Andrej Karpathy is back in the trenches. The prominent AI researcher, known for his stints at OpenAI and Tesla, has joined Anthropic as a hands-on lab contributor. The move ends a period of independent projects and signals his return to the front lines of frontier model development.
From OpenAI to Anthropic
Karpathy was a founding member of OpenAI in 2015. He left in 2017 for Tesla, where he led Autopilot vision, then rejoined OpenAI in 2023 for about a year before departing again in February 2024. That second exit kicked off a stretch of solo work that included launching Eureka Labs, an AI education startup, and building a loyal following through his 'Zero to Hero' video series on neural networks and language models.
Now he's at Anthropic, the company founded in 2021 by former OpenAI researchers. Anthropic has positioned itself as a safety-focused alternative to OpenAI, and its Claude model family now competes directly with OpenAI's GPT line. The company shipped Opus 4.7 in April, which brought stronger long-form reasoning and vision capabilities. OpenAI quickly countered with GPT-5.5, pitched as its most capable system for autonomous, multi-step work.
Vibe Coding to Agentic Engineering
Karpathy coined the term 'vibe coding' in February 2025. He used it to describe a workflow where the model writes code and the user accepts changes without reading the diff. The phrase caught on, reflecting a growing reliance on AI-assisted development. But Karpathy didn't stop there. He later refined the concept into 'agentic engineering', where humans focus on high-level specifications while autonomous agents handle execution. It's a shift from passive acceptance to active delegation.
At Anthropic, he'll be working on exactly that kind of frontier — systems that can reason, execute multi-step tasks, and operate with greater autonomy. The company's safety focus aligns with his interest in building AI that can be trusted to act on behalf of users.
Education Work Continues
Karpathy's move doesn't mean he's leaving teaching behind. He plans to resume his education work alongside his Anthropic role. Eureka Labs, his AI education startup, remains active. The 'Zero to Hero' series, which taught thousands how to build neural networks from scratch, will continue to get new material.
That dual track — building cutting-edge models while explaining them to a broader audience — has defined much of his career. It's a rare combination. Most researchers either ship code or teach. Karpathy does both, and he's betting he can keep doing it from inside one of the most closely watched labs in the field.
What that means for Anthropic's competition with OpenAI is the open question. Karpathy knows the GPT family from the inside. He knows how they think about safety, scale, and deployment. Now he's on the other side, building Claude.