Andrej Karpathy, one of the original co-founders of OpenAI and a highly respected figure in artificial intelligence research, announced on Monday that he has joined Anthropic, the company behind the Claude family of large language models. The move represents a significant talent acquisition for Anthropic as it competes to stay at the forefront of AI development.

Karpathy will work on Anthropic's pre-training team, which is led by Nick Joseph. In his new role, Karpathy will build a new group focused on a recursive and ambitious goal: using Claude itself to accelerate pre-training research. Pre-training is the most expensive and computationally intensive phase of building frontier AI models, where the model learns core knowledge and capabilities from vast datasets. Finding efficiencies in this process could reshape the economics of the AI industry.

In a post on X that garnered over 13 million views, Karpathy expressed his belief that the next few years at the frontier of large language models will be especially formative. He also noted that he remains deeply passionate about education and plans to resume that work in due time. This move puts Karpathy back at the heart of cutting-edge AI research after a brief hiatus with his education startup.

Karpathy's Career Arc

Karpathy's career has touched nearly every major milestone in modern AI. He earned his PhD at Stanford under Fei-Fei Li, the computer scientist behind the ImageNet dataset, focusing on deep learning and computer vision. He was among the 11 people who co-founded OpenAI in 2015, where he worked on deep learning research before leaving in 2017 to join Tesla as Director of AI.

At Tesla, Karpathy led the computer vision teams behind Full Self-Driving and Autopilot, critical programs for the electric carmaker's autonomous vehicle ambitions. He left Tesla in July 2022, returned to OpenAI for about a year, and then departed again in 2024 to found Eureka Labs, a startup dedicated to applying AI assistants to education. With his move to Anthropic, Eureka Labs' work is now on hold.

Timing and Industry Context

The timing of Karpathy's hire is noteworthy. Anthropic has become a magnet for top-tier technical talent, especially as its chief rival, OpenAI, has experienced a series of high-profile departures. Over the past two years, OpenAI has lost more than a dozen senior executives and researchers, including CTO Mira Murati, reinforcement learning pioneer John Schulman, and three executives who left on a single day in April 2026. These departures have raised questions about OpenAI's direction and culture.

For Anthropic, landing Karpathy signals the company's ability to attract talent of the highest calibre as it scales both its research and commercial operations. The firm, led by CEO Dario Amodei, has drawn investor interest at a valuation of roughly $800 billion and is reportedly exploring an IPO that could come as early as late 2026.

Karpathy's new role also underscores a broader trend in frontier AI: the use of existing models to improve the next generation. If Claude can meaningfully speed up its own pre-training pipeline, it would mark a practical demonstration of recursive self-improvement, a capability that the AI safety community has long watched closely. Whether this prospect excites or unnerves observers may depend on how much trust they place in the safety-minded culture Anthropic has cultivated since its founding.

Andrej Karpathy's career began with a fascination for neural networks. After completing his PhD at Stanford, where he worked on convolutional neural networks and reinforcement learning, he joined OpenAI as one of its first researchers. His contributions there included work on unsupervised learning and generative models. At Tesla, he built the team that developed the perception systems for Autopilot, using large-scale neural networks to process camera and sensor data. His return to OpenAI was brief but included work on multimodal models and alignment research.

Eureka Labs, his most recent venture, aimed to create AI-powered tutors for students. The startup's pause suggests that Karpathy sees immediate opportunities at Anthropic that are more impactful than educational tools. His deep technical expertise in pre-training and model optimization will be directly applied to improve Claude's capabilities.

Anthropic's pre-training team focuses on the foundational stage of model development, which involves training on massive datasets using hundreds or thousands of GPUs. This stage determines the model's language understanding, reasoning ability, and factual knowledge. By using Claude itself to generate synthetic data or optimize training hyperparameters, the team hopes to reduce the time and cost required for future versions of the model. This approach is part of a larger industry trend toward self-play and self-improvement in AI systems.

The potential impact of Karpathy's work extends beyond Anthropic. If his team succeeds, it could lower the barrier for developing frontier models, allowing more players to compete with current leaders. It could also accelerate the pace of AI research, as faster pre-training enables more experiments and iteration. However, it also raises concerns about the alignment and safety of models that improve themselves recursively. Anthropic's emphasis on safety and interpretability may help mitigate these risks, but the field remains divided on how to handle such systems.

Karpathy's decision to join Anthropic rather than return to OpenAI or start another venture reflects his confidence in the company's mission and technical strategy. Anthropic's commitment to constitutional AI, which guides models to be helpful, harmless, and honest, aligns with Karpathy's own expressed values. He has often spoken about the importance of thoughtful AI development and the need for transparency in model capabilities.

The news has already sparked discussions among AI researchers and enthusiasts. Many see it as a validation of Anthropic's approach and a sign that the company is becoming the preferred destination for top talent. Others note that Karpathy's move could influence other researchers who are considering leaving OpenAI. The AI industry is watching closely to see how this hire affects the competitive dynamics between the two companies.

In the coming months, Karpathy will build his team and begin work on recursive pre-training methods. Anthropic has not disclosed specific milestones or goals, but the company's track record suggests that significant advances are expected. As Karpathy himself said, the next few years at the frontier of LLMs will be especially formative, and he is now in a position to help shape them.

Source: TNW | Anthropic News