New tiered verifiable math puzzles debut, Grok fails first test

Reddit user jarekduda released 9 verifiable math puzzles graded from human-level to ASI-level, a new benchmark format between fixed benchmarks and open problem lists. In the first test, Grok mistook numerical guessing for proof and drew criticism.

2026-10-25 ~ 2026-10-25 · 2 related posts

• Nine checkable math puzzles from human to ASI level; Grok passed solved ones but mistook a float estimate for a proof — jarekduda · 2026-10-25
• AI puzzles: a graded human-to-ASI problem ladder with checkable proofs — Grok fails first field test — jarekduda · 2026-10-25