How we score
Each build is scored out of 100. 70 points are for the app, 30 for the run that made it.
| Part | Points | What it measures |
|---|---|---|
| Passes the checklist | 30 | Each build comes with a list of things the app must do. Points are split across that list. |
| Looks and works right | 15 | Layout, mobile, loading and error states. |
| Code | 10 | Readable structure, no secrets in the repo, a README. |
| Deploys | 10 | Works at a public URL on the first deploy. |
| Lighthouse | 5 | Accessibility and performance scores, averaged. |
| Time | 10 | Fastest tool gets 10. Three times the fastest time or slower gets 0. |
| Cost | 10 | Same scale as time, using real dollars spent. |
| Manual fixes | 10 | Starts at 10. Minus 2 each time I had to edit code by hand. |
Rules
- Every tool gets the same spec, pasted in word for word.
- I use the cheapest paid plan that can finish the build.
- The clock starts at the first prompt and stops when every checklist item passes.
- 2 hours max. If it isn't done by then, it's marked as not finished.
- I record the screen for every run and report raw credits next to the dollar cost.