OpenAI co-founder Karpathy has joined the pre-training team at Anthropic.

      Andrej Karpathy, one of the original 11 co-founders of OpenAI, has joined Anthropic's pre-training team. In this new role, he will establish a group that aims to utilize Claude to expedite the most costly stage of frontier model development.

      Karpathy, widely recognized as one of the leading AI researchers globally, announced his transition to Anthropic on Monday. This is a significant recruitment success for Anthropic as they strive to maintain a competitive edge in the large language model sector.

      He will become part of Anthropic's pre-training team, which is headed by Nick Joseph. Karpathy's new team will focus on an innovative goal: leveraging Claude to enhance pre-training research. Pre-training, being the compute-intensive phase that imparts fundamental knowledge and abilities to frontier models, is the most time-consuming aspect of creating systems like Claude. Optimizing this process could transform the financial dynamics of the AI industry.

      In a post on X that garnered 13.6 million views, Karpathy expressed his belief that "the next few years at the frontier of LLMs will be especially formative." He mentioned that he remains "deeply passionate about education" and intends to return to that work eventually.

      This career shift caps a journey that has navigated many significant milestones in contemporary AI. Karpathy completed his PhD at Stanford under Fei-Fei Li, known for ImageNet, where he specialized in deep learning and computer vision. He was among the 11 co-founders of OpenAI in 2015 and contributed to deep learning research before leaving in 2017 to take on the role of director of AI at Tesla.

      At Tesla, he led the computer vision teams responsible for Full Self-Driving and Autopilot—key programs for the company's autonomous vehicle aspirations. After leaving Tesla in July 2022, he returned to OpenAI for almost a year, then left again in 2024 to start Eureka Labs, a company focused on using AI assistants in education. Work at Eureka Labs is currently on hold as Karpathy dedicates his efforts to Anthropic.

      The timing is significant since Anthropic has become a hub for elite technical talent while its main competitor, OpenAI, faces multiple high-profile exits. Over the last two years, OpenAI has seen more than a dozen senior leaders and researchers depart, including CTO Mira Murati and reinforcement learning expert John Schulman, along with three executives leaving on the same day in April 2026.

      Karpathy's recruitment signifies that Anthropic can attract top-tier talent as it expands its research and commercial endeavors. Led by CEO Dario Amodei, Anthropic has caught the attention of investors with a valuation of around $800 billion and is reportedly considering an IPO possibly as early as late 2026.

      Karpathy’s new position also highlights an emerging trend in advanced AI: using existing models to enhance the following generation. If Claude can effectively accelerate its own pre-training process, it would represent a tangible example of recursive self-improvement, a concept closely monitored by the AI safety community. Whether this development is seen as promising or concerning may hinge on the level of trust observers have in the safety-focused culture that Anthropic has developed since its inception.

      For now, it seems Karpathy is where he wishes to be: back in the lab, working on models at the cutting edge.