Andrej Karpathy's voice is one of the most quoted in the AI field — partly because of his pedigree (OpenAI founding team, Tesla Autopilot), partly because his explanations are unusually clear, and partly because he's willing to say what other lab insiders won't. His X threads, lectures, and zero-to-hero series have shaped how a generation of practitioners think about neural networks, agents, and the actual mechanics of model training.

Coverage on this page tracks what Karpathy is saying about model behavior, agent design, the limits of current scaling, and whatever skeptical reads on industry hype he's filing this week.