Who is Andrej Karpathy: From Tesla through OpenAI to Anthropic

Andrej Karpathy is among the absolute elite of artificial intelligence researchers in the world. He was born in Bratislava and grew up in Toronto, Canada, where he later graduated from the University of Toronto. He earned his PhD at Stanford University under the supervision of Professor Fei-Fei Li, one of the most influential figures in computer vision.

In 2015, he became one of the founding researchers of OpenAI, where he specialized in deep learning and computer vision. In 2017, Elon Musk personally recruited him to Tesla, where Karpathy led the development of Autopilot and Full Self-Driving (FSD) systems — key technologies for autonomous driving. Musk at the time called him on Twitter "one of the world's best experts in computer vision and deep learning."

Karpathy returned to OpenAI in 2023 — at a time when the company was experiencing rocket growth thanks to ChatGPT. After a year, however, he left again and in 2024 founded the startup Eureka Labs, focused on using AI assistants in education. On his YouTube channel, he regularly published lectures on large language models and how they work. He is also the author of the now-legendary course Neural Networks: Zero to Hero, which has taught thousands of students how to build neural networks from scratch.

What Karpathy will do at Anthropic

According to TechCrunch, Karpathy started at Anthropic on Monday, May 19. He works on the pre-training team under Nick Joseph's leadership, and his main task will be to build a new team focused on using Claude to accelerate pre-training research.

For laypeople: pre-training is the phase where the model learns from a massive amount of text data to acquire basic knowledge about language, facts, and logic. It is the most expensive and computationally demanding part of developing a language model — a single training run can cost tens of millions of dollars. It is precisely the quality of pre-training that determines whether the resulting model will be smart or merely average.

With this, Anthropic is sending a clear signal: it doesn't want to compete with OpenAI just through sheer computational power, but through a smarter approach to training. Karpathy is one of the few people in the world who understands both the theory of large language models and the practical side of their massive training.

What this means for the competitive battle in AI

Karpathy's move comes at a time when Anthropic is experiencing its historically strongest period. According to CNBC, the company is heading toward $10.9 billion in revenue in the second quarter of 2026 and is expected to become profitable for the first time. Just a few weeks before Karpathy's announcement, Anthropic introduced the Claude Opus 4.7 model, which outperforms OpenAI's GPT-4.5 in benchmarks. The company also launched the highly guarded Mythos model, which according to The New York Times set off alarms at central banks and intelligence agencies around the world.

For comparison: OpenAI dominates the market with ChatGPT, which in May 2026 surpassed a 50% share of the South Korean search market — the first time in history an AI assistant has overtaken traditional search engines in any country. Google with Gemini remains a strong player, and Meta is building its own AI assistant for billions of users. On the horizon is also the anticipated GPT-6, which OpenAI plans to release in the coming months.

Karpathy's arrival at Anthropic thus comes at an exceptionally competitively charged time. The fact that Anthropic managed to win over a co-founder of OpenAI — someone who was there at the founding of the company it now competes against — is a significant symbolic and practical victory.

Slovak roots and Czech relevance

For Czech and Slovak readers, Karpathy is interesting on a personal level: he was born in Bratislava and is proud of his Slovak heritage. In interviews, he has repeatedly mentioned that his family emigrated to Canada when he was 15 years old. He is one of the few European natives to have reached the absolute top of American AI research.

For Czech users, it's significant that Claude from Anthropic is fully available in Czech — both in the free version and in the paid Claude Max plan for approximately $100 and $200 per month. Claude performs very well in Czech, including understanding specific Czech cultural context and grammar. In user tests on Czech forums, it often comes out comparable to GPT-5, and in some tasks even more accurate.

Karpathy's educational activities — especially the Zero to Hero course and his YouTube lectures — are freely available to Czech students and developers as well. They represent one of the highest quality free resources for understanding how neural networks and language models work under the hood.

Karpathy is not alone: Anthropic is strengthening across disciplines

Simultaneously with Karpathy, Anthropic also announced the arrival of Chris Rohlf, a cybersecurity veteran with more than twenty years of experience. Rohlf previously worked on Yahoo's security team (known as "The Paranoids") and spent the last six years at Meta. At Anthropic, he joined the frontier red team — a special unit that tests advanced AI models for their ability to withstand serious cyber threats.

"We have a real opportunity ahead of us to dramatically improve cybersecurity using AI," Rohlf wrote on X. "I can't imagine a better company or team to join at this critical moment."

These two hires show that Anthropic is systematically strengthening both its research excellence (Karpathy) and security infrastructure (Rohlf). This is no coincidence — the company, founded by former OpenAI researchers precisely because of concerns about AI safety, continues to build on its original DNA.

What will happen to Eureka Labs?

The fate of Karpathy's educational startup Eureka Labs remains unclear after his move to Anthropic. Karpathy himself wrote on X: "I remain deeply passionate about education and plan to return to it in time." This suggests that Eureka Labs is currently on the back burner — but not dead.

Karpathy's announcement about joining Anthropic was typically brief and matter-of-fact: "I think the next few years at the frontier of large language models will be particularly formative. I'm excited to join this team and return to research and development."