San Francisco, May 19, 2026, 09:03 PDT
Anthropic has picked up Andrej Karpathy, one of OpenAI’s founders and ex-Tesla AI lead, as it looks to boost its research team amid tough competition for AI talent. The Claude maker is adding a high-profile technical hire as rival labs continue hiring sprees.
Karpathy is set to join Anthropic’s pre-training team, the same unit that handles the big compute-intensive training runs behind Claude’s core features, before the model gets fine-tuned for users. TechCrunch said he’ll report to Nick Joseph and plans to start a team focused on using Claude to help speed up pre-training research.
“I’ve joined Anthropic,” Karpathy posted on X Tuesday. He said the next few years for large language models would be “especially formative.” Karpathy said he’s going back to research and development, and plans to get back into education after that. Business Insider
Anthropic said Karpathy is set to start this week. Axios reported his job will focus on using Claude to speed up pretraining research, pointing to a push to automate more of model development instead of only spending on chips or adding engineers.
Bloomberg said Karpathy is taking on research and development work at Anthropic, tied to training new AI models. That puts him in the middle of Anthropic’s race with OpenAI, where model gains come down to how they train, the quality of their data, and how much compute they can get.
Karpathy has worked in multiple high-profile AI groups. He lists himself as a founding member at OpenAI, Tesla’s lead for Autopilot’s computer vision, and then back to OpenAI in 2023 before leaving again to create AI education content.
Anthropic’s new hire brings in a public technical lead as AI firms face scrutiny on more than just new model releases. Lately, investors and rivals want to see if companies can hire people who have actually trained big systems in the past. OpenAI and Google are still the main comparisons here. The fight is over talent, but also over who knows best what to train next and how to do it.
Anthropic bringing on Andrej Karpathy puts a name on the board, but it doesn’t end the model race. Pre-training still eats up big money and compute for anyone trying to stay at the AI edge. said it’s also unclear if Karpathy will keep working on Eureka Labs, the AI education startup he started after OpenAI.
Anthropic is also bringing in more experts. TechCrunch said Chris Rohlf has joined its frontier red team, which pressure-tests advanced AI models for major threats. Georgetown’s Center for Security and Emerging Technology names Rohlf as a cyber researcher with experience in vulnerability research, exploitation, and software engineering since 2003.
Anthropic now faces the question of how quickly Karpathy’s team can bring faster or improved Claude training. For competitors, the takeaway is clear: the AI race remains a talent fight, run by those who build the tech.
