Andrej Karpathy has joined Anthropic, the AI researcher announced on Tuesday. The move brings him back to lab work after more than a year away.
The hire puts impressive figures on AI research in direct competition with OpenAI. Karpathy was a founding member there before holding two shares in the company.
From founding member of OpenAI to Anthropic Researcher
Karpathy joined the original OpenAI team in 2015. He left in 2017 to Tesla and rejoined OpenAI in 2023 for about a year. He left again in February 2024.
Karpathy planned the exit as a personal decision and not because of any internal conflict. He moved into freelance work soon after.
He founded Eureka Labs, an AI education startup. He also built a large following through his Zero to Hero video on neural networks and linguistics.
Anthropic’s announcement marks his first return to a large laboratory since his departure. He explained the directions X.
I have joined Anthropic. “I think that the next few years on the LLM side will be very progressive… I am very passionate about education and I plan to resume my work in due course,” he said. he shared in the post.
Follow us on Xto hear the latest news as it happens
Coining Vibe Coding Between Stints
Karpathy coined a term that also changed the way developers talk about AI-powered software. In a February 2025 post, he introduced the term vibe coding.
It is defined as the process of running a task in which the model writes the code. The user accepts the changes without reading the diff. The word entered the mainstream within weeks.
The spread of BeInCrypto has registered its operation in build a crypto trading bot in one weekend. Similar work helped to separate Winning crypto exchanges from laggards.
Also lowered the barrier of entry for Web3 developers without programming background.
Karpathy has since refined the idea into what he calls technical engineering. People are focused on monitoring and monitoring while independent agents are in charge of execution.
Frontier LLM Search for the Next Topic
Anthropic was founded in 2021 by former OpenAI researchers. The company has positioned itself as a more security-focused solution than its biggest rival.
Claude’s model family directly competes with GPT. Anthropic posted Opus 4.7 in the month of April with strong long-range thoughts and visions.
Karpathy’s hiring follows the steady flow between Anthropic and OpenAI in recent years. OpenAI counted the days after Opus 4.7 release is GPT-5.5.
That version was positioned as the best version of OpenAI for autonomous, multitasking. The battle for talent between the two labs has intensified.
Neither Karpathy nor Anthropic have revealed the team or the projects they will be working on. His public comments refer to LLM courses and agency practices.
His previous research in neural network design, computer vision, and artificial intelligence has direct applications in those areas.
Karpathy also said that he wants to resume his academic career in due course. Eureka Labs and related products will continue along with its Anthropic division.
The coming months will reveal the areas of Anthropic Karpathy research. Its arrival could also change the way the company offers its frontier services to manufacturers.
A note OpenAI Worker Who Created Vibe Coding Joins Anthropic appeared for the first time BeInCrypto.
