Via newsbytesapp.com
The former VP of research and safety returns to the company she left in late 2024, this time focused on AI systems that can improve themselves
Lilian Weng spent nearly seven years at OpenAI before departing on November 15, 2024. She rose to become VP of Research and Safety, a role that put her at the intersection of OpenAI’s two most existential priorities: making models smarter and making sure they don’t go sideways.
After leaving OpenAI, Weng co-founded Thinking Machines Lab in February 2025 alongside former OpenAI CTO Mira Murati and other collaborators. She also took on a Distinguished Fellow position at Fellows Fund in December 2024. She stepped down from Thinking Machines Lab in July 2026, citing health reasons and burnout.
On July 4, 2026, Weng published a detailed exploration of recursive self-improvement on her blog, Lil’Log. That blog has long been a go-to resource for AI researchers and practitioners. Her post laid out frameworks for thinking about how self-improving systems could actually be built and constrained.
Recursive self-improvement is the idea that an AI system can make itself better, and then that improved version can make itself even better, and so on. It sits at the heart of what some researchers call the path to artificial general intelligence, and raises profound safety questions about control and alignment.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our