Andrej Karpathy, a name synonymous with the early days of modern artificial intelligence, confirmed on Monday that he has joined Anthropic’s pre‑training team. The former OpenAI co‑founder will head a brand‑new group tasked with using Anthropic’s own Claude model to accelerate the compute‑heavy pre‑training stage that underpins large language models.
Karpathy’s X post about the move drew more than 13 million views. In it he wrote that the next few years at the frontier of large‑language‑model (LLM) development will be especially formative, and that he remains deeply passionate about education – a cause he plans to revisit once his new role settles.
His career reads like a timeline of AI milestones. After earning a Ph.D. at Stanford under Fei‑Fei Li, Karpathy co‑founded OpenAI in 2015 and helped shape its early deep‑learning research. He left the organization in 2017 to become Tesla’s director of AI, where he led the computer‑vision teams behind Full Self‑Driving and Autopilot. He departed Tesla in July 2022, returned briefly to OpenAI, and in 2024 launched Eureka Labs, a startup that applied AI assistants to education. That venture is now on pause while Karpathy throws his weight behind Anthropic.
Anthropic, led by CEO Dario Amodei, has been on a talent‑acquisition spree as its chief rival, OpenAI, contended with a series of senior exits. Recent departures from OpenAI include CTO Mira Murati, reinforcement‑learning pioneer John Schulman, and three executives who left on a single day in April 2026. Karpathy’s arrival signals that Anthropic can lure top‑tier researchers even amid this turbulence.
The company’s valuation is reported at roughly $800 billion, and insiders say an initial public offering could materialize as early as late 2026. By building a team that leverages Claude to speed up its own pre‑training pipeline, Anthropic hopes to demonstrate a practical form of recursive self‑improvement – a concept that has long intrigued the AI‑safety community.
If successful, Claude could meaningfully reduce the time and cost required to train the next generation of models. That would reshape the economics of the AI industry, where pre‑training alone accounts for the bulk of compute spend. Anthropic’s safety‑first culture, cultivated since its founding, may also help assuage concerns about accelerating AI capabilities without adequate oversight.
For Karpathy, the move feels like a return to the lab. “I’m excited to be back building models at the frontier,” he wrote. The combination of his deep‑learning expertise, experience scaling vision systems for autonomous vehicles, and recent work in AI‑driven education positions him to push the boundaries of what Claude can achieve.
Industry analysts see the hire as a bellwether for the broader talent war between AI powerhouses. As Anthropic continues to attract high‑profile researchers, it may force rivals to double down on retention strategies and explore novel approaches to model efficiency.
Whether Anthropic’s Claude‑centric pre‑training strategy will deliver on its promise remains to be seen, but Karpathy’s involvement adds a layer of credibility to the company’s ambitious roadmap. For now, the AI community watches closely as one of the field’s most recognizable figures steps into a role that could redefine how large language models are built.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory SEO vous aide à automatiser le contenu d'actualités pour votre site.