Anthropic was founded by OpenAI exiles. It’s adding one more. On Tuesday, former OpenAI co-founder Andrej Karpathy announced on X that he will be joining Anthropic, seemingly to work on the lab’s research and development team.
“Personal update: I’ve joined Anthropic. 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. I remain deeply passionate about education and plan to resume my work on it in time.”
Karpathy had an on-again, off-again relationship with OpenAI, serving as a founding member of the group and research scientist from 2015 to 2017, before Elon Musk poached him to serve as Tesla’s director of artificial intelligence. He rejoined OpenAI in 2023 and hung on for about a year before stepping away again in the wake of the very dramatic and short-lived ouster of Sam Altman as CEO—though Karpathy denied his exit had anything to do with that whole situation.
It’s hard not to read Karpathy’s move to Anthropic in the context of the Cold War between the two major frontier labs. Not only is Anthropic made up largely of OpenAI send-offs, but it’s also started to act in an “enemy of my enemy is my friend” fashion. Earlier this month, Anthropic announced that it was going to start using computing power made available by SpaceX (née xAI) despite the fact that CEO Elon Musk has repeatedly referred to the company as “evil.” Turns out he must think it is less evil than a Sam Altman-led OpenAI, and given that his lawsuit to punish Altman failed, might as well lend support to his biggest rival.
Karpathy, the creator of the term “vibe coding,” will reportedly be joining Anthropic’s pre-training lab, per Axios, and will help to launch a new team that will be focused on using Claude, the company’s flagship AI model, to accelerate pre-training research.
Just how that’ll go, we’ll have to see. Karpathy’s credentials are pretty much unassailable, but he’s certainly gotten a bit high off the supply of AI hype at times. He called Moltbook, the social network for AI agents, “genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently”—only for it to be revealed that most of the content on the site was actually human-generated. Earlier this year, he described himself as being in a “state of AI psychosis” and said that he’s embraced “tokenmaxxing,” the practice of using as many tokens as possible to maximize your AI use, whether it’s actually necessary or not.
Whether Anthropic is getting industry icon Karpathy or AI slop-ified Karpathy, simply getting him is something of a coup for the company in terms of attracting talent. It signals that the serious minds in the space are gravitating toward Anthropic—or at least away from Altman. Whatever the price tag to bring him in might be, it could be worth it simply to bolster the perception that Anthropic is the place to be.
