Andrej Karpathy, the prominent AI researcher who co-founded OpenAI and formerly led AI at Tesla, has officially joined Anthropic. “I think the next few years at the frontier of LLMs will be especially formative,” Karpathy posted on X, expressing his excitement to return to R&D at this critical juncture.
Karpathy started his tenure this week within Anthropic’s pre-training team, working under team lead Nick Joseph. Pre-training is the foundational phase responsible for the large-scale training runs that provide Claude with its core capabilities. It remains one of the most compute-intensive and costly stages of frontier model development.
An Anthropic spokesperson confirmed that Karpathy will spearhead a new team focused on utilizing Claude to accelerate pre-training research. This strategic move suggests Anthropic believes that AI-assisted research, rather than raw compute alone, is the key to staying competitive against rivals like OpenAI and Google. Karpathy is uniquely positioned to lead this effort, given his rare ability to bridge the gap between deep learning theory and large-scale practical implementation.
Karpathy’s career spans pivotal roles at OpenAI and Tesla, where he led the Full Self-Driving (FSD) and Autopilot programs. Most recently, he founded Eureka Labs, an AI-integrated education startup. While Karpathy noted he remains passionate about education and plans to resume that work in the future, his immediate focus has shifted back to the frontier of model development.
In a parallel talent acquisition, Anthropic has also hired cybersecurity veteran Chris Rohlf for its frontier red team. Rohlf, formerly of Meta and Yahoo’s "The Paranoids," brings over two decades of experience to stress-test advanced models against severe cyber threats. This double hiring highlights Anthropic's dual focus on scaling model intelligence and ensuring robust security frameworks.
