The frozen prompt
Every run received this text and nothing else. No follow-ups, no corrections.
Weekender v1 One prompt. One starter. No steering. 17 finished apps.
Pick a frame to use the app. Arrow keys move across the sheet.
Every run received this text and nothing else. No follow-ups, no corrections.
What this set of runs supports, and what it does not.
So the only variables left are the model and how long it thinks.
The score reviews engineering. It does not score how the app looks.
Credits are the billing unit applied to token usage. Plan allowances are a measured snapshot, not a fixed quota.
Estimates use the token rate card checked 14 July 2026. Consumption varies with model, token mix, task complexity, and codebase size. Rate card
Runs are captured locally and published as static previews. The public page never starts a model run.