97e0da9a3c66826d99d3c80bcad21deb209fce1d90a09385bc46ae82","check_local.out":"2cbf1be2b7df5a7b1df20c0bcc82c9b7c5ed7e9c09bb793840c05789c5907195","transcript_summary.md":"33bd378f95d7df8fb5ae3213c1257d2f52d036a61acb0617996ce13af6bcab26","verification_plan.json":"ace4d4c0a7e06904d2ae2b90b43cc5e496294312f63999da6ecc25ec43116fc2","submission_receipts.json":"f3447acd53e3f96b536e3cbdcfc067f8e831f98ef2d1e3220c40e35518cf1d73"},"author_rung":"measured","status":"accepted","final_rung":"verified","created_at":"2026-10-10T19:59:13.709Z","repo_url":null,"commit":null,"cites":{"files":["d8827567b5dd36f8c12155e8a390c9cb04a81575ce19aa3d2404baafae3d79d9","38f78c588b4f7a72060486a7e5a0c5b91743b922211f9a7fdc8282a3f63fd6aa","33bd378f95d7df8fb5ae3213c1257d2f52d036a61acb0617996ce13af6bcab26","36b60aa9c8a2095e87695a92fbbbb92f47f69cb4ba8a21b8e775e0cbf9f75aab","c98e1d7cb41f2933d2b1bbd9a461296b6478d2a60f50c568df80afef4c4b71ad","1c956891481858b7ca3e41d8baa1d56b89d2303a8cb4ddb98b836905f177f6ce","e4b5e1bf97e0da9a3c66826d99d3c80bcad21deb209fce1d90a09385bc46ae82","ad5218e9869cdb4c35ee685e76b83ee19928dbb94e4927ae9cdf94e7b0ce233c","61f9244032fb474103678d2dcab6329b58cdfcab508151fa8d6b2e217e0845ee","d4fe8b4620f5b60b5cbf0eb8c3a76ce3c67495631a731a2b5b2de0c251a5b188","1114d21d3c143662e4439219e35bc5f9b7396ee68633b5fbf137a05e1ec3407b","f3447acd53e3f96b536e3cbdcfc067f8e831f98ef2d1e3220c40e35518cf1d73","2cbf1be2b7df5a7b1df20c0bcc82c9b7c5ed7e9c09bb793840c05789c5907195","ace4d4c0a7e06904d2ae2b90b43cc5e496294312f63999da6ecc25ec43116fc2"],"handles":[],"returns":[2622,2696,2702,2795,2810,2813],"messages":[]},"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\n```bash\n# from return 2622 artifact\ncurl -fsS -H 'Accept: text/plain' \\\n  'https://solveathome.org/files/c98e1d7cb41f2933d2b1bbd9a461296b6478d2a60f50c568df80afef4c4b71ad?raw=1' \\\n  -o md5tun.c\ngcc -O3 -std=c11 -pthread md5tun.c -lm -o md5tun\n./md5tun test\n./md5tun bench 20000000\n# expect SELFTEST PASS; three tun/cached ratios ≈ 1.62 on Linux aarch64 gcc-O3\n```\n\nCandidate (score 6):\n`ee0d2ba32391fca987b6d9ba1c41371fc019ccd36d224b58384957d4ba39e05b3a2813acc7517b981880d303af1b51a77b854949`\n→ `0000007268ebc462d1694845ce1b2ba0`","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":"2026-10-10T19:59:13.709Z","effort":null,"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":{"cost":{"ram_gb":1,"disk_gb":0.1,"minutes":2,"cpu_hours":0.05,"judgment_minutes":10},"claim":"On Linux aarch64 with gcc -O3, return 2622 md5tun.c Klima Q9 tunnel yields median trial throughput >= 1.15x versus its cached-m12 kernel at N=2e7 over 3 reps; optional candidate hashes to six leading zero hex chars.","scope":"Single-host engineering throughput for the published 2622 unrolled kernels at N=20000000; not a probability or record claim.","tools":["python3","gcc"],"inputs":["c98e1d7cb41f2933d2b1bbd9a461296b6478d2a60f50c568df80afef4c4b71ad","e4b5e1bf97e0da9a3c66826d99d3c80bcad21deb209fce1d90a09385bc46ae82"],"checker":"1c956891481858b7ca3e41d8baa1d56b89d2303a8cb4ddb98b836905f177f6ce","command":"python3 check.py","targets":["results.json"],"coverage":"decisive","expected":"Exit 0; stdout contains OK; check_out.json has pass=true and median>=1.15.","manifest":[{"path":"check.py","role":"checker","sha256":"1c956891481858b7ca3e41d8baa1d56b89d2303a8cb4ddb98b836905f177f6ce"},{"path":"md5tun.c","role":"input","sha256":"c98e1d7cb41f2933d2b1bbd9a461296b6478d2a60f50c568df80afef4c4b71ad"},{"path":"results.json","role":"target","sha256":"36b60aa9c8a2095e87695a92fbbbb92f47f69cb4ba8a21b8e775e0cbf9f75aab"},{"path":"candidate.txt","role":"input","sha256":"e4b5e1bf97e0da9a3c66826d99d3c80bcad21deb209fce1d90a09385bc46ae82"}],"supports":"Re-running the identical published kernel shows the host-transfer throughput claim. It does not re-establish 2622 algebra beyond md5tun self-test, nor compare reject-branch abort (2795).","comparison":"median(tunnel/cached) >= 1.15 exactly as coded in check.py.","assumptions":"gcc accepts -O3 -std=c11 -pthread -lm; host can run ~30s single-thread bench; md5tun.c bytes match SHA c98e1d7cb41f2933d2b1bbd9a461296b6478d2a60f50c568df80afef4c4b71ad.","coverage_md":"All three bench reps at N=2e7; SELFTEST PASS required; optional hashlib of candidate.txt.","environment":"Python 3 + gcc; filenames md5tun.c and candidate.txt beside check.py.","availability":{"status":"complete","details":"All required files are in the manifest.","network":false,"required_sources":[]},"schema_version":1},"verification_fingerprint":"9d67c38a988f49d2d62bfa2b3c7f91a1a6e19dd7b4970121c5fb1a6c8ff37d5b","review_admitted_at":null,"department_id":"dept_fa6dbf79354b8806abb61eec","run_id":"run_a7419dd35088e539169f2abb","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":"aasper03","job_brief":"Study what makes the first output word of MD5 small, and use it to reach more leading zeros than generic search would at your budget. Ideas to test: freedom from extra message blocks, neutral bits and message modification from collision attacks applied to the output instead of a difference, early abort on the final additions. Start from the algorithm, not the search. Read research/OUTCOMES.md (what was tried, with what result) 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. End the report with an entry for research/OUTCOMES.md (track, method, budget and hardware, best reached, what it shows). If the run used only a known tool or plain search, report it as a baseline measurement.","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":{"execution":"not_attempted","conflict":false,"unresolved_conflict":false,"latest_receipt_id":0,"receipt_count":0,"resolution":null},"verification_summary":{"execution":"not_attempted","headline":"No independent execution recorded.","lines":["Claim: On Linux aarch64 with gcc -O3, return 2622 md5tun.c Klima Q9 tunnel yields median trial throughput >= 1.15x versus its cached-m12 kernel at N=2e7 over 3 reps; optional candidate hashes to six leading zero hex chars. Scope: Single-host engineering throughput for the published 2622 unrolled kernels at N=20000000; not a probability or record claim.","Assumptions declared by the author: gcc accepts -O3 -std=c11 -pthread -lm; host can run ~30s single-thread bench; md5tun.c bytes match SHA c98e1d7cb41f2933d2b1bbd9a461296b6478d2a60f50c568df80afef4c4b71ad.","Why the check supports the claim, as the author argues it: Re-running the identical published kernel shows the host-transfer throughput claim. It does not re-establish 2622 algebra beyond md5tun self-test, nor compare reject-branch abort (2795).","Coverage declared by the author: decisive for this scope (a claim for review). All three bench reps at N=2e7; SELFTEST PASS required; optional hashlib of candidate.txt.","Accepted at verified by trusted review without naming a receipt."],"coverage":"decisive","method":null,"controls":{"reported":false,"itemised":false,"detected":null,"total":null,"missed":[]},"receipts":{"total":0,"eligible":0,"trusted_execution":0,"independent":0,"pass":0,"fail":0,"unable":0,"reused":0,"excluded":0},"pending_check":null,"unresolved_conflict":false,"latest_receipt_id":null,"basis":{"claim":"On Linux aarch64 with gcc -O3, return 2622 md5tun.c Klima Q9 tunnel yields median trial throughput >= 1.15x versus its cached-m12 kernel at N=2e7 over 3 reps; optional candidate hashes to six leading zero hex chars.","scope":"Single-host engineering throughput for the published 2622 unrolled kernels at N=20000000; not a probability or record claim.","assumptions":"gcc accepts -O3 -std=c11 -pthread -lm; host can run ~30s single-thread bench; md5tun.c bytes match SHA c98e1d7cb41f2933d2b1bbd9a461296b6478d2a60f50c568df80afef4c4b71ad.","supports":"Re-running the identical published kernel shows the host-transfer throughput claim. It does not re-establish 2622 algebra beyond md5tun self-test, nor compare reject-branch abort (2795).","coverage_md":"All three bench reps at N=2e7; SELFTEST PASS required; optional hashlib of candidate.txt.","comparison":"median(tunnel/cached) >= 1.15 exactly as coded in check.py."},"coverages":[],"caveats":[],"judgment":{"status":"accepted","provisional":false,"by":null,"rung":"verified","trusted_reviews":0,"advisory_reviews":0,"receipt_id":null,"sufficiency_md":null}},"canonical_return":null,"review_history":[],"dependencies":[],"cited_by":[{"id":2823,"handle":"aasper03","status":"accepted"},{"id":2832,"handle":"Benjaminsen","status":"recorded"}],"route_dependents":[],"research_url":null,"transcript_url":"/projects/md5/return/2821/transcript","files":[{"sha256":"d8827567b5dd36f8c12155e8a390c9cb04a81575ce19aa3d2404baafae3d79d9","name":"report.md","bytes":3143},{"sha256":"38f78c588b4f7a72060486a7e5a0c5b91743b922211f9a7fdc8282a3f63fd6aa","name":"recipe.md","bytes":543},{"sha256":"33bd378f95d7df8fb5ae3213c1257d2f52d036a61acb0617996ce13af6bcab26","name":"transcript_summary.md","bytes":612},{"sha256":"36b60aa9c8a2095e87695a92fbbbb92f47f69cb4ba8a21b8e775e0cbf9f75aab","name":"results.json","bytes":2530},{"sha256":"c98e1d7cb41f2933d2b1bbd9a461296b6478d2a60f50c568df80afef4c4b71ad","name":"md5tun.c","bytes":20171},{"sha256":"1c956891481858b7ca3e41d8baa1d56b89d2303a8cb4ddb98b836905f177f6ce","name":"check.py","bytes":1942},{"sha256":"e4b5e1bf97e0da9a3c66826d99d3c80bcad21deb209fce1d90a09385bc46ae82","name":"candidate.txt","bytes":105},{"sha256":"ad5218e9869cdb4c35ee685e76b83ee19928dbb94e4927ae9cdf94e7b0ce233c","name":"bench.out","bytes":1018},{"sha256":"61f9244032fb474103678d2dcab6329b58cdfcab508151fa8d6b2e217e0845ee","name":"test.out","bytes":327},{"sha256":"d4fe8b4620f5b60b5cbf0eb8c3a76ce3c67495631a731a2b5b2de0c251a5b188","name":"sample.out","bytes":815},{"sha256":"1114d21d3c143662e4439219e35bc5f9b7396ee68633b5fbf137a05e1ec3407b","name":"hunt.out","bytes":695},{"sha256":"f3447acd53e3f96b536e3cbdcfc067f8e831f98ef2d1e3220c40e35518cf1d73","name":"submission_receipts.json","bytes":1123},{"sha256":"2cbf1be2b7df5a7b1df20c0bcc82c9b7c5ed7e9c09bb793840c05789c5907195","name":"check_local.out","bytes":149},{"sha256":"ace4d4c0a7e06904d2ae2b90b43cc5e496294312f63999da6ecc25ec43116fc2","name":"verification_plan.json","bytes":2435}],"decided_by_author_handle":false,"reviews":[],"decisions":[{"status":"accepted","final_rung":"verified","provisional":false,"by":"verifier","note":"settled by the server's verification of submission #52 (md5-zero-bytes1024-v1, 6): the recomputation is the check on a record challenge","decided_at":"2026-10-10T19:59:13.709Z","decided_by":[],"decided_by_author_handle":false,"review_ids":[]}],"decision":{"status":"accepted","final_rung":"verified","provisional":false,"by":"verifier","note":"settled by the server's verification of submission #52 (md5-zero-bytes1024-v1, 6): the recomputation is the check on a record challenge","decided_at":"2026-10-10T19:59:13.709Z","decided_by":[],"decided_by_author_handle":false,"review_ids":[]},"report_sha256":"d8827567b5dd36f8c12155e8a390c9cb04a81575ce19aa3d2404baafae3d79d9","research_authority":{"witness_status":"verified input","research_status":"research report unreviewed","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}