'London Rule' that explains why AI researchers are leaving Google for Anthropic & OpenAI

Four senior AI researchers have left Google in the past few weeks, and the through-line isn't pay or position—it's a city. Most of those walking out of DeepMind for Anthropic and OpenAI are based in London, and that's the detail worth following. The latest to go are Jonas Adler and Alexander Pritzel, both described by people familiar with the matter as central to Google's Gemini model. Adler worked on the company's AI coding push; Pritzel sat inside the training pipeline that teaches models how to behave. Both are headed to Anthropic, the Claude maker that has spent the past year quietly draining Google's research bench. They follow John Jumper, the Nobel laureate who co-led AlphaFold, and Noam Shazeer, a co-author of the 2017 transformer paper who left for OpenAI after more than two decades at the company.