{
 "tm": "1RB0RA_1LC0LD_1LE1LD_0LB0RE_1RF1LB_---0RA",
 "bucket_reason": "Cryptid flag after the funded prove_l9 stage: L9 cannot close by a local invariant, and #28 turns out to be #10 (already a cryptid candidate) mirrored and started in E, with the same orbit from epoch 350,328 on. A priority call.",
 "question": "#28 is #10 mirrored and started in E, and the two orbits merge: park it with #10 as one cryptid candidate, fund the shared Rocq rule reduction now, or fund research on the shared Lemma X?",
 "options": [
  {
   "id": "park_with_10",
   "label": "Park with #10 as one cryptid candidate",
   "consequence": "Status stays CRYPTID_CANDIDATE, with the rules, the conjugacy, the merge point and the conditional proof.md on file. A future proof of #10's Lemma X on the shared orbit transfers to #28 through the mirror map, plus a finite check of #28's epochs before the merge. No agent time now."
  },
  {
   "id": "joint_reduction",
   "label": "One stage (at most 2 h): Rocq rule reduction for #10, its twin and #28",
   "consequence": "One .v proves the round rules and the epoch map as AntiHydra.v does, so that halting reduces to an eat that sees an s block. This is the cryptid-protocol deliverable that #10 still owes. It is not a kill: there is no nonhalt lemma."
  },
  {
   "id": "lemma_x_research",
   "label": "Fund a research stage on the shared Lemma X",
   "consequence": "Tries to show that the orbit never hits the exact-carry patterns (PC = 2^D plus a sparse tail) that #10's family uses. #10's stage found those patterns form an infinite family, so a finite case analysis is ruled out. This is a long shot."
  }
 ],
 "recommended": "park_with_10",
 "rationale": [
  "Same table under mirror + renaming; the orbits coincide from our epoch 350,328.",
  "#10's counter-family satisfies #28's full invariant, then reaches an unsafe eat.",
  "L9 is proved except one exact-carry case: #10's open Lemma X, so its call applies."
 ],
 "evidence": [
  {
   "kind": "fact",
   "text": "Conjugacy: A..F -> E,A,B,F,C,D mirrored, 12/12 entries, 3,000,000 steps lock-step; merge: our epoch 350,328 = #10's cascade state at t = 350,272",
   "href": "scripts/conj10.json"
  },
  {
   "kind": "fact",
   "text": "#10's family sigma0(B) in #28's rules: invariant holds, L9 then fails twice, and the third flush is late (an eat sees an s block), for every B = 8..18",
   "href": "scripts/family10.json"
  },
  {
   "kind": "fact",
   "text": "Real machine at B = 8: 779/779 round starts match, the unsafe eat comes at N = 194, then the machine recovers (no halt in 2.8e6 steps)",
   "href": "scripts/tm_family_B8.json"
  },
  {
   "kind": "fact",
   "text": "Debris lemmas W1-W5 (L9 everywhere but case H): 100,000 random states, 0 violations; in case H, 1,836 of 34,699 successors exceed s = 1",
   "href": "scripts/l9_reduce.json"
  }
 ],
 "effort_if_yes": "park_with_10: none. joint_reduction: one stage (at most 2 h), a new file with no nonhalt claim. lemma_x_research: open-ended, likely several stages with a low chance of success.",
 "updated": "2026-10-11",
 "answer": {
  "option": "park_with_10",
  "note": "Dennis: apply recommendations (2026-10-12)",
  "date": "2026-10-12"
 }
}