{
 "tm": "1RB0RA_1RC1RF_1LD0LC_1RA1LE_1RA0LC_---0RD",
 "bucket_reason": "Med confidence after proof_now: proof.md is a conditional proof, and one of its three flagged lemmas (F3, the MS overflow bursts) is a third nested counter level with no rule.",
 "question": "Spend the next stage on F3 (the unbounded overflow bursts into the throttle), on F1 and F2 (the pause counter and the absorption tail), or park?",
 "options": [
  {
   "id": "f3_first",
   "label": "One stage (2 h) on F3, the riskiest lemma",
   "consequence": "Find a closed invariant family for the throttle under bursts: 6/40 random bursts leave dead slots or big-region leftovers, though all were safe. Decode the big-region counter (0011/0111/0001), write zone-T rules for B^j A^n and test j up to 1e6 natively. If no closed family emerges, the far-future regime (m ~ 236 on) is open: park or a priority call."
  },
  {
   "id": "f1f2_first",
   "label": "One stage (2 h) on F1 and F2",
   "consequence": "Close F2 as 40 finite symbolic runs, using a shift rule over the big region. Write the pause-mode left counter as per-trip rules in the now-autonomous left zone. This leaves only F3 open, but F3 is where the proof could fail."
  },
  {
   "id": "park",
   "label": "Park",
   "consequence": "No spend. It leaves a conditional proof.md with a full zone decomposition, exact middle and throttle rules, and three flagged lemmas."
  }
 ],
 "recommended": "f3_first",
 "rationale": [
  "F1, F2 are routine (F2: 40/40 reachable throttles); F3 is unbounded in burst size, no rule.",
  "The throttle breaks on arbitrary visit words (B A^7 B A^8 B), so F3 hinges on burst shape.",
  "Closing the riskiest lemma first decides if formalization is worth starting."
 ],
 "evidence": [
  {
   "kind": "figure",
   "text": "Fig 3: the m = 14 creep. The thin front is the glider, which leaves the middle as five B visits.",
   "href": "img/creep_special_1.487e8.png"
  },
  {
   "kind": "fact",
   "text": "Zones: b18 visits only AC/DC (833,504) and b618 only AE/BE (803,803) to 1.3e10; middle rules exact on 14,428/14,428 visits.",
   "href": "data/proof/check_proof_long_seed1.txt"
  },
  {
   "kind": "fact",
   "text": "F3: 40/40 random bursts B^j A^2600 safe, but only 34/40 end canonical (dead slots, big-region leftovers).",
   "href": "check_proof.py"
  },
  {
   "kind": "link",
   "text": "proof.md (conditional; flagged lemmas in section 8)",
   "href": "proof.md"
  }
 ],
 "effort_if_yes": "One 2 h stage, light native compute (zone T simulations in C). Formalization is several stages after F1-F3 are closed.",
 "updated": "2026-10-10",
 "answer": {
  "option": "f3_first",
  "note": "Dennis: go with recommendations (2026-10-11)",
  "date": "2026-10-11"
 }
}