Andrej Karpathy, a well-known researcher and original co-founder of OpenAI, has officially taken a new job at Anthropic. Known for his past work running computer vision at Tesla and building the learning startup Eureka Labs, Karpathy shared the news online this Tuesday. He will step into a research and development position focused on large language models as Anthropic grows its team. He believes the next few years in this space will bring massive changes.
Karpathy leads a new research group using Claude for pre-training
At Anthropic, Karpathy joins the pre-training team under Nick Joseph. Pre-training is the heavy lifting phase where models get their basic knowledge before any fine-tuning happens. Anthropic hopes that having top talent here will help it build stronger systems.
He will create and guide a specific group tasked with using the Claude model to speed up pre-training research itself. Instead of just throwing more computing power at the problem, Anthropic wants to use AI to help make its own models better. Joseph publicly welcomed Karpathy, noting he is perfectly suited for this exact challenge.
The researcher pauses his educational startup to focus on models
Karpathy is highly regarded for his ability to explain hard technical topics to the public. Recently, he founded Eureka Labs, a startup designed to rethink schooling with digital assistants. This move to Anthropic means Eureka Labs will take a back seat for now.
However, Karpathy made it clear that he still cares deeply about teaching. He plans to return to his learning projects in the future when the time is right. For now, he is fully focused on his new company and the work ahead.
Adding a recognized name like Karpathy gives Anthropic another major win. As the company competes directly with OpenAI and Google, bringing in a co-founder from its biggest rival shows how seriously it views the race to improve artificial intelligence.
