Andrej Karpathy, one of the founding partners of OpenAI and formerly the artificial intelligence leader at Tesla, announced that he has joined Anthropic. In his statement on the X platform, Karpathy stated that he is excited about the future of large language models and is pleased to return to research and development processes.

Starting this week, Karpathy started working in the pre-training team at Anthropic, led by Nick Joseph. The pre-training process, which provides the basic capabilities and knowledge of the company’s model named Claude, is known as one of the most costly and high-process power-requiring parts of the model development steps.

A new era in artificial intelligence research

Anthropic officials announced that Karpathy will establish a special group that will accelerate pre-training research using the Claude model. The fact that Karpathy is one of the rare experts who can bridge the gap between the theory of large language models and large-scale educational applications is considered a critical step for Anthropic’s strategy.

This move shows that the company aims to gain a competitive advantage over rivals such as OpenAI and Google with artificial intelligence-supported research, rather than focusing only on pure process power.

Karpathy, who focused on deep learning and computer vision at OpenAI in his career, moved to Tesla in 2017 and managed the company’s Full Self-Driving and Autopilot programs.

Karpathy, who worked at OpenAI for another year after leaving Tesla in 2022, started an initiative called Eureka Labs in 2024 with the aim of developing artificial intelligence assistants in the field of education.

Karpathy stated that his passion for education continues and that he plans to return to his work in this field in the future.

New reinforcement to the security team

In addition to Karpathy, Anthropic also added an experienced name in the field of cyber security to its team. Chris Rohlf, who previously worked at companies such as Yahoo and Meta, joined Anthropic’s border security group, which tests advanced artificial intelligence models against serious threats.

Rohlf, who has more than twenty years of cyber security experience, stated that artificial intelligence technologies offer a valuable opportunity to improve cyber security.

These transfers of Anthropic reveal the value the company attaches to both model development and security processes.

How do you evaluate Andrej Karpathy’s new mission within Anthropic and the strategy the company follows?