Karpathy Moves to Anthropic for Frontier LLM Work

*Andrej Karpathy announced he has joined Anthropic to focus on core research after leaving his AI education startup.*

Andrej Karpathy stated on X that he has joined Anthropic. He expects the next several years at the leading edge of large language model development to shape the field in important ways and wants to return to hands-on research.

Karpathy’s path through major AI organizations is well documented. He helped start OpenAI in 2015. He then served as director of AI at Tesla from 2017 to 2022. He returned to OpenAI in 2023 before departing again in 2024 to found Eureka Labs, an AI-focused education company. In February 2025 he introduced the phrase “vibe coding” to describe a particular style of model-assisted development.

The move places a prominent researcher inside one of the companies most directly competing on frontier model capabilities. Anthropic has positioned itself around safety-conscious scaling, while Karpathy has previously worked across both research labs and applied deployment at Tesla. His stated priority is R&D rather than education for the immediate term, though he noted he remains committed to teaching and intends to pick that work up again later.

No public statements from Anthropic accompanied the announcement. Coverage on Hacker News and technology sites has so far centered on Karpathy’s track record rather than any specific project he will lead.

Implications for the field

Karpathy’s decision underscores how concentrated talent remains around a handful of organizations pushing the largest models. His experience spans the earliest days of OpenAI, automotive autonomy, and independent education efforts. Placing that background inside Anthropic adds senior research capacity at a moment when several labs are racing to extend context lengths, improve reasoning reliability, and reduce training costs.

Observers have noted that researchers with his profile rarely stay independent for long. The pattern suggests that meaningful progress on next-generation systems still requires the compute, data, and engineering infrastructure only the largest labs can assemble. Karpathy’s choice aligns with that reality while leaving room for future education projects once the current research cycle settles.

The announcement itself was brief and personal. It contained no technical details or timelines. Readers following the public record will watch for any later papers, talks, or internal updates that clarify the exact scope of his role.

---

Sources:

• Hacker News (front page)
• Daring Fireball

{
  "excerpt": "Andrej Karpathy has joined Anthropic to return to frontier LLM research after founding an AI education startup.",
  "suggestedSection": "ai",
  "suggestedTags": ["andrej-karpathy", "anthropic"],
  "imagePrompt": "An abstract composition of layered translucent neural network diagrams floating above a dark wooden workbench scattered with precision tools and open notebooks. Soft directional light catches the edges of the diagrams and instruments without revealing any faces or text, muted color palette, cinematic lighting, 16:9."
}