Top Google AI Researchers Adler and Pritzel Head to Anthropic
Jonas Adler and Alexander Pritzel, two senior AI researchers at Google DeepMind, are departing for Anthropic — the latest in a string of high-profile exits that includes Noam Shazeer and Nobel Prize-winner John Jumper.
Original sourceGoogle DeepMind is losing two more prominent researchers to a direct competitor. Jonas Adler, known for his work on diffusion models and scientific ML, and Alexander Pritzel, a contributor to memory-augmented neural network research, are both joining Anthropic. The departures follow those of Noam Shazeer, who co-invented the Transformer architecture and returned to found Character.AI before a landmark acquisition, and John Jumper, the AlphaFold lead who won the 2024 Nobel Prize in Chemistry.
The pattern is hard to dismiss as coincidence. Google has long been considered the gravitational center of AI research — the place where foundational work happened — but that reputation is under pressure. Anthropic, co-founded by former Google and OpenAI researchers including Dario and Daniela Amodei, has built a culture explicitly positioned around safety-focused, rigorous research, which appears to be a draw for scientists who want their work to carry both intellectual weight and perceived mission alignment.
For Google, the concern isn't just losing individual contributors — it's losing the density of talent that makes breakthrough research possible. Research organizations operate on clustering effects: the best researchers attract the next generation of best researchers. Each high-profile departure raises the question of whether Google DeepMind can maintain the critical mass that produced AlphaFold, Gemini, and the Transformer in the first place.
Anthropichas been on an aggressive hiring push, backed by substantial investment from Amazon and others. The company's headcount growth, combined with its track record on Claude, appears to be making it a credible long-term research home — not just a startup bet. Whether these departures represent a structural shift in where foundational AI research happens, or just normal talent churn at an accelerated rate, is the defining question for the research ecosystem over the next few years.
Panel Takes
The Skeptic
Reality Check
“The Transformer was invented at Google. AlphaFold was invented at Google. The researchers who built those things are now leaving Google. That's not normal attrition — that's a thesis statement about where the serious work is moving. The prediction here is simple: Anthropic wins the research credibility race within 18 months, and Google's moat becomes infrastructure and distribution, not scientific primacy. That's a survivable position for a trillion-dollar company, but it's a different company than the one that defined the last decade of ML.”
The Futurist
Big Picture
“The thesis embedded in these moves is that research autonomy and mission clarity compound over time — that the best scientists will consistently choose environments where their work has directional coherence over environments where it has maximum resources. If that's true, the second-order effect is that Anthropic becomes the new attractor for foundational talent, and the papers that define the next architecture generation get written there, not at Google. The dependency is that Anthropic's safety framing stays intellectually serious rather than becoming regulatory theater — the moment it drifts into the latter, this talent calculus reverses.”
The Founder
Business & Market
“From a pure business lens, Google's real risk isn't the researchers themselves — it's the signal effect on recruiting pipelines for the next five years. Top PhD candidates choose labs partly based on where their role models land, and right now that answer is increasingly 'not Google.' The moat Google has is compute, distribution, and enterprise relationships, none of which requires being the best research org in the world — but if they lose the narrative entirely, they lose the ability to attract the researchers who build the next-generation models that keep Gemini competitive. That's a slow bleed, not a cliff, but it's real.”
The PM
Product Strategy
“The job these researchers are being hired to do at Anthropic is clearly defined: build models with a safety-first research culture and ship them as Claude. At Google, the job-to-be-done for a top researcher is genuinely murkier — is it Gemini, is it DeepMind research, is it Google Brain legacy work? When a product has too many 'and/or' answers to what it's doing, it loses focus; the same is true for research organizations. Anthropic's singular product focus is doing real recruiting work here, and that's a product strategy lesson Google's leadership should be taking seriously.”