Two of Google's most prominent Gemini researchers, Jonas Adler and Alexander Pritzel, are leaving the company for Anthropic, according to a Bloomberg report. Both played key roles in the development of Gemini, Google's flagship frontier model. Their exit lands in the same month that Google lost Noam Shazeer to OpenAI and Google DeepMind Director John Jumper to Anthropic — three departures in a matter of days, all to Google's two closest rivals.
The pattern is sharper than any single hire. Shazeer announced last week that he was leaving Google for OpenAI, ending a tenure that began in 2000. The only break in those 26 years was the three years Shazeer spent building Character.AI, the consumer chatbot startup that Google effectively acquihired for $2.7 billion — a deal whose central purpose, by most read-throughs, was to bring Shazeer back to work on Gemini. That investment now looks like a very expensive bookmark.
Days after Shazeer's announcement, Jumper said he was leaving Google DeepMind for Anthropic. Jumper shared the 2024 Nobel Prize in Chemistry with DeepMind CEO Demis Hassabis for their work on AlphaFold, the system that predicts 3D protein structures from amino acid sequences. AlphaFold is among the most-cited applied AI achievements of the past decade, and Jumper's name is attached to it as closely as Hassabis's. Losing him is not a routine engineering departure — it is a signal about where the most ambitious scientific AI work is now perceived to be done.
Key facts
- 01Jonas Adler and Alexander Pritzel, both key contributors to Gemini, are leaving Google for Anthropic.
- 02Noam Shazeer left Google for OpenAI last week, ending a tenure that began in 2000 (minus three years at Character.AI).
- 03Google effectively acquihired Character.AI for $2.7 billion in part to bring Shazeer back to work on Gemini.
- 04John Jumper, 2024 Nobel laureate in Chemistry for AlphaFold, is leaving Google DeepMind for Anthropic.
Adler and Pritzel sit one rung down in public profile but are arguably more operationally costly to lose. Gemini is the model Google is shipping into Search, Workspace, Android, and its cloud business simultaneously. Researchers who shaped its architecture carry context that is difficult to replace on a quarterly timeline, and Anthropic gets that context by writing an offer letter.
Google did not respond to a request for comment. The company has not publicly addressed the cluster of exits, and there is no indication yet of counter-offers, retention grants, or organizational changes in response. Anthropic has not detailed what Adler, Pritzel, or Jumper will work on, though the company's recent hiring has concentrated on frontier model research and agentic systems built on Claude.
The timing is not incidental. Both OpenAI and Anthropic are preparing to go public, which means private-company equity grants made now convert into liquid stock on a known horizon. For a senior researcher weighing a move, pre-IPO equity at a company growing into a listing is a different financial instrument than additional Google RSUs at a $2T-plus market cap. The recruitment pitch writes itself.
It is also a referendum on where the interesting work happens. Google still publishes more peer-reviewed AI research than any other company and runs DeepMind, one of the two or three most decorated AI labs in the world. But the researchers leaving are not citing access to compute or publication freedom. They are going to companies whose entire corporate identity is the frontier model — places where the model is the product, not a feature wedged into seven existing products with seven existing product managers.
The counterargument from Google's side is that the company has depth. Gemini's roadmap is set by hundreds of researchers and engineers, not four, and the model has continued shipping major updates through this period. DeepMind's bench in protein modeling, robotics, and reinforcement learning is genuinely deep. A handful of high-profile exits, even painful ones, do not by themselves break a research program of Google's scale. Whether they break momentum is a different question, and one that will be visible in the cadence of Gemini releases over the next two quarters.
For Anthropic and OpenAI, the talent inflow compounds. Each senior hire from a rival lab brings tacit knowledge about that lab's training infrastructure, data pipelines, and unsolved problems. It also makes the next hire from the same lab easier, because the social network of who-works-where bends toward the destination. Google spent $2.7 billion to bend that network back toward Mountain View. Three weeks later, the bend has reversed.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
