Karpathy joins Anthropic to lead pre-training push

Andrej Karpathy, the co-founder of OpenAI and former head of AI at Tesla, has joined Anthropic to lead its pre-training efforts. Karpathy confirmed the appointment on social media platform X on Tuesday, stating he is excited to return to research and development at the frontier of large language models. He will lead a new team under Nick Joseph, focusing on using the Claude model to accelerate pre-training research.

Pre-training represents the most compute-intensive phase of building a frontier model, responsible for establishing the core knowledge and capabilities of the system. According to an Anthropic spokesperson, Karpathy will spearhead a team dedicated to leveraging Claude to streamline this expensive process. The company views this move as a strategic emphasis on AI-assisted research as a method to maintain competitiveness against rivals such as OpenAI and Google.

Karpathy brings a rare combination of theoretical knowledge and practical experience in large-scale training. He previously co-founded OpenAI, where he worked until 2017, before leading Tesla’s Full Self-Driving and Autopilot programs from 2017 to 2022. He returned to OpenAI for a one-year stint between 2022 and 2023, after which he founded Eureka Labs, a startup focused on applying AI assistants to education.

The appointment raises questions regarding the future of Eureka Labs. Karpathy has not provided recent updates on the startup, and it remains unclear whether he will continue his involvement given his new responsibilities at Anthropic. He noted on X that he remains passionate about education and intends to resume related work in the future, though no specific timeline was provided.

In a separate development, Anthropic has hired Chris Rohlf for its frontier red team, which stress-tests advanced AI models against severe threats. Rohlf brings over 20 years of cybersecurity experience, including roles at Yahoo and Meta, and previously worked on the CyberAI project at Georgetown University. He stated that he sees a significant opportunity to improve cybersecurity with AI and described Anthropic as the ideal team to join at this critical moment.

Karpathy is also widely recognised for his educational contributions, including the online course Neural Networks: Zero to Hero and his YouTube lectures on artificial intelligence. His transition to Anthropic marks a significant consolidation of talent in the sector, as leading figures from major tech and AI firms converge on specific strategic priorities in model development and safety.