Google has lost two more senior architects of its Gemini model. Jonas Adler and Alexander Pritzel, who played key roles in Gemini’s development, are joining Anthropic, Bloomberg reported on June 24.
The departures land just days after two other high-profile exits. Noam Shazeer, who had been at Google since 2000 and returned via the $2.7 billion Character.AI acqui-hire, announced he was leaving for OpenAI. Days later, John Jumper, a Google DeepMind director who shared the 2024 Nobel Prize in Chemistry with DeepMind CEO Demis Hassabis for the AlphaFold protein-structure work, said he was also heading to Anthropic. TechCrunch cited the Bloomberg reporting on all four departures.
Four named, senior departures in roughly a week is not routine attrition. It is a public signal about where researchers believe they can do their best work. Google built Gemini with exactly this talent. Losing the architects while the model is still in active development creates both an institutional knowledge gap and a reputational narrative that is difficult to reverse.
The structural incentive is straightforward. As TechCrunch noted, Anthropic and OpenAI are both preparing to go public. Pre-IPO equity at a company on a strong growth trajectory is a concrete financial offer that a large public company cannot match dollar-for-dollar inside its compensation bands. Google can offer salary and existing equity; it cannot offer the asymmetric upside of a private-company grant at a pre-IPO valuation.
What makes the Adler and Pritzel moves particularly pointed is the destination. Shazeer went to OpenAI. Jumper, Adler, and Pritzel went to Anthropic. That pattern says something about where Anthropic sits in researcher perception: not merely competitive with Google, but credible enough to attract people who built the model Anthropic’s Claude is measured against on every benchmark.
For teams evaluating which frontier model to build on, researcher flow is a lagging indicator of capability trajectory. The people who designed a model’s architecture carry intuitions about its limits, its failure modes, and its headroom. When those people leave for a competitor, they do not bring code. They bring judgment. That judgment will shape what Anthropic builds over the next two to three years.
Google has not publicly commented on the departures, according to TechCrunch. The company has significant resources, deep infrastructure, and its own pipeline of researchers. A few departures do not determine the outcome of a multi-year model competition. But the public clustering of exits, the seniority of those leaving, and the timing relative to Anthropic’s IPO preparation all point in one direction.
Teams currently committed to Gemini APIs should watch the next two Gemini release cycles closely: researcher departures of this kind typically surface in capability deltas six to eighteen months later, which is exactly the window their next contract decision will cover.
Reported by TechCrunch, citing Bloomberg, on June 24, 2026.
