Anthropic says it has met its own threshold for AI-accelerated research

Anthropic has declared it has reached its own threshold for AI-driven acceleration in research progress, based on future model expectations and recent evaluations. An independent group, METR, found that Anthropic's model improvements nearly doubled in speed, attributing most of this acceleration to AI. However, AI does not yet fully automate the training pipeline, and some data in the evaluation was redacted by Anthropic.