Overview
- Lilian Weng said she stepped down from her cofounder role at Thinking Machines Lab because sustained stress and repeated illness left her unable to keep up with the pace of a startup.
- OpenAI confirmed it has hired Weng to lead a team that will accelerate internal research on recursive self-improvement, the process by which models help design and improve later models.
- Weng framed the move as a shift to a more predictable, scoped role rather than cofounder-level operational pressure and did not disclose a specific medical diagnosis.
- Thinking Machines has seen substantial early-team attrition since its February 2025 launch, and reports note uncertainty about whether company leadership knew Weng planned to rejoin OpenAI.
- The case highlights two industry trends: high personal cost from startup ‘sprint’ culture and continued consolidation of top AI talent into large labs, a shift that also focuses attention on the safety stakes of recursive self-improvement.