Bloomberg reported on June 24 that Jonas Adler and Alexander Pritzel, both contributors to Google's Gemini effort, are planning to join Anthropic. Adler worked on Google's AI-for-code program; Pritzel was on the team training Gemini models. Neither move was officially announced as of publication, and Google had not commented.
This is the fourth senior DeepMind departure in roughly a month. Nobel laureate John Jumper, of AlphaFold, left for Anthropic on June 22 to lead an AI-for-science effort. Noam Shazeer — co-author of the 2017 paper that introduced the Transformer — left for OpenAI on June 21. Losing the people who built Gemini's training stack and code stack back-to-back is a more pointed signal than any single resignation.
Bloomberg's read is that Anthropic's pre-IPO equity is doing what cash compensation alone cannot: putting researchers within reach of a one-time payday before the company prices its shares. The departures also come as Gemini 3.5 Pro slipped its June general-availability deadline and as Google leans more on Anthropic, Broadcom, and TPU partnerships — a posture that complicates any internal pitch that Gemini itself is still the prize.
A takeaway for learners: talent maps onto roadmaps. When you see a cluster of named contributors moving to one lab, expect that lab's near-term capabilities to shift in their direction — coding, training methodology, biology. Skim author lists on the foundational papers occasionally; the people in those credits explain a lot of why specific products show up where they do, and they are often a better leading indicator than press releases.