Generate fresh grid puzzles that match the measurable characteristics of the public ARC I1 evaluation set without publishing another benchmark.

🔒 This use case is in the locked catalog.

The description, method, result and source video — the free sample is the first 50; the rest come with a library pass.