{
 "tm": "1RB1RF_0RC0LE_1RD0RF_1LB---_1LB1LE_1RA0LC",
 "bucket_reason": "Priority call (cryptid protocol) after Wave 6 stage 1. Halting comes down to exact-timing coincidences between wall births at the left end and wall arrivals far to the right. No CTL certificate.",
 "question": "Park W6-7 as a cryptid candidate, or fund one stage for an accelerated simulator that hunts for the halt?",
 "options": [
  {
   "id": "park",
   "label": "Park: CRYPTID_CANDIDATE, no more stages",
   "consequence": "Nothing more is spent. The write-up is the particle picture, the halting chain, scripts/stage1/chain_checks.py and the figure. Siblings 248 and 582 run the same table, so they are parked with it."
  },
  {
   "id": "hunt",
   "label": "One 2 h stage: accelerated simulator and halt hunt",
   "consequence": "Build an event-level simulator for the particle picture, check it tape-for-tape against direct simulation (1e8, 1e9, 3e10), run it as far as it goes, and log every frozen-'1:' episode. Births thin out over time (1.55% to 0.54% of bounces), so the expected halt is at least about 1e16 steps and the likely outcome is no halt yet. A halt found this way would settle the question, but it is not a kill: certifying a halt that far out needs a runner we do not have."
  },
  {
   "id": "lock",
   "label": "One 2 h stage looking for a timing lock (the only route to a non-halting proof)",
   "consequence": "Look for an invariant that ties left-end birth bounces to distant wall arrivals and rules out both coincidences. The data already show no lock (same-bounce birth and annihilation 14 times, against 22.7 expected), so STUCK is likely."
  }
 ],
 "recommended": "park",
 "rationale": [
  "Only changing when 3 defects reach the left edge makes the real tape (step 1.24e9) halt.",
  "Nothing in the data rules those timings out, and no CTL-style certificate can.",
  "A proof or a certifiable halt is out of reach in one stage; parking frees budget for others."
 ],
 "evidence": [
  {
   "kind": "figure",
   "text": "The halting chain on the real tape: a wall leaves a 000 (red) at row 52, and a second wall reaches it at row 105, where the machine halts. Magenta ticks mark the 3 edge-timing edits.",
   "href": "img/halt_chain_rows.png"
  },
  {
   "kind": "fact",
   "text": "chain_checks.py: a wall reaching a frozen '1:' makes a 000 (36/36); a wall reaching a frozen 000 halts (105/105); a wall in a pure field moves one run per bounce (324/324); the right end absorbs walls safely (216/216).",
   "href": "work/stage1/chain_checks.json"
  },
  {
   "kind": "fact",
   "text": "Real run to 3e10 (5.2M bounces): no interior 000; 11 '1:' episodes, none frozen. That matches chance, since births happen on only 0.5-1.5% of bounces.",
   "href": "work/stage1/ones_3e10.txt"
  },
  {
   "kind": "fact",
   "text": "CTL62 sweep: 235 runs (NG, NG with history, RWL_mod), 0 true.",
   "href": "scripts/ctl/sweep_A.jsonl"
  }
 ],
 "effort_if_yes": "park: none. hunt: one 2 h stage, plus one core for the long run. lock: one 2 h stage.",
 "updated": "2026-10-14",
 "answer": {
  "option": "park",
  "note": "",
  "date": "2026-10-14"
 }
}