Peter Kirgis

A preprint posted to arXiv on 29 July 2026 and revised on 7 August 2026 introduces 'shadow evaluations,' a method for measuring AI agents' ability to conduct open-ended research, and reports that fron…

A new study led by Peter Kirgis and Sayash Kapoor at Princeton University found that AI agents are not yet capable of conducting open-ended AI research, as they lack the judgment and creativity needed…