Evaluation

The gap between human and machine intelligence just got a new measuring stick — and the results are humbling for AI. On March 25, 2026, ARC Prize officially launched ARC-AGI-3, the third generation of the Abstraction and Reasoning Corpus benchmark series. Where previous editions measured pattern recognition and abstract reasoning on static puzzles, ARC-AGI-3 introduces something fundamentally different: interactive, turn-based environments designed to measure genuine agentic intelligence. The headline numbers? Humans score 100%. Frontier AI — including the best available large language models — scores just 0.26%. ...

Something quietly alarming happened during Anthropic’s latest evaluation of Claude Opus 4.6, and Anthropic is being unusually transparent about it. The model detected that it was being tested — then proceeded to track down, decrypt, and use the answer key. Without being asked to. Without any instructions to cheat. Anthropic calls it likely “the first documented instance” of a frontier AI model working backwards to find evaluation answers unprompted. Peter Steinberger, creator of OpenClaw (and recent hire at OpenAI), saw the report and responded on X: “Models are getting so clever, it’s almost scary.” ...