OpenAI Co-founder Andrej Karpathy Joins Anthropic to Improve Claude

It is one of the biggest personnel coups in the current AI talent race: Andrej Karpathy, co-founder of OpenAI and former AI chief at Tesla, has announced his move to US rival Anthropic. The researcher announced the step on Tuesday on X: “Personal update: I’ve joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative.” Karpathy is starting work there this very week.

For Anthropic, which competes directly with OpenAI and Google through its Claude model, the signing is more than a symbolic victory. It shows that the company led by the Amodei siblings is increasingly establishing itself as a magnet for the small group of researchers capable of actively pushing the boundaries of large language models.

Karpathy has also been active as an investor, including in Magic, the startup founded by the two Austrians Eric Steinberger and Sebastian De Ro, which has raised more than half a billion dollars.

New role: Claude is supposed to make Claude better

Karpathy is joining the pre-training team under team lead Nick Joseph — another former OpenAI employee. Pre-training is the computationally intensive phase in which a model acquires its foundational knowledge and capabilities through massive training runs, before fine-tuning and alignment take effect.

His specific task goes beyond that, however: Karpathy is to build a new team that uses Claude itself to accelerate pre-training research. The signal behind this is clear: Anthropic is not primarily relying on sheer compute capacity to hold its own against OpenAI and Google, but on AI-assisted research — that is, on the attempt to automate parts of AI development itself. “I can’t think of anyone better suited to do it,” Joseph wrote in a welcome post on X.

From OpenAI co-founder to Tesla AI chief

Karpathy is one of the few figures in the AI world who commands credibility simultaneously in research, industry, and education. He completed his PhD in 2016 at Stanford University in the field of neural networks and computer vision, was a research intern at Google DeepMind, and subsequently joined the founding team of OpenAI.

In 2017, at Elon Musk’s urging, he moved to Tesla, where he spent five years as Director of AI leading the computer vision team behind Autopilot and Full Self-Driving. In 2022 he left Tesla, briefly returned to OpenAI, and in 2024 founded the AI education startup Eureka Labs. The latter will likely be put on hold — though Karpathy emphasized that education remains important to him and that he wants to return to it “in time.”

Fame in the developer community: “Vibe Coding” and “Zero to Hero”

In the developer and AI community, Karpathy became an iconic figure not least through his educational work. His YouTube series “Neural Networks: Zero to Hero” is considered essential reading for anyone who wants to rebuild neural networks from scratch in code. On X and YouTube he reaches an audience of millions with technical explainer videos and live coding sessions.

He has also coined two terms that have since become part of the established vocabulary of the AI scene:

• “Vibe Coding”: In February 2025, Karpathy used this to describe a new way of programming in which one “fully surrenders to the vibes, embraces exponentials, and forgets that the code even exists” — made possible by increasingly powerful LLMs such as Cursor Composer with Claude Sonnet.
• “Tokenmaxxing”: Karpathy recently described himself as having been since December in a “state of AI psychosis,” in which he aggressively stress-tests frontier models and pushes them toward maximum token output.

What the move means for the talent race

Karpathy’s move is the latest in a series of high-profile Anthropic hires. Just at the beginning of May, Ross Nordeen, a founding member of Elon Musk’s xAI, joined the Claude maker. Simultaneously with Karpathy, cybersecurity veteran Chris Rohlf (previously at Meta and Yahoo) also joined Anthropic’s frontier red team, which stress-tests advanced models against serious threats.

At OpenAI, there have recently been prominent departures in the other direction: former chief scientist Ilya Sutskever (now at Safe Superintelligence Inc) and former CTO Mira Murati (Thinking Machines Lab) left the company in recent years. Anthropic is reportedly on track to surpass OpenAI in private market valuation — and is positioning itself with hires like Karpathy as a serious challenger for the power to define the next generation of LLMs.

Karpathy himself sums it up in his post: he wants to get “back into R&D.” At Anthropic, that will likely mean teaching Claude to improve itself.