Andrej Karpathy Joins Anthropic - The OpenAI Co-Founder Is Now Building Claude

Andrej Karpathy has joined Anthropic's pre-training team, marking another significant move in the AI talent wars. The OpenAI co-founder announced the news himself, stating he's returning to frontier LLM research and development with the team building Claude. Karpathy will work under Nick Joseph, building out a group focused on using Claude to accelerate Claude's own training process. Pre-training represents the most computationally expensive phase of model development, involving massive compute runs that give models their core knowledge base. Anthropic's decision to place Karpathy in this critical role signals their strategic priorities. His career trajectory reads like a chronicle of AI development. After co-founding OpenAI in 2015, Karpathy led Tesla's Autopilot programme until 2022, returned to OpenAI for a year, then launched AI-native education startup Eureka Labs. He indicated he'll resume that educational work "in time". The hire reveals Anthropic's strategic bet: AI-assisted research, rather than raw computational power, will determine competitive advantage against OpenAI and Google. By hiring Karpathy to use Claude for building the next generation of Claude, Anthropic is making this philosophy explicit. The temporary pause of Eureka Labs suggests the opportunity at Anthropic was compelling enough to delay his educational ambitions.