Fighter · visual qualification
The interface did what the tests could see. The tests could not see what the player felt.
The wrong question was easy to test.
The checks asked whether the interface fit its container and whether its geometry remained inside expected bounds. It did. The result still lacked hierarchy and compositional intent.
That was not a failure of the geometry checks. They answered the question they were built to answer. The mistake was treating their answer as a complete product judgement.
Initial belief
Responsive geometry and screenshot checks would prevent the important visual failures.
What the evidence could not see
A page can avoid clipping and still put the wrong thing first. A screen can be dimensionally correct while the player has to work to understand what matters, what changed or where to go next.
The factory therefore added separate visual forensics, full-journey review and an explicit unslop check. Each asks a human reviewer to identify the focal point, hierarchy, competing elements, context and quality of the complete experience.
The lesson travelled
The same distinction later mattered in Bubbles: a technically valid interface was still challenged against its actual product bar. The factory's lesson was not to discard automation. It was to stop asking one instrument to certify a different kind of quality.
What changed in the factory
- Visual review became a named qualification stage.
- Whole journeys are reviewed with the implementation context in view.
- Responsive quality is judged at the states and orientations where people use the product.
- Passing a machine check is reported only as the claim that check supports.
Record boundary
Some source repositories are private. This publication does not reproduce raw issue text or private conversation transcripts.