IMPLICATOR
Nicholas Joseph, Anthropic's head of pretraining, gave Andrej Karpathy his assignment in an X post Tuesday. Karpathy will join the pretraining team and build a group using Claude to accelerate pretraining research itself, Joseph wrote. Karpathy confirmed the move in a post of his own: "I've joined Anthropic," he wrote, adding that "the next few years at the frontier of LLMs will be especially formative."
The significance is the assignment, not the celebrity hire: Anthropic is putting a researcher with OpenAI, Tesla and synthetic-data experience inside the process that produces Claude's base capabilities. The company says Claude Code reached more than $2.5 billion in run-rate revenue by its February funding announcement after becoming generally available in May 2025. Anthropic said in the same post that company run-rate revenue had reached $14 billion, up from about $1 billion at the start of 2025; the New York Times later reported that Amodei had put the figure at $30 billion.
Key Takeaways
- Karpathy joins Anthropic's pretraining team under Nick Joseph.
- Joseph says Karpathy will build a team using Claude to accelerate pretraining research.
- Anthropic links the hire to a business already scaling around Claude Code.
- The move shifts Karpathy's public education work inside a proprietary frontier lab.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
The pretraining assignment
TechCrunch reported Tuesday that Karpathy started this week under Joseph on the Anthropic group responsible for the large-scale runs that give Claude its core knowledge and capabilities. The same report described pretraining as one of the most expensive, compute-intensive phases of building a frontier model. Joseph's post gave the internal version: "He'll be building a team focused on using Claude to accelerate pretraining research itself."
Anthropic is hiring an OpenAI alumnus back into frontier-lab work, but the role is more specific than the name. In February, Anthropic said it had raised $30 billion at a $380 billion post-money valuation. Last week, the New York Times reported that the company was in talks to raise $30 billion to $50 billion at up to $950 billion, above OpenAI's reported $852 billion March valuation.
The proposed team is Anthropic's test of whether research labor, not only more compute, can shorten the path to the next model.
The résumé matches the machine
Karpathy's career is unusually close to the problem Anthropic just assigned him. VentureBeat, citing his website, reported that he helped create Stanford's CS231n course, worked at OpenAI, led Tesla's Autopilot computer-vision team from 2017 to 2022, then returned to OpenAI from 2023 to 2024 to build a group focused on midtraining and synthetic data generation.
That last line is not biography filler. It is exactly the connection between model theory and model operations that TechCrunch identified when it wrote that Karpathy is one of the few researchers who can move between LLM theory and large-scale training practice. Pretraining sets the base model. Midtraining and synthetic data can shape downstream behavior and reasoning style, while products like Claude Code also depend on post-training and the surrounding agent harness.
Track the AI talent wars
Strategic AI news from San Francisco. No hype, no "AI will change everything" throat clearing. Just what moved, who won, and why it matters. Daily at 6am PST.
No spam. Unsubscribe anytime.
Karpathy has spent the past year teaching that shift in public. The Implicator covered his February claim that AI coding agents made programming "unrecognizable" since December. Business Insider also quoted his label for the shift: "Many people have tried to come up with a better name for this to differentiate it from vibe coding, personally my current favorite 'agentic engineering.'"
Small detail, large signal: VentureBeat also noted that Karpathy interned at Google Brain, Google Research and DeepMind before this whole cycle began.
The talent war moved to training
Reuters put Karpathy's move next to another one: John Schulman, another OpenAI co-founder, left for Anthropic in 2024, while former OpenAI chief scientist Ilya Sutskever and former technology chief Mira Murati also left the ChatGPT maker. The Decoder called Karpathy's choice of Anthropic over a return to OpenAI "a clear loss for his former employer."
The loss is not only symbolic. Karpathy had been one of AI's most useful public explainers while building Eureka Labs, the education startup he announced in 2024. On Tuesday he said, "I remain deeply passionate about education and plan to resume my work on it in time." VentureBeat read that as a sign that at least some education work will pause while he digs into Anthropic.
VentureBeat put the open-work tension in practical terms: Anthropic has supported Model Context Protocol, and it still ships proprietary models and products such as Claude and Claude Code.
Karpathy helped popularize workflows that spread because he explained them in public. Inside Anthropic, that habit now has a narrower assignment from Joseph: use Claude to accelerate pretraining research. The next model release or funding announcement will show how much of that workflow Anthropic is ready to put in writing.
Frequently Asked Questions
What role is Andrej Karpathy taking at Anthropic?
Karpathy is joining Anthropic's pretraining team under Nick Joseph. Joseph said Karpathy will build a team focused on using Claude to accelerate pretraining research itself.
Why does the hire matter for Anthropic?
Pretraining is one of the most expensive parts of building frontier models. Anthropic is putting Karpathy's OpenAI, Tesla and synthetic-data experience inside the process that produces Claude's base capabilities.
What did Karpathy say about the move?
Karpathy wrote on X that he had joined Anthropic and that the next few years at the frontier of LLMs would be especially formative. He also said he remains passionate about education and plans to resume that work in time.
How does this connect to Claude Code?
Anthropic says Claude Code had reached more than $2.5 billion in run-rate revenue by its February funding announcement. A stronger Claude model directly affects the coding and agent products driving that growth story.
Is Karpathy leaving his education work behind?
He did not say he is abandoning it. He wrote that he remains deeply passionate about education and plans to resume that work in time, which suggests at least some of it pauses while he works at Anthropic.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: [email protected]
