Decoded Health

How is rapid progress in AI reshaping healthcare and how should clinicians, researchers, and builders think about what comes next? In our first episode of Decoded Health, we explore the frontiers of model reasoning with Niklas Muennighoff, AI researcher at Stanford. We discuss how modern large language models learn to reason, the role of reinforcement learning, and what recent breakthroughs mean for the future of work. We also dive into AI’s real-world impact on healthcare—from virtual cell models and biological foundation models to radiology and medical imaging—and what it takes to translate cutting-edge research into clinical value. Niklas also shares his personal path into AI, how he develops research taste, and what distinguishes impactful research from hype. Topics covered: LLM reasoning & reinforcement learning AI and the future of jobs Virtual cells & AI in healthcare Building research taste in AI Subscribe for conversations at the intersection of technology, medicine, and the future of human health. Niklas Muennighoff: Niklas Muennighoff, born in Munich, Germany, is pursuing a PhD in computer science at Stanford School of Engineering. He graduated top of his class from Peking University with a bachelor’s degree in business administration. Niklas aspires to contribute to developing artificial general intelligence that can one day help cure human diseases. He led 23 researchers at the nonprofit Allen Institute for AI in building OLMoE, the best open language model for its efficiency when it was released. He previously led the creation of MTEB, an open-source AI evaluation software used by OpenAI and Google with over a million downloads in 2024. At Peking University, Niklas built an AI system to reduce hate speech and placed second among 3,300 teams in a Meta-sponsored competition. His research has earned Best Paper awards at ACL and the ICLR AGI Workshop, and a NeurIPS Outstanding Main Track Runner-Up.