Andrej Karpathy Joins Anthropic to Lead Pre-Training Acceleration

Andrej Karpathy, an OpenAI founding member and one of the most-followed AI researchers in the world, has joined Anthropic. The move, announced Monday on Karpathy's own X account, lands him on Anthropic's pre-training team under team lead Nick Joseph — with a specific mandate to use Claude to speed up Anthropic's own pre-training research.

The hire is a notable signal in a year that has seen Anthropic, OpenAI, and Google trade senior researchers at a pace that resembles an NBA free agency. Karpathy is the highest-profile mover so far.

Why This Hire Matters

Karpathy is not a typical senior research hire. He was a founding member of OpenAI in 2015, led Tesla's Full Self-Driving and Autopilot programs from 2017 to 2022, returned to OpenAI for a year, then left in 2024 to start Eureka Labs, an AI education startup.

His public output — the nanoGPT codebase, the "Let's build GPT" YouTube series, the LLM lectures at Stanford, and his widely shared writing on the state of model training — has shaped how a generation of researchers and engineers think about large language models. Hiring him is partly a research move and partly a talent-market signal.

For Anthropic, the message to the broader research community is straightforward: this is where frontier work is happening.

The Role: Claude Training Claude

Karpathy's stated focus at Anthropic is unusual. He's starting a team aimed at using Claude to accelerate pre-training research itself.

That phrasing matters. Anthropic and other frontier labs have publicly discussed using their own models for code review, experiment design, and ablation analysis inside research workflows. Karpathy's team appears to be a formal investment in pushing that pattern as far as it will go — having Claude play an active role in the design and execution of the next generation of Claude.

Why this matters: The hypothesis behind every frontier lab's compute spend is that models will eventually accelerate their own research. If that pattern works at scale, the lab whose model does the best job of accelerating its own research pulls ahead. Karpathy is being hired to test that hypothesis directly.

Karpathy's Statement

In his announcement post, Karpathy framed the move as a return to research after several years of teaching and entrepreneurship.

"Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D."

He added that he remains "deeply passionate about education and plan to resume my work on it in time" — a signal that Eureka Labs is paused, not shuttered.

The Broader Anthropic Story

The hire lands in a moment when Anthropic's growth numbers are arguably the most aggressive of any major AI lab.

The company is on track for its first profitable quarter, with revenue projected to more than double to $10.9 billion in Q2 2026 and an estimated $559 million operating profit — arriving roughly two years ahead of internal projections.

Anthropic also expanded its compute partnership with SpaceX this month, agreeing to spend roughly $1.25 billion per month through 2029 for access to the Colossus supercomputing infrastructure.

Against that backdrop, hiring Karpathy reads as a confidence move. Anthropic has the revenue, the compute, and now the headline research hire.

Industry Implications

For OpenAI, losing a founding member to the closest direct rival is not a small story. Sam Altman has historically been able to count on alumni goodwill even when researchers leave. Karpathy's choice of destination signals something about where serious researchers think the frontier is right now.

For Google DeepMind, the hire intensifies an existing talent competition. Demis Hassabis has been actively recruiting senior alumni from OpenAI and other labs, and Anthropic just landed a name that would have been near the top of any short list.

For the broader research community, Karpathy's choice will move careers. His move is the kind of signal that has historically caused other researchers and engineers to follow.

Expert Perspectives

Reaction on X focused on three things. Researchers noted the specific framing — "using Claude to accelerate pre-training" — as the most technically interesting part of the announcement. Industry watchers focused on the symbolism of an OpenAI co-founder choosing Anthropic. Educators wondered what happens to Eureka Labs.

Several investors framed the hire alongside Anthropic's $900 billion valuation and the SpaceX compute deal as evidence that the company is consolidating frontier talent ahead of a major model release.

What's Next

Karpathy started this week. Anthropic has not detailed the team's size, timeline, or specific research goals beyond the "Claude accelerating pre-training" framing.

The next public signal will likely come in Anthropic's next model release. If Karpathy's team is contributing to the pre-training of the next Claude generation, expect Anthropic to say so.

Bottom Line

Karpathy moving to Anthropic is the clearest individual talent signal in AI this year. It reinforces that Anthropic, not just OpenAI, is now seen as the place serious frontier researchers want to be — and it sets up Anthropic's next major model release as the test of whether Claude can actually help build the next Claude.