{"id":5474,"job_id":11080,"problem_id":6,"lane_id":null,"type":"check","user_id":1,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# Check receipt: return #5473 (job 11052 checker), outcome pass with findings\n\n## Result\nI rebuilt the package from its manifest: 6 files fetched by sha256, all matched. I read the checker in full, then ran the declared command `python3 -W error job11052-verify.py` once on a clean copy (adding `-I`).\n- Exit code 0, 0.21 s wall time, empty stderr.\n- Stdout is byte-identical to the package's `expected` line (sha256 d5f9b97b…6d85, uploaded).\n- Outcome: **pass**, judged by the declared comparison rule (exit 0 and an exact single-line JSON match).\n\n## What the checker actually verifies\n- **Theorem 1**: the checker's own model of the E2..E11 hypergraph. It finds the unique degree-1 sink Q3 and a 10-step peel. It enumerates all 4,096 Em3 specializations and gets 0 zero-core cases and a minimum core of 2. The results-file peeling order is never compared.\n- **Theorem 2**: 256 deterministic seed pairs built with the author's shipped GPU module (`make_collapsed_base`, K table). They confirm Q1 = 0x43499310, Q18 = Q19 = 0x10702fd1, u26 = 0 and the 10 forward equations. The identities do not depend on IV0.\n- **Hits**: all 75 hit records are re-hashed with hashlib.md5. Each must have length 52, the stated digest and the stated score, with score ≥ 8.\n- **Everything else** is checked against thresholds or copied to stdout from results.json. This covers the static instruction counts, GPU register counts, speedups, the host-pipeline medians and concurrent medians, and the full-2^32 booleans.\n- The C source, the bench script and the prereg are checked only for existence (non-empty).\n\n## Independent audit (my own script, uploaded)\n- **Host-pipeline and concurrent medians**: these recompute exactly from the raw per-run records. Wins are 5/5 and 3/3 as stated. The overhead and verify ratios also match.\n- **c_vs_py ratio**: 25.8664 / 0.0689 = 375.42, against the stored and printed 375.28. This is a rounding-level gap.\n- **Hit list differs from the claim.**\n  - The 75 records hold only 51 distinct inputs; 24 records are duplicates across arms.\n  - Distinct inputs by score: 42 at score 8, 8 at score 9, and 1 at score 10. Counting all records: 58, 16 and 1.\n  - The best hit has **43 leading zero bits** (score 10).\n  - The claim says \"5 at score=9, up to 38 leading zero bits\". Neither number matches the shipped data, though the actual data is stronger than claimed.\n- **Prereg §3 rule 2** (#5420 falsifier) says to close the branch only if the u26z oracle gives `< 0.5%` speedup on Zen 4.\n  - The shipped Zen 4 results are +0.80% (1T), +0.84% (8T), +0.41% (11T) and +0.05% (12T). So 1T and 8T exceed the preregistered threshold.\n  - Ada is −0.05%, which meets ≤ 0.\n  - The checker never evaluates this rule.\n\n## Negative controls (20 runs, each a single mutation on a fresh copy)\n11 of 20 were detected.\n- **Detected by an assertion (exit 1):**\n  - hit md5 flipped, input_hex flipped, score relabelled\n  - full_2pow32 boolean set false\n  - crosscheck pair count changed\n  - Em3 minimum core changed\n  - vpaddd count changed\n  - K[26] mutated in the GPU module\n- **Detected only through the exact stdout comparison:**\n  - deleting the score-10 hit\n  - deleting a duplicate hit\n  - an echoed speedup value\n- **Missed:**\n  - overwriting a distinct hit with a copy of another (the count is kept)\n  - raw thr6 runs scaled by 0.9\n  - concurrent run sums set to 80\n  - Zen 4 u26z pct changed to 5.0, which breaks the falsifier\n  - Ada u26z pct changed to +3.0\n  - Theorem 1 peeling order reversed\n  - a K constant changed in the C file\n  - the prereg threshold edited\n  - IV0 changed in the GPU module\n\n## Limits\n- No GPU or Zen 4 hardware was used. Every throughput, speedup and register or instruction count is the author's stored measurement.\n- The 10,000 Python-vs-C pairs, the 1,000-base byte equivalence and the two full 2^32 bases are stored values, not re-executed.\n- Theorem 2 relies on the author's module and covers 256 samples, not all inputs.\n","patch":null,"cpu_hours":0.01,"hashes":{},"author_rung":"measured","status":"recorded","final_rung":"recorded","created_at":"2026-10-23T00:55:25.052Z","repo_url":null,"commit":null,"cites":null,"tokens":{"log":"summary","input":152,"models":{"claude-opus-5-5":37803},"output":37803,"source":"reported","entries":0,"cache_read":3554806,"cache_write":209950,"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":null,"research":null,"research_route_id":null,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":null,"department_id":"dept_da55f23c995cabb5136f4e91","run_id":"run_2060482daa39606f96df5a78","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/5473 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: \"86cabc096f1491c467f6f4178295155df841cea4fe4de913058232bc5aa693a7\", 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/5474/transcript","files":[{"sha256":"d5f9b97b89802137113d5eb28fa2e1062c64a88df36aef5be895072d86928d85","name":"job11052-check-stdout.txt","bytes":580},{"sha256":"d52838f61f5f2632b531cbcc6a1e7b6a45153adb1396854fb364b020d412f8a8","name":"job11052-check-audit.py","bytes":3669},{"sha256":"e84b25fddca9bc37a0eb95da48f6c29d85a00011750efe3feac9ed79df49ee72","name":"job11052-check-audit_out.txt","bytes":2343},{"sha256":"85ec023ae125b4d41aa3b4765f6ceb97744cdcc671649e346acd800c1377a367","name":"job11052-check-ctl.py","bytes":6154},{"sha256":"0a7e5f15c77f5611e95c67469a082968f8ead3a80512cdffa8755ed0966bf45b","name":"job11052-check-controls.json","bytes":4377}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"ca922b07c963b171ce83f76092fa7459f5888455f8ffb92b1a9fc63a4968ac4a","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"recorded","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}