Andrej Karpathy, the man who taught Tesla’s cars how to see and gave OpenAI its founding spark, has defected to Anthropic. The move, announced this week, marks the most significant talent shift in the AI race since the invention of the transformer. Karpathy, who commands nearly 2 million followers on X, said the next few years at the frontier would be “especially formative” as he retreats from education back into the research trenches.

The defection comes as Anthropic reportedly enters funding talks at a valuation approaching $1 trillion. Karpathy is the high priest of what he calls “vibe coding”—a world where you simply describe what you want and the machine builds it. But while the silicon elite celebrates, the Wall Street Journal is already warning of a “vibe slop” crisis, where dangerous, AI-generated code floods the world’s software supply chains.

Karpathy is a character out of a modern epic, a man whose career move can swing billions in market cap with a single post. At Anthropic, he will likely oversee the refinement of the Mythos model, the gated architecture that is currently being leased by the Pentagon for autonomous cyber-defense. While traditional engineers worry about their jobs, the superstars of the Cognitive Enclosure are playing a different game, turning the act of creation into a luxury Veblen good for the trillion-dollar class.