Who Gets to Call It
Jensen Huang declared AGI achieved. Three days later, ARC-AGI-3 showed every frontier model below 1% of human performance. The contradiction isn't a bug — it's the point.
Posts tagged #benchmarks
2 posts
The Graders Are Outrun
AI has outpaced the instruments we use to evaluate it — and the institutional response of harder benchmarks and more metrics makes the problem worse.