Squid

Screenshot-To-React Benchmark

A useful benchmark should measure more than visual similarity. It should show whether the generated React app is understandable, reversible, exportable, and fair to run.

Scoring Criteria

Score each tool against the same prompt, input screenshot, and acceptance checklist.

  • Visual fidelity to the screenshot
  • Resolved internal imports
  • Accessible controls and labels
  • Rollback or checkpoint support
  • Export bundle completeness
  • Credit behavior for failures and repairs

Evidence To Publish

Publish generated files, diagnostics, screenshots, export contents, and the exact credit cost shown before generation.

Why This Matters

The first preview can be impressive while the artifact is hard to own. A benchmark should make ownership and maintainability visible.

Run the benchmark in Squid