[State of RL/Reasoning] IMO/IOI Gold, OpenAI o3/GPT-5, and Cursor Composer — Ashvin Nair, Cursor

This podcast episode features an interview with Ashvin Nair, formerly of OpenAI and currently with Cursor, recorded at NeurIPS. The discussion begins with Ashvin's background in robotics and his transition to language models, including his early experiences at OpenAI. The conversation covers the potential of robotics, the progress and limitations of reinforcement learning (RL) in language models, and the challenges of achieving general AI (AGI). Ashvin shares insights on OpenAI's approach to scaling models, the importance of governance in AI development, and the shift away from a "one model fits all" strategy. The episode concludes with a discussion of Ashvin's move to Cursor, the company's focus on co-designing products with models, and the potential for continual learning in AI systems, including a recruitment call for Cursor.