Google’s Mind Is Now Less Deep

To me, at least until we have more actual reporting on the matter, the most interesting element of the Google bombshells today is simply the timing.

Just Jeff Dean leaving Google would obviously be a massive story by itself. I mean, he's been there almost 27 years. He was employee #30. As everything you read today will repeat, he was not just vital, but instrumental to a lot of what made – and makes – Google, well, Google. He's one of just two Google "Senior Fellows" – the company's highest technical honor.¹

Of course, the other Google Senior Fellow, Sanjay Ghemawat, is also leaving. Less well-known externally, but just as critical to the history of the company (and industry) for his technical work alongside Dean over decades, he's now joining his compatriot in forming a new startup.

So yes, just news of his departure alone would also be massive. But together, as Steven Levy so colorfully puts it: "that’s like Mick Jagger and Keith Richards ditching the Rolling Stones to start a new band."

But those two are joined by two other seemingly critical people to Google's current AI efforts in Oriol Vinyals, the VP of research at DeepMind (and a technical lead for Gemini), and Quoc Le, a cofounder of Google Brain. As an aside, sort of harsh that Sundar Pichai's note calls out Dean and Ghemawat but not Vinyals and Le? Then again, a lot of senior people have been leaving Google recently...

Which leads me to...

These departures were announced the same day that it was announced that Demis Hassabis would be stepping back – sorry "stepping up" or "aside"? – from his day-to-day work at DeepMind and into a more nebulous and higher minded role as Alphabet's Chief Scientist (though also still leading day-to-day at Isomorphic Labs, his other Alphabet project where he's CEO). I mean, what is going on at Google?!

The answer, at least in what you read right now, is perhaps less dire than it may seem on the surface. Dean and team, after collective decades on the inside, seemingly wanted to try their hands at a startup in the hottest space during the hottest time to have a startup: the AI Boom. And they seemingly picked the hottest current sector to start with: recursive self-improvement.

Yes, RSI is the new AGI, at least until RSI leads to AGI. Maybe. One day.

Anyway, on the DeepMind side, everything you read suggests that Hassabis had already effectively handed off day-to-day operations to DeepMind SVP Koray Kavukcuoglu. The fact that Kavukcuoglu was (and will remain) Google's 'Chief AI Architect' – a title which he may or may not have received when a certain Mark Zuckerberg came calling with a "Godfather" offer – lessens any blow to actual operations even further. He already reported directly to Pichai. He will continue to report to Pichai.

At the same time, he will not be the new CEO of DeepMind – because no one is getting that title. Because DeepMind is now clearly going to be run less autonomously and more as a part of Google's overall AI efforts.²

And that's probably a good thing, to be honest. Because after already having to course correct when they fell behind in AI once, well, it sure seemed to be happening again. The reporting there this time points less to timidity and more the usual corporate in-fighting and bureaucracy. Google always has too many cooks in the kitchen, but the actual chef can't seem to keep them corralled and on the mission to make the meal. That's why we used to get a dozen chat apps, many of which are half-baked. And now a dozen AI coding projects.

Hopefully Kavukcuoglu has the mandate to unite the cooks. And to get the meal prepped and plated.

We'll see. But what remains wild to me is that all of these departures were announced at once. Couldn't Google convince either Hassabis or Dean's team to wait a week or two to blunt the impact – certainly on the market, but also internally? Maybe something was leaking. Or maybe they decided that packaging them together would actually help soften the blow versus the alternative.

Because the alternative not only includes all of the above people, but also includes Noam Shazeer, unexpectedly jumping ship a couple months ago – to join OpenAI, no less. And John Jumper – another Nobel Prize winner at Google alongside Hassabis³ – john-jumping ship, to join Anthropic, no less. And Peter Norvig – Google’s head of research for the past 25 years – heading to Recursive Superintelligence, working on, you-know-what.

You add all this up and it doesn't look just like a trend, it looks like an exodus. A brain bleed the likes we perhaps have never seen in such a short amount of time.

So yeah, had we gotten hit with Dean, Ghemawat, Vinyals, and Le – followed by Hassabis (even yes, with him technically staying, but in a new, less day-to-day role), it would have actually seemed even more catastrophic. Like Google couldn't stop the bleeding. Better to rip the band-aid off, perhaps?

Regardless, it's still an optics nightmare. I mean, these people were just on stage together a couple months ago talking about the future of AI at Google as three of the four co-leads of Gemini. Now they're all gone – except for Kavukcuoglu.

Again, I think there's a way to read this that is not so catastrophic and could actually end up as a good thing for the company and the AI efforts overall. Well, at least in so far as you think the need to productize and commercialize AI is vital to Google right now, with research and a push beyond LLMs (areas which have always clearly interested Hassabis far more) taking a bit of a backseat.

It's going to be hard for people to see this right now. I mean, just look at today's headlines matched with the other headlines about the departures and the fact that Google has clearly flubbed the Gemini 3.5 Pro model and... yikes.

But the flip side is that things seemingly weren't going great with the status quo so... a shakeup can be a good thing sometimes. You hate to lose talent, let alone historic talent, but the way forward is also likely to be paved with new technology and talent. Now Google needs to rally the troops to ensure they don't lose more. And that's probably what today's announcements were about. Take the lumps quickly, make it all seem quite orderly, and get everyone back to the kitchen. It's time to cook.

¹ Cue the "these go to 11" jokes as that's the level inside Google at which a "Senior Fellow" sits. Again, there were two of them yesterday. Now there are zero. ↩

² Will this lead to more retention issues? We'll see... ↩

³ Fun fact: there were actually three before Jumper left, with Michel Devoret, the 2025 Nobel Prize winner in Physics, leading up Google's quantum computing efforts. ↩