{
 "tm": "1RB0LD_0RC1RA_1RD1RF_1LA0LE_0RA0LC_0RA---",
 "bucket_reason": "Mechanism confidence med, so proof work waits for you. The counter, both edge modes and the reset are unified and validated, but the left zone (zero-length runs) has no parametric rules.",
 "question": "Fund one more deepen stage on #9's left zone (parametric rules for zero-length runs, plus validated marker episodes), aiming to reach high confidence and go on to proof.md?",
 "options": [
  {
   "id": "left_zone",
   "label": "One targeted stage on the left zone, then proof.md",
   "consequence": "About 2 h of agent time and 1-2 processes. Fit parametric rules where the carry meets zero-length runs (D-D adjacency, D2 0 X, D3 P^0 X/D). Validate the 3 marker-episode shapes and the start/end episodes on random instances. Draft the generation invariant. If these close: proof.md, then a formalization in the style of Longitudinal/SBCv6a."
  },
  {
   "id": "proceed",
   "label": "Start proof.md now, at med confidence",
   "consequence": "Writes the invariant around the validated parts. The left zone still has to be ruled before the proof is complete, so this mostly reorders the same work."
  },
  {
   "id": "park",
   "label": "Park",
   "consequence": "Costs nothing. It drops a machine whose core is validated and which has an exact trajectory twin."
  }
 ],
 "recommended": "left_zone",
 "rationale": "Almost all of #9 is validated. The sweep model matches 3,732 of 3,732 and 16,391 of 16,391 real right-edge sweeps (steps 3.8e7-3e9), with nothing unmodelled. Round 2 showed that the reset is the same P<->Q conversion every sweep makes; the Q-edge rule passes 600/600 and the post-regrowth transient 400/400. The DA/reset alternation fits all 10 non-initial left events and predicted the next two in the fast-forward. Run lengths only change timing, so nothing points to a cryptid. One region keeps confidence below high: the left zone, where new digits are born and the marker episodes happen. It has been seen only as fixed-shape episodes (21 real and 20 capped instances), not as rules. Fitting those rules is ordinary work with the existing tools, and the generation invariant needs them whichever option is chosen. A proof also covers the exact trajectory twin 1RB---_0RC1RA_1RD1RF_1LE0LF_1RB0LD_0RE0LC. The formal proof will be large (no template within 8 transitions; SBCv10-sized), so pin down the left zone first.",
 "evidence": [
  {
   "kind": "validation",
   "text": "Gen-2 sweep model predict2: 3,732/3,732 (steps 3.8e7-2e8) and 16,391/16,391 (2e8-3e9) real sweeps match exactly, 0 unmodelled.",
   "href": "scripts/deepen_check_gen2_2e8_3e9.json"
  },
  {
   "kind": "validation",
   "text": "Q-edge rule (regrowth) 600/600 random instances, post-regrowth transient 400/400, wall-crossing rules 300/300.",
   "href": "deepen.json"
  },
  {
   "kind": "sim",
   "text": "Real gen-1 reset: 1,001 of 1,065 inner sweeps follow the Q-edge model, and the other 64 fall into 8 fixed-shape episodes. 2nd reset (runs capped at 64): 1,081 of 1,442, with the rest in 25 episodes.",
   "href": "notes.md"
  },
  {
   "kind": "analysis",
   "text": "Left-overflow type rule: a tape starting with 7 ones gives a DA overflow, one starting with 9 ones a reset. It fits all 10 non-initial left events of gen 1 and correctly predicted DA, then reset, in the gen-2 fast-forward."
  },
  {
   "kind": "risk",
   "text": "The left zone has no parametric rules: zero-length runs, the raw prefix (94 cells after gen 1, 194 after gen 2), the 3 marker shapes (observed only) and the start/end episodes. Claims past 2e10 steps rely on run-capped simulation."
  },
  {
   "kind": "sim",
   "text": "No halt to 2e10 steps by direct simulation (56,806 sweeps)."
  },
  {
   "kind": "doc",
   "text": "Twin: 1RB---_0RC1RA_1RD1RF_1LE0LF_1RB0LD_0RE0LC has exactly the same trajectory.",
   "href": "../../reports/targeting.md"
  },
  {
   "kind": "figure",
   "text": "Fig 3: digit lines cross the reset walls without bending.",
   "href": "img/phase_1e9_left.png"
  }
 ],
 "effort_if_yes": "One deepen stage: about 2 h of agent time and 1-2 heavy processes. If the left zone closes: proof.md, then a formalization of about 50 KB (SBCv10-sized), over several stages.",
 "updated": "2026-10-10",
 "answer": {
  "option": "left_zone",
  "note": "Dennis: go with all recommendations (2026-10-10)",
  "date": "2026-10-10"
 }
}