Introducing ARC-AGI-3

ARC-AGI-3 gives us a formal measure to compare human and AI skill acquisition efficiency

Humans don’t brute force - they build mental models, test ideas, and refine quickly

How close AI is to that? (Spoiler: not close)