Lilian Weng, a co-founder of Thinking Machines Lab, has left the AI startup and returned to OpenAI, the company she departed roughly a year earlier to help start Thinking Machines.

Weng announced her exit from Thinking Machines in an internal Slack message, citing the toll of startup-level workload on her health. “I don’t feel I’m able to continue at the pace a startup requires,” she wrote, adding that the consistent stress and workload had “pushed me beyond what my health can sustain physically.”

Within days of that announcement, reports confirmed Weng was rejoining OpenAI to lead an internal research team focused on recursive self-improvement, a research area centered on building AI systems capable of improving their own capabilities.

Weng previously served as OpenAI’s vice president of AI safety research before leaving to help found Thinking Machines Lab in February 2025. The startup was founded by Mira Murati, OpenAI’s former chief technology officer, alongside Weng and four other industry veterans, including chief scientist John Schulman.

With Weng’s departure, only Murati and Schulman remain from the six founding members who launched Thinking Machines Lab a year and a half ago. The company has not commented publicly on how it plans to fill the gap left by Weng’s exit or whether additional leadership changes are expected.

The move underscores how competitive and demanding the frontier AI research field has become, even for well-funded startups founded by former executives at leading labs. Thinking Machines Lab had positioned itself as a serious contender in foundation model research, drawing significant investor interest on the strength of its founding team’s pedigree.

Weng’s return to OpenAI puts her back inside the company she helped build safety practices for during its earlier years, this time focused on a research direction that touches directly on how future AI systems might accelerate their own development.