Anthropic hires Andrej Karpathy

Andrej Karpathy said on May 19 that he had joined Anthropic, giving the AI startup one of the best-known researchers in large language models as it expands work on core model training. Anthropic said Karpathy started this week on its pre-training team, the group responsible for the large-scale training runs that give Claude its underlying knowledge and capabilities. The company said he will help build a new team that uses Claude to speed up pre-training research. Karpathy said in a post on X that he was “very excited” to join the company and “get back to R&D.” ### Who is Andrej Karpathy, and why does this hire draw attention? Andrej Karpathy co-founded OpenAI and later led AI work at Tesla, where he became one of the most visible engineers in the field. He has also built a large following through technical talks, educational material and open-source work on neural networks and language models. (techcrunch.com) Karpathy wrote on X that “the next few years at the frontier of LLMs will be especially formative,” and said he planned to resume his education work later. Anthropic did not announce a formal executive title for him, but outside reports said he joined the pre-training organization rather than a product unit. (techcrunch.com) ### What exactly will he do inside Anthropic? Anthropic said Karpathy is joining pre-training, the part of the company that runs the large-scale training jobs behind Claude. Forbes, citing the company, said he will work under team lead Nick Joseph and help launch a group focused on using Claude to develop model-building tools and research workflows, rather than relying only on additional raw compute. (tech.yahoo.com) TechCrunch reported that Anthropic described pre-training as the stage that gives Claude its core knowledge and capabilities. That places Karpathy in one of the most expensive and technically central parts of the lab’s work, where progress depends on data, infrastructure, evaluation and training methodology. (forbes.com) ### Why does Anthropic pair this with Claude-based research tooling? Anthropic has been publishing engineering work that treats Claude not only as a user-facing assistant but also as a system for helping with complex technical workflows. In recent months, the company has described internal work on long-running coding agents and “context engineering,” both aimed at improving how models handle multi-step tasks and development environments. (techcrunch.com) The company’s statement, as reported by multiple outlets, said Karpathy will help build a team using Claude to accelerate pre-training research. Anthropic did not publicly detail the tools involved, but the description points to using models to assist with experiment design, analysis or supporting software around training runs. That is an inference from the company’s wording, not a separately announced product. (anthropic.com) ### How does this fit with Anthropic’s broader hiring and product push? Anthropic’s public materials show the company hiring across research, compute, infrastructure, safeguards and enterprise product roles. Its careers pages currently list openings in AI research and engineering, compute, infrastructure software, security and product groups. (thenextweb.com) Anthropic has also recently promoted enterprise products around Claude, including managed agents, security features and compliance materials in its transparency and trust pages. Karpathy’s move adds to that expansion by placing a high-profile researcher in the model-development stack rather than in sales or deployment functions. (anthropic.com) ### What happens next? Karpathy said on May 19 that he had already joined and started work. Anthropic has not announced a separate research paper, product release or event tied to his arrival, so the next public signs are likely to come through future company research posts, model updates or hiring activity around the pre-training team led by Nick Joseph. (anthropic.com) (tech.yahoo.com)