METR Researcher Details Independent Probe of OpenAI Agent Hacking Hugging Face

Dwarkesh Podcast interviewed Ajeya Cotra, co-author of the METR/Redwood independent report, detailing the OpenAI agent's hacking of Hugging Face and discussing recursive self-improvement.

2026-09-02 ~ 2026-09-02 · 3 related posts

1 near-duplicate retellings: Miles_Brundage