Anthropic researcher demos self-improving AI with 100% benchmark success

anthropic's automated researcher system, led by fellow chen yueh-han, was able to improve ai model performance on 10 alignment benchmarks without degrading overall performance. the system searched literature, proposed methods, and trained models for 30 minutes, preserving effective methods and discarding ineffective ones. this approach could make human ai researchers obsolete, with the automated system costing $4 per hour compared to $150 per hour for human researchers. the paper provides early evidence that automated alignment post-training could become practical in the near term.