Real approach through the arch, followed by a late environment rewrite / replacement into ruins.

The five clips were concatenated directly with no editorial transitions. Perceived camera speed changes abruptly at some shot boundaries, most obviously A→B / B→C. A speed change can be expressive, but today it is accidental. Cinematographer planning should eventually consider exit velocity from one shot and entry velocity into the next. Accidental speed jumps should not simply be repaired in post.

Shot duration should not necessarily be globally fixed. The ~5-second generative cadence in Digging in the Dirt became perceptible as a repeated rhythm. Future direction: Screenwriter owns narrative pacing; Cinematographer decides practical shot duration and physical camera pace within provider limits. If a move cannot be covered naturally in one supported duration, subdivide rather than forcing it. No schema yet.

This run is neither “Camotion solved traversal” nor “Camotion failed.” 01.12 established the practical limit of strong baked-exposure motion encoding on Ghost Library. 01.8 remains the current Camotion baseline. 01.5 / strength 0.08 remains the conservative directed video baseline where appropriate. The complete system can nevertheless produce genuinely compelling traversal shots. Camotion artifacts remain visible in conditioned stills, but they can become less perceptually important during successful moving shots. The movie-level metric is increasingly: does the model shoot the route? — not: is A′ aesthetically clean as a still?

Do not reopen the 01.12 brute-force strength/sigma/kernel branch. Scene-aware selection from {.02,.04,.08} is completed Phase 1 evidence; Camotion Phase 1 is frozen. See chapter 08.