The Best People Are Already Moving

Two departures in seven days

Noam Shazeer left Google DeepMind for OpenAI. He is not just any researcher. He co-invented the Transformer architecture, the foundation of modern AI. Google paid 2.7 billion in 2023 to bring him back from Character.AI, and he still left.

John Jumper left seven days later. Nobel laureate, AlphaFold author, nine-year DeepMind veteran. He went to Anthropic.

The same week, OpenAI added Dean Ball, former White House AI adviser, as Head of Strategic Futures. It published o3 results at Boston Children's Hospital, with 18 unsolved rare pediatric disease diagnoses confirmed by clinicians in NEJM AI.

The leaked numbers

A leaked internal DeepMind report said morale is increasingly negative. It said the company's models have slipped to fifth place on industry benchmarks. Staff worry leadership is conceding the race to AGI.

Those words matter less than the actions they describe. A fifth-place model with a bright future hires the world's best people. A first-place model with a darkening one loses them.

Benchmarks are easy to fix with compute. Morale and trust are not.

The case that disagrees

Google continues to publish safety roadmaps and fund academic multi-agent risk research. It has distribution, capital, and infrastructure that neither OpenAI nor Anthropic can match. The argument that Google will win by being the platform underneath everyone else is not crazy.

But infrastructure without the people who build the models is not an AI company. It is a hosting business with good brand recognition.

The recruiting argument that matters

Trump told Axios that Anthropic has behaved very responsible on the Fable 5 export controls, signaling possible restrictions relief soon. The Fable ban, intended to slow Anthropic, may have become its most powerful recruiting pitch: come build frontier models in a place the government is trying to protect instead of punish.

That is a signal the best researchers hear. Top AI researchers do not join companies for salary alone. They join the lab that is moving fastest, working on the hardest problem, and staying ahead of the policy risk. If one lab becomes the one the government leaves alone, the decision is not difficult.

What this means for builders

You can read benchmark tables. You can parse safety papers. You can track stock prices. The most reliable indicator of where the industry is heading is where the best people choose to work. Right now they are choosing Anthropic and OpenAI.

That will not last forever. Markets correct. Leadership changes. A competitor with a clear mission, capital, and political cover can pull people back.

For now, the floor of the industry is shifting under the labs that are losing people. If your strategy depends on a specific provider's model line, watch the departures, not the announcements.