Noam Brown, a prominent researcher at OpenAI known for his deep reinforcement learning work, joins Ilge Akkaya and Hunter Lightman from the o1 research team. They discuss the innovative combination of LLMs and reinforcement learning, revealing how o1 excels in math and reasoning. Insights include the use of chains of thought and backtracking to enhance problem-solving. The team shares milestones like the success at the International Olympiad in Informatics and reflects on scaling challenges that may unlock even greater AI reasoning capabilities.