{"id":4076,"job_id":8327,"problem_id":6,"lane_id":35,"type":"check","user_id":1,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# Check receipt: #4072 verification package (route 286, dBB GPU pipeline + d3/d4)\n\n**Outcome: pass at the checker's actual coverage.** The declared command `python3 job8276-verify.py` was run unchanged on a hash-verified copy of the 5-file manifest (Python 3.14.8, `-I`, no network). It took 0.02 s, exited 0, and its stdout (1040 B, sha256 d6bf418b…) is byte-identical to `verification_plan.expected`.\n\n## What the checker really recomputes\n- The SHA-256 and byte size of the prereg, the OpenCL source and the host .py, compared with the hashes recorded in results.json.\n- For each of the 20 deepest d3 records, a pure-Python RFC 1321 compression of block 1 for M and M' (M' = M with word 5 + 2^10). It checks ihv1/ihv2 and the dIHV = 0x80000000 pattern on all four words. It then runs block 2 (134 B padding block, m0 from b_cand) and checks the MSB conditions for rounds 1–2 and the step-48..63 stop_step. The maximum is 60.\n- For both d4 hits, `hashlib.md5` equality of the 68-byte pair plus the block-1 dIHV pattern.\n\n## Read from the JSON, not recomputed\n- The H1 stream identity (768 bases, 193 solutions, 0 mismatches) is read from flags and counts the producer wrote; the GPU/CPU comparison itself is not replayed.\n- The H2 benchmark gets only threshold assertions on the three summary means and on the B2 speedup. The B3 speedup is printed but never asserted, and the raw per-run rates are never read.\n- The campaign total of 258,565 IHVs and the d4 scan count of 5,185 get only threshold checks.\n- The \"verified\" status of submissions 6474/6475 is read from the local file, not from the server.\n\nSo the package's `supports` sentence overstates what it covers: it says the checker \"independently verifies\" the stream identity and the throughput ratios. It does neither. Those claims are transcribed. The exact-stdout comparison does pin the transcribed values to the expected text.\n\n## Negative controls (each in its own fresh copy)\n| mutation | exit | stdout = expected | detected |\n|---|---|---|---|\n| d3 record stop_step 60→61 | 1 (AssertionError) | no | yes |\n| B2 mean rate 2039.06→2039.07 | 0 | no | yes (exact stdout only) |\n| drop the last deep d3 record | 0 | no (19 records) | yes (exact stdout only) |\n| drop the last d4 hit | 1 | no | yes |\n| flip one hex digit of a d3 m1_hex | 1 (ihv1 mismatch) | no | yes |\n| flip one hex digit of a d4 hex_b | 1 | no | yes |\n| flip one bit of prereg.md | 1 (SHA-256 mismatch) | no | yes |\n| scale one raw benchmark run field ×0.1 | 0 | **yes** | **no** |\n\n## Off-receipt consistency arithmetic (my own, not part of the package)\n- Per-seed solutions 63+61+69 = 193.\n- Each summary mean equals the mean of its three raw runs: B1 2197.18, B2 2039.06, B3 2621.48.\n- 2039.06/593.0 = 3.44 and 2621.48/593.0 = 4.42.\n\n## Limits\nThis is a rerun of the author's own checker on an aarch64 musl host with no GPU or OpenCL. It establishes the MD5 facts (20 d3 trails, 2 collision pairs) and that the package agrees with itself. It does not establish the GPU/CPU bit-identity, the RTX 4070 SUPER throughput or the campaign totals. Those remain producer-reported.\n\n**Conflict of interest:** an earlier session on this account wrote the lane upkeep that promoted #4072, based on its own receipt recount and d3 replay. This check reuses none of that work.\n","patch":null,"cpu_hours":0.001,"hashes":{},"author_rung":"measured","status":"recorded","final_rung":"recorded","created_at":"2026-10-15T21:56:18.844Z","repo_url":null,"commit":null,"cites":null,"tokens":{"log":"summary","input":90,"models":{"claude-opus-5-5":21222},"output":21222,"source":"reported","entries":0,"cache_read":2067568,"cache_write":141507,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":null,"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":[{"sha":"81e8321dfeec5001d9f153f15e1b1f71cc9faad57a68067f47d1cf8cd60f517f","name":"check8327-ctl.py","notes":["prints what looks like progress or timing to stdout on line 50 (\"print(json.dumps(dict(name=name, exit=p.returncode, stdout_match=(p.stdout == EX\"): stdout is the artifact and must reproduce byte for byte elsewhere; send progress, timing and rates to stderr. This one is a guess from the text, not a measurement: if the output is already identical from run to run, say so in your return and leave the file alone."]}],"research":null,"research_route_id":null,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":null,"department_id":"dept_da55f23c995cabb5136f4e91","run_id":"run_f440085037062845592a0803","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":"Benjaminsen","job_brief":"Reconstruct the immutable package from GET <project base>/return/4072 in a clean directory using ONLY its manifest and declared runtime/source requirements. Fetch each file by SHA from /files/<sha> to its relative manifest path. Inspect the checker before executing it within your person's limits. The checker must consume the submitted target, not only regenerate an unrelated expected answer. Check actual coverage and the comparison rule. Run negative controls in separate temporary copies: corrupt a value in the target, remove a record, alter the certificate, and record for each whether the checker detected it. A control the checker misses is a finding, not a failure of yours. Preserve the original files and results. Do not redo discovery. Return report_md, transcript, and check_receipt: {fingerprint: \"5aa053374f76e86d04d4ecc2adace25a8e6d60996f3ae2b25522c744918ec46b\", outcome: \"pass|fail|unable\", observed: \"actual output and differences\", elapsed_seconds: <actual time>, stdout_sha256: \"<uploaded actual output>\", exit_code: <integer or null if unable>, environment: \"observed versions\", coverage_md: \"exactly what ran, exclusions and seeds\", execution_policy: \"authenticated-contributor-v1\", attestation_md: \"for Lean: personally observed execution, actual artifacts, negative controls and limitations; unable attests only the blocker\", method: \"rerun|independent_implementation\", shared_components_md: \"shared algorithm, code, parser or library\", controls_md: \"negative controls and their observed outcomes\", controls: [{name: \"what you corrupted\", detected: true|false, note: \"exit code and message\"}], limits_md: \"what this execution does not establish (an unpinned producer, an unread input, a scope the checker skips)\"}. The itemised controls and limits_md feed the generated summary reviewers read first; write them for a reader who will not open the transcript. If execution cannot proceed, use outcome unable and blocker: {kind: \"capability|package\", required_tools: [], required_sources: []}. Use capability only when another worker with the named tools or source access can run the unchanged package; include at least one missing capability identifier. Use package for missing artifacts, undeclared dependencies or defects requiring repair, and describe the defect in observed. A capability gap permits one targeted reassignment; package defects and unresolved second attempts go to judgment. A repair requires a new package. Execution receipts remain worker-reported evidence at their stated coverage, not mathematical verdicts.","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/4076/transcript","files":[{"sha256":"d6bf418bcaae2d27a0a94d9b1eab43a508920032c9432c6b4bbeba4ae6c3f2f9","name":"job8276-verify-stdout.txt","bytes":1040},{"sha256":"408aed8ec35269dc82099a006a0709b0c3e89164d9badff2949451025b5d0f21","name":"check8327-controls.jsonl","bytes":1212},{"sha256":"81e8321dfeec5001d9f153f15e1b1f71cc9faad57a68067f47d1cf8cd60f517f","name":"check8327-ctl.py","bytes":2435}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"244ec9017348b44749639226b9b17d4229904d65653b5b2181195af3219c418c","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"recorded","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}