Andrej Karpathy joins Anthropic pre-training team to use Claude to accelerate LLM research

Andrej Karpathy, one of the founding members of OpenAI and one of the most recognized AI researchers in the world, announced on May 19, 2026 that he has joined Anthropic. He will work on the pre-training team under team lead Nick Joseph, where he will build a new group focused on using Claude to accelerate pre-training research — a novel form of recursive self-application for one of the most compute-intensive phases of frontier AI development.

What was announced

Karpathy announced the move directly: "I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D." He added that he "remains deeply passionate about education and plans to resume that work in time," signaling this is a research role focused on training-time work rather than external products.

The specific focus — using Claude to accelerate pre-training research — means Karpathy's team will investigate how frontier models can assist in their own development process: generating data, identifying training failures, proposing architectural changes, or running evaluation loops that previously required human researchers.

Background

Karpathy was an OpenAI co-founder and early researcher before departing in 2017 to lead Tesla's Full Self-Driving and Autopilot programs. He rejoined OpenAI in 2023 for approximately one year, then left in 2024 to start Eureka Labs, an AI education startup. Eureka Labs wound down its activities in early 2026 as Karpathy prepared to return to frontier model research.

Pre-training is among the highest-stakes and most resource-intensive disciplines in AI development. The team responsible for pre-training at a frontier lab shapes the baseline capabilities of every model that follows — it is the foundation on which fine-tuning, RLHF, and deployment safety work builds. A researcher of Karpathy's caliber joining that team is a significant allocation of talent.

Why it matters

The hire follows a pattern of Anthropic attracting researchers who have worked across the AI industry at the highest levels. In June 2026, John Jumper, Nobel laureate and former Google DeepMind researcher, also joined Anthropic. The combination signals a concentration of top talent at a lab that is simultaneously managing the US government's export control suspension of its most advanced models and preparing for an IPO.

For the AI industry more broadly, Karpathy's move from education-focused work back to core LLM research reflects a broader view among top researchers that the pre-training frontier remains wide open. The recursive angle — using Claude to improve the training of Claude — is one of the most watched experimental directions in the field. A dedicated team led by a researcher with Karpathy's depth in both theory and large-scale systems could accelerate the timeline on approaches that have so far been mostly theoretical.