The ARC-AGI-3 challenge launches today as the first interactive reasoning benchmark which stumps current frontier LLMs. Karl ...