Andrej Karpathy joins Anthropic on the pre-training team.

Andrej Karpathy joins Anthropic to lead a pre-training team under Nicholas Joseph, using Claude to accelerate LLM research and automate model training.

Andrej Karpathy joins Anthropic. A recognized figure in the field, he was one of the co-founders of OpenAI, then led AI at Tesla, where he spearheaded Autopilot and computer vision, before a second stint at OpenAI. He is also known for his online courses.

According to his announcement and specialized media, he is joining the pre-training team, responsible for the massive initial training of models, under the direction of Nicholas Joseph. His mission: to build a team dedicated to using Claude to accelerate pre-training research, an iterative loop where a language model is used to improve the training of subsequent ones. The researcher states he wants to fully immerse himself in R&D at the cutting edge of LLMs, which he considers particularly formative for the coming years, while retaining a commitment to teaching, to which he plans to return later.

His arrival highlights the battle between Anthropic, OpenAI, Google, and Meta to attract top talent, particularly concerning scaling laws, training data, and model architecture. The post sharing the news quickly garnered millions of views.