Shazeer to openai, then John Jumper (the alphaFold nobel guy) to anthropic, plus Adler and Pritzler out the same door within a week. Every time one of these drops the framing is google is bleeding. I think people are reading it backwards. If the people who actually trained the thing can leave and instantly matter at a competitor, the weights were never the asset. The judgment about how to steer a model, what to eval it on, where it breaks, that stuff lives in heads not in checkpoints. Hardware you can buy. That you cannot. What it means for the rest of us is simpler than the talent drama. If capability is going to keep walking between labs every few months, betting your whole stack on one provider's model is a bet on that lab keeping its people, which is the one thing you cannot control. I...
No Comments 
