Andrej Karpathy, a founding member of OpenAI and one of the highest-profile machine-learning researchers in industry, announced on Tuesday evening that he was joining Anthropic's pre-training organisation. The move is the most consequential single talent transfer between the two AI labs since the 2021 departures of Anthropic's founders from OpenAI.
Mr Karpathy, 39, said in a post on X that he would help launch a new Anthropic team that uses Claude itself to accelerate pre-training research — work that ranges from synthetic-data generation through to active curriculum design for the next foundation-model run. He framed the move as a return to the technical depth he had stepped away from in 2024 in favour of his Eureka Labs education start-up.
The hire lands at a moment of unusual financial momentum for Anthropic. The company's revenue is on track to more than double to $10.9 billion in the June quarter, with an estimated $559 million in operating profit — what would be its first profitable quarter, arriving roughly two years ahead of internal projections.
Anthropic also expanded its compute relationship with SpaceX this week, committing to roughly $1.25 billion per month through 2029 for capacity on the Colossus supercomputing cluster — a relationship now visible from the other side in SpaceX's IPO prospectus filed Tuesday.
OpenAI confirmed Mr Karpathy's departure briefly without elaboration. A spokesperson said "we wish Andrej well" and pointed reporters to the company's recent geometry-problem announcement, in which a general-purpose reasoning model autonomously solved a long-standing conjecture — an internal milestone OpenAI has used to push back against the narrative that Anthropic has the pre-training lead.