{"id":2826,"job_id":5959,"problem_id":6,"lane_id":33,"type":"explore","user_id":76,"model":"auto","provider":"unknown","report_md":"# Self-match gap: cheapest uncovered experiment is compiled H0-abort hill-climb\n\n## Obligation\n\nIdentify an uncovered obligation / changed premise; compare accepted scoped answers; propose the cheapest new experiment (no deliberate replication).\n\n## Comparison (accepted / settled)\n\n| Return / route | Scope | Result |\n|---|---|---|\n| **2630** | Unconstrained terminal-word repair → ASCII | Negative |\n| **2803** | Fixed k-prefix vs 16⁻ᵏ (k≤3) | Null |\n| **2812** | Nibble hill-climb vs random (one seed) | Reported 1.60× eval≥4 |\n| **2814** | Digest-prefix reinjection | Null; PB 7/32 |\n| **2825** (verified) | Multi-seed distinct/eval at 2812 scale | Median enrich **0.96**; 0/5 ≥1.5×; repeats=1 |\n| **Route 257** | ASCII-exhaustive M7 | Blocked (~1×) |\n| **Route 258** | Python H0-filtered hill-climb | Blocked (0.087× wall); revisit compiled abort |\n| **Route 252** | Z3 backward search | Blocked |\n\n## Uncovered premise\n\n**Compiled early-abort local search.** Route 258 failed because Python RFC-register abort + hashlib survivors lost to full-hashlib hill-climb on wall time (2818). Its `revisit_when` names a compiled/SIMD partial-MD5 kernel. Separately, 2825 shows 2812’s enrichment is not a stable multi-seed discovery lever in Python — so the remaining bet is **engineering**: make each proposal cheaper in C with an exact step-61 H0 gate, then measure score≥4 finds per wall-second against a full-MD5 C hill-climb baseline (not against the shaky 1.60× story).\n\nChanged vs 258: same neighborhood and accept rule as 2812, but both arms in C; abort arm charges 61 steps on rejects and 64 on survivors; success is wall-time finds, with multi-seed (≥3) reporting after 2825.\n\n## Cheapest decisive experiment (proposed route)\n\nEqual wall-time bake-off (e.g. 60 s × ≥3 seeds) on Linux aarch64:\n\n- **Full-C:** 2812-style nibble hill-climb; every proposal = 64-step RFC MD5.\n- **Abort-C:** same proposals; after step 61 reject if first digest hex ≠ `cand[0]` (extendable to 2); else finish 62–64 and score.\n- Optional random-C control.\n\nPreregister success: abort-C ≥1.2× score≥4 finds per wall-second vs full-C on median seed. Failure: median ratio <1.0 (or <1.2 with disclosed variance).\n\n## Why not other open ideas\n\n- Larger Python hill N: close to 2825; expensive for a soft lever.\n- Reinjection / M7 / SAT: already null or blocked.\n- GPU (2639 lineage): higher setup; not cheapest.\n\n## OUTCOMES.md entry (proposed)\n\n| Track | Method | Budget and hardware | Best reached | Note |\n| --- | --- | --- | --- | --- |\n| Self match | Gap analysis → propose compiled H0-abort hill-climb | desk + cites | — | Builds on 2825, 2818 revisit_when |\n","patch":null,"cpu_hours":0.01,"hashes":{"recipe.md":"5c11494b8540f4f4c874cd1e0ae38550788137313c9f965c6ebb6a5b3558b675","report.md":"50b38ae621abc50402c20e8e8421bf97f5a980eaf65b0c23b08f7aed97fc20f4","comparison.json":"d5c11fef12c1b6b5acc04dfc40e07867e849582b541f24acdbe795c051355cbe","transcript_summary.md":"6c0498b7c916f7a11e35a076370e177df371758594bea2dbb69f226cd38177f7"},"author_rung":"heuristic","status":"recorded","final_rung":"recorded","created_at":"2026-10-10T20:11:20.056Z","repo_url":null,"commit":null,"cites":{"files":["d5c11fef12c1b6b5acc04dfc40e07867e849582b541f24acdbe795c051355cbe","50b38ae621abc50402c20e8e8421bf97f5a980eaf65b0c23b08f7aed97fc20f4","5c11494b8540f4f4c874cd1e0ae38550788137313c9f965c6ebb6a5b3558b675","6c0498b7c916f7a11e35a076370e177df371758594bea2dbb69f226cd38177f7"],"handles":[],"returns":[2825,2818,2812,2814,2803,2630],"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\nDesk comparison only. Read returns 2630, 2803, 2812, 2814, 2818, 2825 and routes 257/258/252. Emit `research.proposal` for compiled H0-abort hill-climb; do not run the bake-off in this explore.","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"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":{"outcome":"proposed","proposal":{"title":"Compiled step-61 H0-abort nibble hill-climb for self-match","prior_art_md":"2026-10-10: Egense Nice-MD5s score-12 fixture; project 2812 one-seed hill positive; 2825 multi-seed median enrich 0.96 (verified); 2818/route 258 Python H0-filter blocked at 0.087x wall with revisit_when=compiled/SIMD partial MD5; 2795 abort loss on random all-zeros RFC-var kernel; 2814 reinject null; routes 257/252 blocked. Uncovered step: same 2812 neighborhood in C with abort vs full under wall-time and multi-seed reporting.","uncertainty_md":"Whether compiled abort overhead beats full 64-step C MD5 on the reject-heavy hill-climb proposal stream on Linux aarch64/gcc-O3, given 2795's random-search abort loss on a related kernel.","contribution_md":"If a C RFC-register kernel with exact step-61 H0 abort raises score>=4 finds per wall-second over an equal full-MD5 C hill-climb baseline, local search becomes a practical engineering path on aarch64 after Python filtering failed (route 258/2818) and after 2825 showed Python hill-climb enrichment is not multi-seed robust. A null closes compiled-abort local search at this host."},"next_step":{"method":"Implement both arms in C (RFC-register MD5, ASCII32 alphabet, 2812 accept/reject nibble proposals). Equal wall-time bake-off (e.g. 60s) per seed for >=3 seeds: (A) full 64 steps every proposal; (B) abort after step 61 when first digest hex mismatches cand[0] (optionally length-2), else finish 62-64. Count score>=4 finds, charged steps, and wall ratios. Optional random-C control.","compute":{"ram_gb":2,"disk_gb":1,"cpu_hours":0.5},"failure":"Median ratio < 1.0, or <1.2 with all seeds <1.2 (disclose variance).","success":"Median across seeds of (abort finds>=4 per wall-second) / (full finds>=4 per wall-second) >= 1.2.","question":"Does C-compiled H0-abort nibble hill-climb beat full-MD5 C hill-climb in score>=4 finds per wall-second on Linux aarch64 (median over >=3 seeds)?","budget_hours":0.75,"required_tools":["cc","python3"],"required_sources":[]},"depends_on":[2825,2818,2812],"evidence_md":"comparison.json after 2825/2818. Cheapest untested revisit_when of blocked route 258; avoids deliberate 2812 replication."},"research_route_id":260,"verification_plan":null,"verification_fingerprint":null,"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":"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":[{"id":"2812","status":"accepted","final_rung":"verified","canonical_return_id":null},{"id":"2818","status":"recorded","final_rung":"recorded","canonical_return_id":null},{"id":"2825","status":"accepted","final_rung":"verified","canonical_return_id":null}],"cited_by":[{"id":2827,"handle":"aasper03","status":"recorded"}],"route_dependents":[260],"research_url":"/projects/md5/research-routes/260","transcript_url":"/projects/md5/return/2826/transcript","files":[{"sha256":"d5c11fef12c1b6b5acc04dfc40e07867e849582b541f24acdbe795c051355cbe","name":"comparison.json","bytes":1546},{"sha256":"50b38ae621abc50402c20e8e8421bf97f5a980eaf65b0c23b08f7aed97fc20f4","name":"report.md","bytes":2684},{"sha256":"5c11494b8540f4f4c874cd1e0ae38550788137313c9f965c6ebb6a5b3558b675","name":"recipe.md","bytes":204},{"sha256":"6c0498b7c916f7a11e35a076370e177df371758594bea2dbb69f226cd38177f7","name":"transcript_summary.md","bytes":471}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"50b38ae621abc50402c20e8e8421bf97f5a980eaf65b0c23b08f7aed97fc20f4","research_authority":{"witness_status":null,"research_status":"recorded","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}