Lilian Weng, co-founder of the startup Thinking Machines Lab and formerly VP of AI safety research at OpenAI, has left the company citing her health — and then returned to OpenAI. Her move was announced on July 29, 2026. At OpenAI she will lead a team that accelerates the company's internal research, including work on recursive self-improvement of models.
Key takeaways
- Lilian Weng left Thinking Machines Lab, citing the health toll of startup work.
- She returned to OpenAI, where she was previously VP of AI safety research.
- At OpenAI she will lead a top-level team accelerating internal research.
- The team will support work on recursive self-improvement — a process by which AI systems iteratively become more capable.
- Thinking Machines co-founder Mira Murati bid her farewell publicly on X.
The reason for leaving
Weng announced the decision in an internal Slack message, pointing to the health cost of the startup's pace. She wrote that she does not feel able to continue at the pace a startup requires, and that constant stress and workload had pushed her beyond what her health can physically sustain.
I don't feel I'm able to continue at the pace a startup requires. The amount of consistent stress and workload have pushed me beyond what my health can sustain physically.
Lilian Weng, co-founder of Thinking Machines Lab, in an internal Slack message.
Return to OpenAI and a new role
Weng is no stranger to OpenAI — she previously led AI safety research there as a vice president. Her return was announced on Wednesday, July 29, 2026. This time she is to lead a top-level team whose goal is to accelerate OpenAI's internal research. At the center of her work is recursive self-improvement — a direction in which an AI system iteratively improves itself, becoming ever more capable. It is a topic Weng has written about before on her technical blog.
Her departure from Thinking Machines Lab drew a comment from company co-founder and former OpenAI CTO Mira Murati. On X she wrote that she will miss Weng and that she is glad Weng is putting her health first. The tone was conciliatory, with no trace of rivalry — even though Weng is returning to a direct competitor.
Why it matters
The departure of a co-founder is a signal for a young startup that reaches beyond a single person. Thinking Machines Lab was founded as one of the most loudly funded ventures of the new wave of AI labs, and losing someone from the founding core — to OpenAI, the very company the team spun out of — is telling. At the same time, Murati's public, supportive tone shows the move happened without open conflict.
For OpenAI it is a reinforcement of expertise in exactly the area the company treats as strategic. Handing Weng the work on recursive self-improvement — one of the most sensitive research directions, tying capability growth to safety — signals how highly OpenAI rates the topic. The return of an experienced safety researcher precisely here is no accident.
What's next?
- Weng takes charge of a new OpenAI team focused on accelerating internal research — the first effects will be visible from the team's composition and publications.
- Thinking Machines Lab faces the question of filling the co-founder's role in safety research.
Sources
- TechCrunch — Thinking Machines co-founder Lilian Weng left the company citing health reasons, then joined OpenAI
- Lil'Log (Lilian Weng) — Lilian Weng's technical blog
