Lilian Weng leaves Thinking Machines and returns to OpenAI to lead internal research

On July 29, 2026, OpenAI confirmed that Lilian Weng would rejoin the company, two days after the Thinking Machines co-founder announced her departure on health grounds. Weng previously served as OpenAI’s vice president of AI Safety Research; she will now lead a top-level team dedicated to accelerating internal research.

Why the departure matters

Weng’s decision exposes a pressure point for AI startups competing for scarce senior researchers. Founders carry responsibility for hiring, fundraising and strategy alongside technical work, making sustained workloads a business continuity risk as well as a personal concern.

In an internal Slack message later shared on X, Weng said she had considered the decision for several months.

“I don’t feel I’m able to continue at the pace a startup requires. After thinking about it for several months, I ultimately have to admit that the amount of consistent stress and workload have pushed me beyond what my health can sustain physically.”

Thinking Machines co-founder and former OpenAI CTO Mira Murati publicly supported the decision to prioritise health. It remains unclear whether Murati knew at the time that Weng planned to return to OpenAI.

A new mandate at OpenAI

An OpenAI spokesperson said Weng will lead a top-level team supporting work across research groups. Its remit includes recursive self-improvement—the process by which an AI system iterates on itself to become more capable.

The return may appear at odds with Weng’s concerns about pace and stress. However, leading research inside an established organisation can involve a different burden from operating as a startup co-founder, even when the scientific mandate remains demanding.

Her exit also changes the context around Thinking Machines’ interactive models and AI governance agenda, as one of the startup’s founding researchers moves back to a major competitor during an intense contest for AI talent.

For business leaders, the practical lesson is to treat sustainable workload and succession planning as operating requirements. Critical expertise concentrated in one founder or researcher can quickly become delivery, governance and retention risk.