Andrej Karpathy announced Tuesday that he has joined Anthropic, ending roughly two years as one of AI's most visible independents. The OpenAI cofounder and former Tesla AI director confirmed the move in a brief personal update on X.
"I think the next few years at the frontier of LLMs will be especially formative," he wrote. "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, which is the kind of hedge that suggests Eureka Labs is not exactly his priority for the next while.
Why Anthropic
The choice is the interesting part. Karpathy could have gone almost anywhere. OpenAI, where he was a founding researcher and put in two separate stints. xAI. Google DeepMind. He picked the company that, five days earlier, had published a policy paper arguing that the next two years will decide the US-China AI race.
That paper, dated May 14, lays out two scenarios for global AI leadership by 2028 and pushes for American labs to lock in a 12 to 24 month frontier lead before the window closes. It is a maximalist document, and Anthropic has been hiring like it believes its own argument. Jan Leike and John Schulman both left OpenAI for Anthropic in 2024. Karpathy makes three high-profile defectors travelling the same way.
The independent years
Karpathy left OpenAI in early 2024 and spent the following stretch running Eureka Labs, an AI-native education startup, and producing widely-watched lecture series on YouTube. His January 2026 post on agentic coding, in which he said roughly 80 percent of his workflow had moved into Claude Code, was the kind of endorsement no marketing budget could buy. It probably did not hurt that the tool he was praising belonged to his future employer.
He is also, depending on whom you ask, the person who popularized the term "vibe coding" in early 2025, then declared it obsolete a year later in favor of "agentic engineering." Take that as you will.
What it signals
Karpathy picked Anthropic over OpenAI, xAI, and Google, all of which have more capital, more compute, or both. The company was last valued around $380 billion in February, after a fundraising sprint and an open fight with the Trump administration that, oddly, seems to have helped its recruiting more than hurt it.
No word yet on what Karpathy will work on, and Anthropic has not posted an official announcement at the time of publication. The next signal to watch is Google I/O, which opens this week with agentic coding near the top of the agenda.
