Noam Brown is a research scientist at OpenAI working on reasoning, reinforcement learning, self-play, and multi-agent AI. Before OpenAI he worked at Meta AI and earned his PhD at Carnegie Mellon, where he built the superhuman poker AIs Libratus and Pluribus and the Diplomacy-playing agent CICERO. He is best known as a key contributor to OpenAI’s o1 reasoning models, which use inference-time compute to “think” before responding. He remains at OpenAI leading reasoning research as of 2026.
Recent
- 2026-08-01 Tweeted that an internal version of Astra – OpenAI’s next major model family – solved 10 major open problems in mathematics, quantum complexity, and theoretical CS, and that total compute cost for all 10 proofs was under $2,000 at Sol API prices. Also noted Astra could not solve the Millennium Prize Problems. X · X