Thomas Kwa leaves METR for OpenAI

Thomas Kwa announced on LessWrong that he was leaving METR to join OpenAI as a Member of Technical Staff. He said he would start on September 1, 2026.

Kwa will work on measuring and modeling recursive self-improvement: the possibility that AI systems could accelerate AI research and thereby help create increasingly capable successors. He wrote that he wanted to give the world better evidence about whether such a feedback loop is imminent.

At METR, Kwa was a lead author of the organization's widely discussed time-horizon research, which estimates the length of software tasks that AI agents can complete reliably. He said he felt very positively about his roughly 1.7 years at METR and continued to recommend it as a high-impact place for AI-safety researchers.