Skip to the article

HUMAN INTELLIGENCE MINING

Journal

Articles from HIM on what people know and machines miss, how our games work, and how results are proven on Solana.

We let a cheap model take our quiz: 98.1%. So we stopped making quizzes
A cheap text-only model answered 255 of 260 HIM questions, median 652 ms, for a third of a cent. What we measured, and what we built instead.
24 October 2026 · 8 min read · measurement, ai-evaluation, game-design

Play a free round

Read the question, then catch the right answer. Free to play.

Play a free round Follow by RSS