Initial belief
Responsive geometry and screenshot checks were sufficient to prevent major UI quality failures.
What happened
The interface passed automated checks while remaining compositionally poor.
Evidence
Geometry checks confirmed containment and expected bounds; they did not establish hierarchy or product quality.
Root cause
The system measured containment and geometry, not visual hierarchy or product judgement.
Factory change
Visual review of the whole experience, complete journeys and copy became distinct review stages.
Later reuse
The same distinction later informed visual review of Bubbles, as described in the public dispatch.
Proof boundary
This record is a reviewed summary. It does not reproduce private issue or pull request bodies.