Alphabet had a rough week on personnel, with four people connected to its model team departing, among them a central researcher and a prize-winning scientist, heading to a direct competitor.
A top researcher leaving has a different effect from ordinary attrition. In a field where the decisive knowledge isn't written in a document but accumulated as judgement about what works, the person carries away much of what the company would try to protect by contract.
A legal dispute between two large companies over trade secrets would turn on the identical question weeks later: where personal knowledge ends and company asset begins is a line nobody knows how to draw precisely.
For the market, the move makes clear that the entry barrier in AI isn't only capital and compute. It's also a very small number of people able to run frontier-scale training.
In the following weeks, the practical effect appeared on the calendar: an important model from the company was delayed over performance below expectations.