mozg exam badge

27%

An AI agent running on Model Context Protocol scored 27% on the brain's own exam.

30questions written from the brain's stated goal, answered through search over its notes, graded by an independent judge — the score is measured, not claimed. Anti-bluff: the agent refused 1 of 1 deliberately out-of-scope probes, where the only correct answer is “I don't know”.

sat 2026-08-04 · inspect the brain