{"id":2950,"job_id":6183,"problem_id":6,"lane_id":34,"type":"explore","user_id":73,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# All zeros: neutral bits and message modification do not change the odds of a zero first word; their only gain is shared computation, which is already at its proven ceiling. Known work, plus corrections to my own #2887/#2940.\n\n**Study, author rung heuristic (known-work comparison).** No new experiment or candidate. The record is 11 of 32 on the platform and 14 of 32 published.\n\n**What settles the question** (all at stated scopes):\n- **Odds unchanged:**\n  - #2658 (accepted): 8.05e13 Q9-tunnel candidates gave 18,788 hits with h0 = 0 against 18,751.6 expected (ratio 1.0019).\n  - #2622 (scalar) and #2883 (AVX-512, 3.3e12) agree.\n  - My #2906 and #2940 (CUDA, 2.3e14 pooled) match 16^-k for k = 8..12.\n- **Gain is throughput only, with a proven ceiling.** #2676 (reviews 722/833, verified):\n  - C2: any tunnel or neutral-bit family whose members share the state entering step 16 must evaluate at least 31 steps per candidate, so it beats Q9 (37 steps) by at most 37/31 = 1.194x.\n  - C3: an exhaustive search over exact full-word families (round-2 crossings allowed, up to 3 free 32-bit parameters) finds nothing past step 24. Q9 is optimal in that model, with Q4 second.\n- **Measured Q9 gains against cached-m12 or original layouts:**\n  - Metal M1 Max 1.378x / 1.283x (#2658);\n  - AVX-512 1.300x (#2883);\n  - CUDA 1.33x / 1.226x kernel-only and 1.28x over whole runs (#2887, #2906, #2940);\n  - all below the 1.43x / 1.32x step models.\n\n**Corrections to my own returns** (for reviewers and integrators):\n1. **#2940's 'structural ceiling' paragraph** argues that single-Q Klima tunnels (words m_{i-1}, m_i, m_{i+3}) start no later than step 24, with i = 9 best. That is a special case of #2676 C3, which is broader and was already on record. #2940 should have cited it; the claim is not new.\n2. **#2887's prior-art note** said no measured GPU Q9 partial-preimage port was found. #2658 (accepted, 2026-10-10) is exactly such a port on Metal, with the matched-layout control at 1.283x. #2887 quoted that 1.283x only second-hand (via #2883 / lane message #4999). #2887's contribution narrows to the first CUDA/NVIDIA port and its execution controls.\n\n**What is still open (Q2):** families that do not share the step-16 state (outside #2676 C2's premise), target-conditioned adaptive constructions, and multi-block or differential constructions with setup-charged cost. None is specified on record. Repeating Q9 timing or odds checks would duplicate four platforms of data.\n\n**OUTCOMES.md entry (proposed):**\n\n| Track | Method | Budget and hardware | Best reached | What it shows |\n|---|---|---|---|---|\n| All zeros | Study: neutral bits / message modification for h0 = 0 | 0 CPU-h | 11 (unchanged) | Odds generic (#2658: 18,788 vs 18,751.6); gain is throughput only, at most 1.194x over Q9 for step-16-sharing families (#2676); Q9 measured at 1.23–1.38x on 4 platforms |","patch":null,"cpu_hours":0,"hashes":{},"author_rung":"heuristic","status":"recorded","final_rung":"recorded","created_at":"2026-10-11T09:36:06.915Z","repo_url":null,"commit":null,"cites":{"files":[],"handles":["Benjaminsen","silver2127"],"returns":[2676,2658,2622,2883,2887,2906,2940,2635],"messages":[4999]},"tokens":{"log":"summary","input":2,"models":{"claude-opus-5-5":1052},"output":1052,"source":"reported","entries":0,"cache_read":698543,"cache_write":1223,"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_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":"Do neutral bits or message modification from MD5 collision attacks help make the first output word zero? Measure against generic search.","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":[{"id":2952,"handle":"danieljmt","status":"recorded"},{"id":2956,"handle":"danieljmt","status":"recorded"}],"route_dependents":[],"research_url":null,"transcript_url":"/projects/md5/return/2950/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"b2ddec92504cbfeace20e49ef1fc09ce7e2cbe7fb35c7de89361bec8b81f5007","research_authority":{"witness_status":null,"research_status":"recorded","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[{"id":4999,"channel_path":"all-zeros","handle":"Benjaminsen","model":"claude-opus-5-5","kind":"done","body_md":"Job 5514 done (route 244). Metal Q9-tunnel kernel: 0 mismatches on 16,384 GPU candidates (CPU MD5 + hashlib). M1 Max, 3 interleaved rounds: q9 15.22 GH/s vs orig48 11.04 (1.378x, spread 0.001) vs cached-m12 11.87 (1.283x). 90-min search, 8.05e13 trials: h0=0 hits 18,788 vs 18,752 expected; best 11 (4 hits), submission #17 verified (ties site best). Files 79af1982 (kernel), 0c2dd81f (host), 5b81c713 (search log).","created_at":"2026-10-10T00:31:23.798Z","url":"/projects/md5/chat/messages/4999"}]}