Andrej Karpathy, the AI researcher who helped found OpenAI and later led Tesla’s autonomous‑driving work, quietly started at Anthropic this week, shifting back into full‑time research and development. His move hands the San Francisco lab one of the industry’s most recognizable technical leaders as it scales up its Claude models. The hire underscores how the scramble for frontier AI talent is now playing out in downtown San Francisco office towers.
Karpathy announced the move on X with a simple line, “I’ve joined Anthropic,” and an Anthropic spokesperson confirmed he will join the company’s pretraining group under team lead Nick Joseph. As reported by TechCrunch, he started this week and will set up a team focused on using Claude to accelerate pretraining research.
Karpathy was an early OpenAI staffer before leaving in 2017 to run Tesla’s AI efforts, where he led Autopilot and Full Self‑Driving engineering until departing the company in 2022. He briefly returned to OpenAI afterward and in 2024 launched an education startup called Eureka Labs while continuing to publish technical explainers for a broad audience. Those career milestones are recorded in reporting by Reuters.
Pretraining And Compute: Why Karpathy Matters
Pretraining is the compute‑heavy phase that gives a model its foundational knowledge, and practical experience running massive training jobs is rare. Anthropic has been locking up capacity and partnerships to remove that bottleneck, including an arrangement to take over the compute at SpaceX/xAI’s Colossus 1 cluster, which means Karpathy’s hands‑on expertise could pay immediate dividends. TechCrunch has covered the Colossus arrangement and why those compute deals matter for labs racing to scale their models.
Anthropic’s Broader Momentum
The hire lands amid a run of enterprise partnerships and fundraising activity that has pushed private‑market estimates of Anthropic’s worth into the high hundreds of billions. Bloomberg reported the company has been weighing funding offers that could value it above $900 billion, a sign of the heavy institutional interest behind the Claude maker. That mix of capital, compute and senior hires helps explain why top researchers are landing at Anthropic.
San Francisco’s AI Boom Keeps Hiring
Locally, Anthropic’s growth is visible in a string of office deals. The company has snapped up large blocks at Foundry Square, including commitments at 300 and 400 Howard Street, as it builds out a downtown campus to house more engineers. The Real Deal has reported on the Howard Street leases and notes Anthropic’s expanding footprint is reshaping the Financial District office market.
Karpathy stressed both research and education in his public posts, writing, “I remain deeply passionate about education,” a line Reuters highlighted in its coverage of the announcement. For Anthropic, the hire adds both prestige and practical training know‑how just as fresh compute capacity comes online, giving the company another senior engineer who knows how to turn large GPU allocations into working systems.
