Andrej Karpathy joins Anthropic

Andrej Karpathy said on May 19 that he had joined Anthropic, adding one of the best-known researchers from OpenAI and Tesla to a lab competing with OpenAI, Google and Meta for frontier-model talent. Anthropic said Karpathy started this week on its pretraining team, the group responsible for the large-scale training runs that give Claude its core knowledge and capabilities. (cnbc.com) Karpathy said in a post on X that “the next few years at the frontier of LLMs will be especially formative” and that he was “very excited” to return to research and development. (techcrunch.com) The hire puts Karpathy inside one of the most expensive and technically central parts of model development. Anthropic told TechCrunch he will start a team focused on using Claude itself to accelerate pretraining research. (techcrunch.com) CNBC reported the company said he starts this week. ### What exactly is Karpathy doing at Anthropic? Anthropic said Karpathy joined its pretraining team, which handles the training runs that give Claude models their base knowledge and capabilities. The company told TechCrunch that he is working under team lead Nick Joseph and will build a team focused on using Claude to speed up pretraining research. (cnbc.com) Pretraining is one of the most compute-intensive stages in building a frontier model. TechCrunch described it as the phase where large-scale training runs establish a model’s core capabilities, and CNBC said Anthropic placed Karpathy in the group responsible for that work. (techcrunch.com) ### Why does this move draw attention beyond a normal executive hire? Karpathy is a founding member of OpenAI and later served as Tesla’s director of AI, according to his website and reporting from CNBC and TechCrunch. His site says he returned to OpenAI in 2023-2024 to build a team working on midtraining and synthetic data generation before leaving again. His background spans research, large-scale training and production AI systems. (techcrunch.com) Anthropic’s decision to place him in pretraining gives the move weight because that group sits near the center of how frontier labs improve model quality before products are built on top. That description of the role comes from Anthropic’s own account of what pretraining does. ### What did Karpathy say publicly? Karpathy wrote on X: “I’ve joined Anthropic.” He added that the next few years in large language models would be “especially formative” and said he was excited to get back to R&D. TechCrunch also reported that he said he remained “deeply passionate about education” and planned to resume that work later. (karpathy.ai) Eureka Labs, the education startup Karpathy introduced in July 2024, remains live on its website. TechCrunch said it was not clear whether he would continue with the startup, and Karpathy’s post said only that he planned to return to education work “in time.” ### How does this fit Anthropic’s recent hiring? CNBC reported that Ross Nordeen, a founding member of xAI and a former Tesla employee, also said earlier in May that he was joining Anthropic. (techcrunch.com) TechCrunch separately reported that Anthropic had brought on Chris Rohlf for its frontier red team, which stress-tests advanced models against severe threats. Those additions come as Anthropic and its rivals compete for researchers with experience in large training runs, model safety and deployment. (techcrunch.com) Forbes described Karpathy as one of the most recognizable names in AI research, while CNBC called the move Anthropic’s latest high-profile hire. (eurekalabs.ai) ### What should readers watch next? Anthropic said Karpathy starts this week and will build a new team around AI-assisted pretraining research. Any next public marker is likely to come from Anthropic’s future research updates or from Karpathy’s own posts about the team’s work and his education projects. (techcrunch.com) (forbes.com) (cnbc.com)