Anthropic has attracted co‑founder of OpenAI—Andrey Karpaty will work on training the Claude model
Andrey Karpathy – the new “pilot” at Anthropic
Who is he?
AI researcher, co‑founder and former employee of OpenAI. He previously headed Tesla’s AI department.
New role
He joined Anthropic as head of Claude pre‑training work—an essential stage that provides the system with foundational knowledge and skills. This is one of the most expensive and resource‑intensive processes in building modern LLMs.
Mission
Build a team that will use Claude to accelerate pre‑training research. Karpathy is considered one of the few specialists who can bridge large language model theory with large‑scale training practice. His appointment underscores Anthropic’s priority: “research, not just compute power” – the key to competitiveness in AI.
Karpathy’s Path
Stage | Activity
OpenAI (until 2017) | Worked on deep learning and computer vision.
Tesla (until 2022) | Led full‑self‑driving (FSD) and autopilot programs.
Return to OpenAI | Spent a year before leaving again.
Eureka Labs (2024–?) | Founded a startup applying AI assistants in education. Details about his role are not yet disclosed; he may continue this project later.
Educational projects | Taught the online course “Neural Networks: From Zero to Hero” and runs a YouTube channel with lectures on master’s programs and AI.
> “I joined Anthropic,” Karpathy wrote on X. “The next few years at the forefront of LLMs will be especially important. I’m very excited to return to research and development,” he added, emphasizing his enduring passion for education and intention to resume work in this area later.
Cybersecurity at Anthropic
Anthropic also expanded its Red Team by bringing on cybersecurity veteran Chris Rohlf:
| Information | Details | Experience |
|---|---|---|
| Experience | Over 20 years in cybersecurity. | |
| Recent positions | Six years at Meta, previously a research fellow at the Center for Security and Emerging Technologies at Georgetown University (CyberAI project). | |
| Mission | Test complex AI models for threats, improving cyber defense. |
In short:
Andrey Karpathy is now working on foundational pre‑training research at Anthropic, while Chris Rohlf strengthens the Red Team to secure AI models. Both bring rich experience from OpenAI, Tesla and Meta, affirming Anthropic’s strategic focus on innovation and data protection.
Comments (0)
Share your thoughts — please be polite and stay on topic.
Log in to comment