{
  "id": "BC-modes",
  "frozen": "2026-10-12T13:35:44+00:00",
  "question": "On a new brief, does a critic that may go bold (name one layer the painter may destroy, and its ulcer) move that layer, compared with a critic that may only refine? Which layer does it name, and where does the painting land?",
  "start": "SEL1-easel (S09: the woman in the shaded pier between lit window and door, the chair in sun on the pavement)",
  "paintings": {
    "BC-P1": "critic chooses refine or bold at each critique (prompts/critic-modes.md)",
    "BC-P2": "critic may only refine"
  },
  "same_in_both": "fresh painter per painting with an identical packet; critiques after material sittings 2, 4 and 6 by a fresh critic each; 10 material sittings, 900 s replay; no completion tests, no arbiter; the painter ends by holding the state it keeps",
  "measures": {
    "per_critique": "mode, layer, ulcer, mode_why (from the record)",
    "per_painting": "geometry: edge-map correlation of the held state with S09's block-in (rgi/geom.py logic); light placement: median luminance in S09's woman, chair, window and shade regions",
    "moved": "a named layer counts as moved if the next painted state changes it visibly (geometry below 0.80 against the state the critique read, for a geometry ulcer; a woman/chair/window luminance order change, for a light ulcer), else recorded as held; read with Russell's eye"
  },
  "prediction_by_claude": "BC-P1's critic goes bold at least once, most likely on light or meaning (the woman reading as an emblem); BC-P2 refines the S09 arrangement and ends close to it (geometry above 0.8).",
  "limits": "One painting per arm: a description, not an effect."
}