Dear Magicians,

Last week, a 2025 Nobel laureate told me, on tape, that Google demoted him. Not for failing. For succeeding.

John Martinis led the team that achieved quantum supremacy — the moment a machine did in minutes what the world’s best supercomputer could not. Then Google “reorganized.” He went from head of hardware to what he called “one-over-N authority” — And that his authority had been reduced to a fraction of a percent of what it used to be a committee role, in effect. He tried it for nine months. He discovered what he calls “negative authority,” which he analogizes to parenting a teenager. Then he left.

“I was too much like Elon Musk to work at Google,” he told me said. Previously he said that “I had already lost control. I wasn’t really leading the hardware group.” His former UCSB students “naturally wanted to leave their advisor and form their own group and be independent.” Management and the team together decided the co-leader model was better. John clearly didn’t — he left 9 months after the re-org and 5 years before he’d win the Nobel Prize. Wonder what Google would have done had they known about the Nobel Prize? Actually, I don’t wonder at all….

Notice what’s actually being said here, because it might be the most important sentence in the entire conversation. The same Google that built him up to win the quantum race could not metabolize the kind of person who wins quantum races. Not the science. The person. The archetype. The ‘founder-as-father’ who makes a thousand decisions a day and owns every one.

The symmetry is striking. The co-authors of “Attention Is All You Need” — the paper that launched modern AI — also all left Google. Same company, same pattern. The institution that enables the breakthrough cannot survive the breacher.

This is not a Google problem. It’s an institutions problem. Physicists have a name for it: decoherence. The irreversible loss of phase coherence between a system and its environment. A coherent signal enters a system. The system interacts with its environment. The signal decays into thermal noise. The system is now stable, predictable, and useless.

John Martinis’ crime wasn’t incompetence. It was coherence. He was too phase-locked to the original problem. He didn’t want to be reorganized. He wanted to build a million-qubit machine, not a roadmap that pleased a steering committee.

The Talmud tells us to carry two notes in our pockets. One says: “I am nothing but dust and ashes.” The other: “The whole world was created for me.” Every Nobel laureate I know carries the second one. Institutions can’t.

Even if you’re not doing Nobel-quality work, here’s what should concern you : The person capable of your next breakthrough is, by structural necessity, the person your institution is most likely to lose. Innovation is, in this sense, a decoherence event. The signal that arrives is the signal that leaves.

Martinis did what physicists do when the apparatus stops working. He built a new one. With Alan Ho, he started a company to do what Google wouldn’t let him do: scale qubits the way you scale semiconductors. Deposition and etch. The way the entire chip industry has worked since 1955.

“The right way to do it,” he says, with the disarming directness of a man who’s already survived a Nobel and a demotion.

The rest of us are still arguing about whether AI can replace us. John Martinis is in a cleanroom.

I think he knows what he’s doing.

Until next time, have a M.A.G.I.C. week.

Brian