{
 "tm": "1RB0LE_1LC1RD_1LA0RB_1RC1RB_1LF0LC_---0RC",
 "bucket_reason": "Priority call. STUCK at informal_proof: the edge lemma (E') is the whole difficulty, and every remaining route is expensive with unknown odds.",
 "question": "Continue investing in the edge lemma (E') or park the machine?",
 "options": [
  {
   "id": "park",
   "label": "Park: STUCK at informal_proof, keep proof.md as the conditional reduction",
   "consequence": "No further spend. Everything except (E') stays checked and documented: rules, bounces, cycle law, and the init at 784,041 steps. Reopen only if a stronger CTL search tool is built for other machines (then run it on #13 in the background) or a closed language for the edge map turns up. Mirror near-twins #34 and R6 share the period-10 bulk and probably the obstacle too."
  },
  {
   "id": "ctl_native",
   "label": "Continue, route 1: native reimplementation of busycoq's stronger CTL modes",
   "consequence": "Days of agent time with unknown odds. mxdys's standard ranges already failed, and plain n-gram CPS fails for n <= 26. If it worked, sections 2-5 of proof.md would become unnecessary."
  },
  {
   "id": "edge_checker",
   "label": "Continue, route 2: edge-only closed set plus a new Coq checker",
   "consequence": "1-3 weeks of agent time on top of finding the closed set, which nobody has. busycoq has no interface for a custom start or a protocol."
  },
  {
   "id": "grammar",
   "label": "Continue, route 3: hand analysis toward a non-regular grammar for the tail",
   "consequence": "Open-ended research with no budget bound."
  }
 ],
 "recommended": "park",
 "rationale": "The tail D is chaotic. It does not repeat in 5.5e6 cycles, its k-factor counts grow about 1.6^k without saturating, and 120 cycles show 3,786 distinct per-cell crossing sequences. Every regular candidate tried so far has failed: SFT_k invariants are refuted with verified chains for every k <= 36, plain n-gram CPS does not close (whole machine n <= 26, edge system n <= 24), and CTL62 searches ran out of memory with no verdict. Nobody knows whether any regular invariant exists, and the cheapest remaining route costs days with unknown odds. The obstacle is not arithmetic, but for proof purposes it behaves like a cryptid: no structure has been found to build an invariant on. The budget does more on machines with validated mechanisms, and the conditional reduction keeps this work reusable.",
 "evidence": [
  {
   "kind": "proof",
   "text": "proof.md 6.4: SFT_k refuted for k <= 22 by sampling and for k = 24-36 by verified chains. Plain CPS does not close. Two CTL62 Nsolve_cert runs reached about 10 GB within 5 min and were killed with no verdict.",
   "href": "proof.md"
  },
  {
   "kind": "refutation",
   "text": "Refutation chains for k = 24, 26, 28, 30, 32, 36, each re-checked in pure Python by scripts/edge_sft/verify_refutations.py.",
   "href": "work/sft_refutations/refute_k36.txt"
  },
  {
   "kind": "data",
   "text": "Orbit to 5.5e6 cycles: no break, no halt, no repeated D, |D| in [144, 564] with no drift, deeper cells chaotic (proof.md 6.3).",
   "href": "proof.md"
  },
  {
   "kind": "rules",
   "text": "Everything except (E') passes check_proof.py: RS/LS/WS/WE and J1-J6 with unknown context, B1-B6, IFACE (the same 32 crossings on 3,000 cycles), INIT at 784,041 steps.",
   "href": "check_proof.py"
  },
  {
   "kind": "status",
   "text": "status.json: STUCK, informal_proof, priority_call = continue on (E') or park.",
   "href": "status.json"
  },
  {
   "kind": "figure",
   "text": "Card Fig 1: one row per bounce. Regular bulk and junction, all the disorder in the tail.",
   "href": "card.html#fig-1"
  }
 ],
 "effort_if_yes": "None now. Status stays STUCK (informal_proof) with the priority call answered. Reopening later costs nothing beyond running whatever new tool exists by then.",
 "updated": "2026-10-10",
 "answer": {
  "option": "park",
  "note": "Dennis: go with all recommendations (2026-10-10)",
  "date": "2026-10-10"
 }
}