Dwarkesh Podcast: John Schulman & Frontier Researchers on RSI, Chinese Labs, and RL Paths to AGI

_AndrewZhao · x · 2026-09-12

Dwarkesh Patel released a new episode sitting down with John Schulman, Cameron O'Reilly, and Beren Millidge — researchers from "openish" frontier companies — to dig into what's actually happening at the frontier. Topics include: steelmanning the case against recursive self-improvement, what's driving Chinese labs' progress, training automated AI researchers, whether long-horizon RL elicits AGI, the sim-to-real gap, how much progress data explains, why RL works so well, Move 37 and entropy collapse, and rapid-fire timelines.

Related event: Schulman and fellow researchers join Dwarkesh podcast to discuss RSI and AGI(5 posts)→

Original post →

More from AGI Musings

AGI Musings channel →