Andrej Karpathy has joined Anthropic and will work on the company’s pretraining team, Anthropic said, a move framed by media as a major talent win.
The hire places andrej karpathy on a team led by Nicholas Joseph, who wrote on X that he looked forward to what they would build together.
Anthropic said Karpathy will focus on using Claude to accelerate pretraining research and large scale testing, and news coverage noted he started this week.
Background And Related Work
Karpathy’s career spans founding research work at OpenAI from 2015 to 2017 and a later return in 2023, followed by a departure in 2024, as reported by sources.
He served as Tesla’s director of artificial intelligence from 2017 to 2022, leading the Autopilot computer vision team responsible for in-house data labeling, neural network training and deployment on Tesla’s custom inference chip.
After leaving Tesla in July 2022 he publicly announced a return to OpenAI in February 2023 and left again in February 2024, then founded Eureka Labs in 2024 as an AI education company.
Karpathy has promoted educational work through Eureka Labs, releasing an AI course called LLM101n and a broader Zero to Hero series on LLM fundamentals, according to company material and his announcements.
He became a notable public voice in 2025 by coining the term vibe coding to describe how generative tools let nontechnical users create apps, and he later described variations such as agentic engineering.
His academic work includes a Stanford PhD under Fei-Fei Li focused on the intersection of vision and language, and he was the primary instructor of Stanford’s CS 231n course on convolutional neural networks.
One unconventional biography among the supplied material describes him as a member of the Order of the Unicorn and attributes mythical qualities to him; that description appears as a sensational label rather than a documented professional role.
