{"n": 1, "what": "first view prints: 12 prints from a 37-anchor stick guide, capsule shapes, riso painting, dot screen in print space (cell 14 print px)", "saw": "profile, rear three-quarter, rear, lying read as one dog; the first look-back print (head over the near shoulder, front three-quarter body) T1 was a tangle (tail over the back, head merged with the chest); L_in head did not read as turned in", "changed": "rear-route turn-back prints (RT3/RT2/RT1, head held on the door); L_in head yaw increased"}
{"n": 2, "what": "first frames: MLS forward mesh + bilinear forward splat", "saw": "rigid skull reported anisotropy 50: rigid MLS cannot absorb the print's 2x scale", "changed": "MLS runs in film units (handles and grid halved first)"}
{"n": 3, "what": "turn frames (3.5-3.8 s) with handles lerped in 2D between view prints", "saw": "torso and neck folded (aniso up to 1800, 60+ folded cells): hidden left/right handle pairs coincide in the profile print and separate in the turn, tearing the print between two coincident points; the neck's handles are collinear in the rear view (affine singular)", "changed": "handles follow the stick guide continuously (monotone cubic in yaw/pitch/posture) instead of a 2D lerp; handles closer than 30 print px are merged; an affine layer whose handles are near-collinear falls back to similarity; torso handles reduced to the centre line"}
{"n": 4, "what": "retreat crouch frame", "saw": "coat anisotropy 23: the throat handle (in the torso set) followed the lowered neck and dragged the chest", "changed": "throat removed from the coat's handles (it belongs to the neck)"}
{"n": 5, "what": "composition at 235 px/m", "saw": "the sitting person's hand ran into the lying dog's head and the person was cut by the frame edge", "changed": "back to 215 px/m with the sill at x=330; the person steps back in depth (scale 0.8) to a bench at the far left"}
{"n": 6, "what": "head carriage as 2D rotations scaled by profile-ness", "saw": "in the rear views the head popped up (giraffe) because the lowered head was faded out with the profile factor", "changed": "neck and head pitch moved onto the guide (so a lowered head is lowered in every view); lie/stand head formulas blended by the lie curve (no pop at the switch)"}
{"n": 7, "what": "look-back at 7.46 over the far shoulder", "saw": "mid-swing (7.50) the head passed edge-on and the motion stretch drew the ears into streaks", "changed": "motion-stretch offset capped at 7 px; eye drawn whenever the eye side faces the viewer at all (threshold -0.2)"}
{"n": 8, "what": "hard proof: the same handle move under all-affine / all-similarity / all-rigid / per-region", "saw": "neck anisotropy 20.9 affine vs 12.3 similarity; fore leg 13.9 affine, 1.95 similarity, 30.9 rigid (folds); skull 1.00 under all", "changed": "neck switched from affine to similarity (measured); legs similarity and skull rigid confirmed; proof/families.jpg"}
{"n": 9, "what": "full render 1 (neck similarity), encoded film checked", "saw": "checker failed I1 on lie_down (measured 26.29 s: the ears stay high when lying with the head up, so the top of the dog did not drop); the head vanished at 3.62-3.70 s (rear view: all head handles merged into one, MLS collapsed it to a point); neck folds at 3.58-3.70 and 7.46-7.70 s", "changed": "lie onset measured on the rendered back (torso top); a layer whose handles merge into one is translated; neck handles reduced to withers, chest, poll, skull; fold guard: a layer that folds (or exceeds 6x anisotropy) is re-solved as a two-handle similarity, logged per frame"}
{"n": 10, "what": "full render 2, encoded film: turn strips, lie-down, look-in, stills", "saw": "ALL OK; folded cells 0 in every family; both turns read as one dog in every frame (rear view at 3.62-3.70 s, back shown from 3.79 s; head round to the door while lying at 7.42-7.75 s, body follows via the rear by 8.6 s); the look-in lifted the head high (risk of reading as a lift)", "changed": "while looking in the neck is kept forward and level (+30 deg); frames 622-697 re-rendered; look-in re-checked in the encoded film: head low in the doorway, back of the skull and both ears to the viewer"}
{"n": 11, "what": "critic round 1: DOES NOT PASS (narrowly): look-in reads as looking at the hand; room all orange; torso ball/area jumps 3.70-3.86; strain readout in the scene, rigid 1.74; giraffe neck 8.22-8.34", "saw": "measured torso area jumps of 27-34% at the print swaps (a print warped 20 deg past its own view over-shrinks); lowering the head off profile hid it (headless frames)", "changed": "4 in-between turn prints + evenly spaced keys; off-profile neck carried at 22 deg (refusal) / 32 deg (turn-back); L_in head yaw 84; hand back to the lap by 26.35; lit doorway aperture with a dim room and the leaf hinged on the aperture post; readout moved to the end card with honest rigid-fit wording; full re-render, verified on the encoded film, ALL OK"}