Lilian Weng returns to OpenAI a day after leaving Thinking Machines Lab due to health

On July 27, Thinking Machines Lab co-founder Lilian Weng publicly shared a farewell letter to employees, stating that sustained stress and workload had exceeded what her body could handle, and she could no longer maintain the pace required by a startup. July 28 was her last day at Thinking Machines. The very next day, OpenAI confirmed to TechCrunch and Business Insider that Weng would return to lead a senior internal research team supporting cross-cutting research including recursive self-improvement, where AI systems help improve training methods, research processes, and next-generation models. Weng first joined OpenAI around 2018, working on the Dactyl robot project, then led applied AI research and Safety Systems, becoming VP of Research and Safety in August 2024 before leaving in November 2024 to co-found Thinking Machines with Mira Murati, John Schulman, Barret Zoph, Luke Metz, and Andrew Tulloch. During her time at Thinking Machines, the lab launched Tinker, Interaction Models, and the open-weight model Inkling, which has 975 billion total parameters and 41 billion active parameters. Weng's personal blog posts in 2025 and 2026 focused on reasoning, test-time compute, and self-improvement, aligning with her new role. Her departure letter made clear she loved AI and was leaving the co-founder role, not the field. Of the original six co-founders, only Murati and Schulman remain; 13 of the initial 42 founding members have left. Murati responded supportively, saying she would miss Weng and was glad Weng prioritized her health.