On May 19, Andrej Karpathy — an OpenAI founding member, former Tesla AI chief, and founder of the education startup Eureka Labs — announced he'd joined Anthropic, the maker of Claude. He starts on the pre-training team, building a group to use Claude itself to accelerate AI research; Eureka Labs is paused, not closed. He's the third prominent OpenAI figure to land at Anthropic in under two years, after Jan Leike and John Schulman.

1. Anthropic Is Winning Because Mission Beats Money (SignalFire)

The best researchers keep choosing Anthropic over bigger checks elsewhere — and the traffic only goes one way.

Karpathy is the third marquee OpenAI departure to land at Anthropic, with no traffic the other way. He follows Jan Leike, OpenAI's former alignment head (May 2024), and co-founder John Schulman (August 2024) — a one-directional flow widely attributed to unease over OpenAI's commercialization and safety pace. Karpathy framed it simply: "the next few years at the frontier of LLMs will be especially formative," and he was "excited to join the team here and get back to R&D."

The money makes the loyalty more striking, not less. Per a SignalFire study, Anthropic retains roughly 80% of its two-year hires while paying below OpenAI at the median — reported figures put OpenAI senior engineers near $1.2 million in total comp versus about $600,000 at Anthropic. To this camp, that gap is the proof: top researchers are choosing mission and research freedom over the bigger check.

2. This Is the Talent Bubble at Fever Pitch (AI talent-war analysts)

When one hire draws millions of views and moves sector narratives, the market is pricing people like franchise athletes.

A single researcher's job change is now a market event. Karpathy's announcement drew nearly 3 million views within an hour, per TechCrunch, and landed in tech-stock roundups as if it moved valuations. Across the industry, labs are handing out nine-figure packages and poaching whole teams, and investor coverage treats individual AI researchers the way sports leagues treat free agents.

The counter sits in the same data. The anti-bubble read is that Anthropic's below-market pay plus high retention shows the top of the market is rational, not frothy — people are staying for reasons money can't fully explain. Either way, the spectacle around one hire is a tell about how concentrated, or how overheated, the AI race has become.

3. This Is Weird (Andrej Karpathy, Dario Amodei)

A man who says AGI is a decade away just took a job at the lab whose CEO thinks it's a year or two.

Karpathy is the industry's most prominent hype-skeptic. In an October 2025 interview he called today's AI agents overhyped "slop," argued AGI is "still a decade away," and reframed the "year of agents" as a "decade of agents," citing fundamental reliability gaps. Now his job at Anthropic is to use Claude to accelerate the pre-training research that builds the next models.

That puts him under a boss who sees the timeline very differently. Anthropic CEO Dario Amodei has written that powerful AI "could come as early as 2026," describing a coming "country of geniuses in a datacenter." So the move reads two ways: either a careful skeptic is exactly who you want building capability at the safety-first lab, or even the skeptics are now racing to push the frontier they warn about. Karpathy joining to make AI that speeds up AI is the tension in miniature.

Where This Lands

Anthropic keeps winning the people who matter, with mission seemingly outbidding money. It's also hard to argue with the idea that an industry treating single hires as multimillion-view events is either gloriously concentrated or dangerously frothy. And it's also hard not to take note that a self-described skeptic of AI hype is taking a job at the lab most convinced superintelligence is near. And he's building the tools to get it there faster. Whether that's reassuring or unsettling is unclear.

Sources