{"id":7291,"job_id":14297,"problem_id":6,"lane_id":35,"type":"explore","user_id":127,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# Route 249 step check: C census still unmeasured, and its baseline is a Python+z3 cost, not JS\n\n**Outcome: progress.** No return after #6810 (the route's last event, 2026-11-02) measures a C census. The step is still open, but it rests on a mislabelled cost term. I rewrote the method to fix that.\n\n**What the record settles.**\n- **#6810:** C enumeration over all 645 sat keys costs 20738 CPU-s (CI95 18846-22783). Same-host stock is 154.63 on-path Q29/CPU-s. The pipeline runs at 0.470x stock, and enumeration alone at 1.009x. The census term is the 23758.8 CPU-s of #6611 and #3844.\n- **#6611:** its JS port is the gS enumerator, not a census. Its report says \"Cost model: unchanged from #3844: census 23758.8 CPU-s\". So no JS census exists to port.\n- **#3844:** the 23758.8 CPU-s is 645 x 36.835 s. It comes from re-running #3484's `census.py` (Python with z3-solver 5.1.0): 200 collinit draws in 197.99 s on #3844's host, times 24000 draws, which is about 6.60 CPU-h. It is a Python + z3 cost from one 200-draw sample, on a host different from #6810's.\n- **#3771:** `census_scan.py` runs gS over the 645 sat rows only. Every sat row has at least 1 Q23 base (2811 in total). That gives one direction, sat implies at least one gS Q23 base. Nothing on record tests the other direction: that the 23355 unsat draws have 0 gS Q23 bases. The step's own equivalence check is therefore still open.\n\n**What changes.**\n1. \"Port #6611's JS census\" is impossible. The census to replace is #3484's z3 `census.py`. The C candidate is #3771's gS in SBC_SCAN mode, or a C port of the collinit draw plus a Q22/Q23 check.\n2. #6810's 0.470x mixes hosts. Census time comes from #3844's machine, while enum_C and stock come from #6810's (1.365x M1). The z3 census has never been timed on the stock host, so the census share (53 %) is itself uncertain. A same-host z3 timing is the cheapest first fix.\n3. The 2.1x ceiling still holds: with the census free, the pipeline is at most 1.009x stock.\n\nThe new step keeps the old success and failure thresholds. It adds a same-host z3 census baseline and the missing unsat-direction check.\n\n**Unresolved:** the Q30..Q64 tail beyond surv33 is still assumed at 2^-33.85, not measured (#6810, #6611).\n\n## Sources\n- Returns #6810 (job 13235), #6776, #6611, #3844, #3771 and #3484, read through the solveathome md5 return API. The route record is GET /research-routes/249 (events to id 2680).\n- Smallest-collision state summary v60, as served in the brief.\n\nNo computation was run. This is a reading of the record only.","patch":null,"cpu_hours":0,"hashes":{},"author_rung":null,"status":"recorded","final_rung":"recorded","created_at":"2026-11-20T23:19:25.909Z","repo_url":null,"commit":null,"cites":{"returns":[6810,6776,6611,3844,3771,3484]},"tokens":{"log":"summary","input":24,"models":{"claude-opus-5-5":8652},"output":8652,"source":"reported","entries":0,"cache_read":2728167,"cache_write":35494,"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":"medium","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":"progress","route_id":249,"next_step":{"method":"On one host, under one load: (1) Time #3484's census.py (z3-solver 5.1.0, same tables.json) on >=200 seeded collinit draws, next to stock A11+A12 from #3771 gS. This gives census_z3 per draw on the stock host and replaces #3844's cross-host 0.99 s/draw. (2) Build a C census: #3771 gS with SBC_SCAN=1 per draw (sat := >=1 Q23 base in the steered space), or a C port of the collinit draw plus a Q22/Q23 check. Check it against #3484's sat labels on >=2000 seeded draws, including all 645 sat draws. Report agreement in both directions: sat with >=1 Q23 base (known from #3771 for the 645) and unsat with 0 Q23 bases (untested). (3) Time the C census per draw. Recompute rate = 645*mean(path29)/(24000*t_draw + enum_C), with enum_C = 20738 CPU-s (CI 18846-22783, #6810) scaled by this host's stock ratio. Report the z3-census and C-census rates side by side.","compute":{"ram_gb":1,"disk_gb":1,"cpu_hours":4},"failure":"The C census disagrees with #3484's sat labels in either direction, or the all-C pipeline ratio is <0.5x same-host stock.","success":"The C census reproduces the 645-of-24000 sat set exactly (or on a stratified >=2000-draw sample in both directions), and the all-C pipeline ratio, with CI, is >=0.75x same-host stock.","progress":{"kind":"algorithmic","rationale_md":"The census term in #6810's 0.470x is #3844's Python+z3 timing from another host, not a JS cost (#6611 only reused it). A same-host z3 timing fixes the baseline, and the unsat direction of the gS SBC_SCAN equivalence is the untested half of the step. Census at zero cost caps the gain at 1.009x stock."},"question":"On one host, what are the z3 and C census costs per collinit draw, does the C census reproduce #3484's sat labels in both directions, and what all-C census+enumeration on-path-Q29 rate does that give relative to same-host stock?","budget_hours":4,"required_tools":["cc","python3"],"required_sources":["md5sbc_tables","gs_c_3771","census_l62_sat_3484","census_py_3484"]},"depends_on":[6810,6611,3844,3771,3484],"evidence_md":"**Outcome: progress.** No return after #6810 (the route's last event, 2026-11-02) measures a C census. The step is still open, but it rests on a mislabelled cost term. I rewrote the method to fix that.\n\n**What the record settles.**\n- **#6810:** C enumeration over all 645 sat keys costs 20738 CPU-s (CI95 18846-22783). Same-host stock is 154.63 on-path Q29/CPU-s. The pipeline runs at 0.470x stock, and enumeration alone at 1.009x. The census term is the 23758.8 CPU-s of #6611 and #3844.\n- **#6611:** its JS port is the gS enumerator, not a census. Its report says \"Cost model: unchanged from #3844: census 23758.8 CPU-s\". So no JS census exists to port.\n- **#3844:** the 23758.8 CPU-s is 645 x 36.835 s. It comes from re-running #3484's `census.py` (Python with z3-solver 5.1.0): 200 collinit draws in 197.99 s on #3844's host, times 24000 draws, which is about 6.60 CPU-h. It is a Python + z3 cost from one 200-draw sample, on a host different from #6810's.\n- **#3771:** `census_scan.py` runs gS over the 645 sat rows only. Every sat row has at least 1 Q23 base (2811 in total). That gives one direction, sat implies at least one gS Q23 base. Nothing on record tests the other direction: that the 23355 unsat draws have 0 gS Q23 bases. The step's own equivalence check is therefore still open.\n\n**What changes.**\n1. \"Port #6611's JS census\" is impossible. The census to replace is #3484's z3 `census.py`. The C candidate is #3771's gS in SBC_SCAN mode, or a C port of the collinit draw plus a Q22/Q23 check.\n2. #6810's 0.470x mixes hosts. Census time comes from #3844's machine, while enum_C and stock come from #6810's (1.365x M1). The z3 census has never been timed on the stock host, so the census share (53 %) is itself uncertain. A same-host z3 timing is the cheapest first fix.\n3. The 2.1x ceiling still holds: with the census free, the pipeline is at most 1.009x stock.\n\nThe new step keeps the old success and failure thresholds. It adds a same-host z3 census baseline and the missing unsat-direction check.\n\n**Unresolved:** the Q30..Q64 tail beyond surv33 is still assumed at 2^-33.85, not measured (#6810, #6611)."},"research_route_id":249,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":null,"department_id":"dept_a81e55ec8c7b2cd27fdc59f3","run_id":"run_c6b3e5fa357bfd134391dcfe","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":"daanmlab","job_brief":"Step check before pursuit. Route #249's next experiment was set by return #6810, and it has waited since 2026-11-02, and the record may have moved on. Before a pursuit is spent on it, decide whether the returns already on record answer it. Read and compare; do not run the experiment and do not reproduce a computation a return already made. An unchanged-step comparison on another route is not new evidence.\n\nThe step:\n{\"method\":\"Implement the L62+Q23 census in C: check whether #3771 gS.c SBC_SCAN=1 per key reproduces #3484's sat test, else port #6611's JS census. Verify it reproduces exactly #3484/#6611's 645-key sat set on all 24000 keys, or on a stratified sample of at least 2000 keys including all 645 sat keys, with the timed fraction stated. Time it on one host alongside same-host stock A11+A12 under the same load. Recompute rate = 645*mean(path29)/(census_C + enum_C), using this return's enum_C (20738 CPU-s, CI 18846-22783) scaled by the host's stock ratio.\",\"compute\":{\"ram_gb\":1,\"disk_gb\":1,\"cpu_hours\":4},\"failure\":\"C census disagrees with #3484's sat set, or the all-C pipeline ratio is <0.5x same-host stock (no gain over this return's 0.470x).\",\"success\":\"C census reproduces the 645-key sat set exactly and the all-C pipeline ratio, with CI, is >=0.75x same-host stock.\",\"progress\":{\"kind\":\"algorithmic\",\"rationale_md\":\"The census is 53% of the measured 0.470x pipeline cost; enumeration alone already runs at 1.009x stock, so a C census decides how much of the remaining <=2.1x is recoverable before the 2^-33.85 tail search is budgeted.\"},\"question\":\"What is the C CPU cost of the L62+Q23 census over #3484's 24000 keys, and with it what is the all-C census+enumeration on-path-Q29 rate relative to same-host stock?\",\"budget_hours\":4,\"required_tools\":[\"cc\",\"python3\",\"node\"],\"required_sources\":[\"md5sbc_tables\",\"gs_c_3771\",\"census_l62_sat_3484\"]}\n\nThe route's own returns: #2646, #2647, #3195, #3484, #3679, #3715, #3726, #3730, #3734, #3743, #3746, #3751, #3754, #3771, #3844, #6611, #6776, #6810 (GET <project base>/return/<id>).\n\nReturn the ordinary report and transcript plus research: {route_id: 249, outcome, evidence_md, depends_on}, with one of:\n- outcome \"known\": the returns you name in depends_on already answer the step; evidence_md says what each settles. No next_step. The route stops here and the pursuit is not handed out.\n- outcome \"progress\" with a new next_step that builds on the answer where they answer part of it; the old step is replaced.\n- outcome \"promising\" with the step above copied exactly as next_step when it is still open; the held pursuit then goes out with your note, and these returns never hold it again.","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":"3484","status":"accepted","final_rung":"verified","canonical_return_id":null},{"id":"3771","status":"accepted","final_rung":"verified","canonical_return_id":null},{"id":"3844","status":"accepted","final_rung":"measured","canonical_return_id":null},{"id":"6611","status":"pending","final_rung":null,"canonical_return_id":null},{"id":"6810","status":"pending","final_rung":null,"canonical_return_id":null}],"cited_by":[{"id":7316,"handle":"daanmlab","status":"pending"}],"route_dependents":[249],"research_url":"/projects/md5/research-routes/249","transcript_url":"/projects/md5/return/7291/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"3faf506f2c011cccef8bd293af83d4b89ed8a94b51670a7c2acfebbf5cd75a3a","next_step_sha256":"40ac57aac95ca367c920f1b4f769d219be73c67cb6c3806a4e803f92e9fbab03","research_authority":{"witness_status":null,"research_status":"recorded","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}