Noam Brown on Dwarkesh: why LLMs might hit a wall

Dwarkesh Patel · youtube · 2026-09-27

OpenAI researcher Noam Brown joins Dwarkesh Patel to explain why LLMs might hit a wall: the diminishing returns of inference-time search and scaling, limits of pretraining data and compute, and what decades of poker/game-playing AI research suggest about the next frontier of reasoning models.

Original post →

More from AGI Musings

AGI Musings channel →