Here is what I keep thinking about after Wednesday: not the press release, not the stock ticker, but the slide. The pitch-deck slide that lists what Jeff Dean's founding team built reads less like a startup pitch and more like a museum catalog for the internet age. Google Search. Google Ads. Gmail. Gemini. AlphaFold. And that was only one page of the flex.
Jeff Dean, Google's chief scientist for 27 years and roughly its 30th employee, is leaving. He is not retiring to a beach. He is co-founding Discovery Loop with three other heavyweights from the same building: Sanjay Ghemawat, his collaborator since the MapReduce era; Oriol Vinyals, the Gemini technical lead; and Quoc Le, the mind behind AutoML-Zero. Four of the most decorated researchers in the field walked out together, and Google helped fund the exit.
The Pitch Deck Everyone Called an Insane Flex
Dean posted parts of the deck on X, and the internet responded with grudging awe. Startups normally pitch potential, projections, and promises. Discovery Loop pitches receipts. One slide lists the founding team's shipping record, and the only word people could find for it was "insane flex."
- Google Search, Google Ads, and Gmail: three products most of humanity touches every day.
- Gemini and AlphaFold: two of the defining AI systems of the decade.
- Google's first neural chatbot, plus a mentoring record that seeded a generation of founders.
Normally a deck like this is hype. This one is a biography. That is the point about the AI talent race that I keep coming back to: we have reached the stage where the most persuasive pitch in the industry is a list of things you already built.
Automate the Loop
The mission, in Dean's own words, is to automate machine learning, science, and engineering. Concretely, Discovery Loop wants to close the experimental loop: propose an experiment, implement it, run it, evaluate it, then do it again thousands of times faster than any human team can. It will start on its own machine learning systems and then generalize to chip design, drug discovery, biology, and materials.
"We were working on different things, but we were all starting to see the possibility of AI being able to automate scientific and engineering loops," Dean told Wired. Vinyals was more specific about the hard part: "One of the things that we'll be obviously very focused on is how these models come up with new ideas to try. That's not something that currently they're super strong at."
This is the part I find genuinely exciting. The bottleneck in modern science is rarely ideas. It is iteration. A graduate student runs dozens of experiments in a good year. An automated loop could run thousands. If Discovery Loop delivers even a fraction of that vision, the most important lab in a field stops being the one with the biggest building and becomes the one with the tightest loop.
It is also a little unsettling. Khosla Ventures and Radical Ventures are in, with Google itself as a founding investor and Cloud partner and a year of free compute thrown in. Vinod Khosla called it an honor to co-lead the round. The company is a public benefit corporation, which sounds noble, but the real test is governance: which discoveries get prioritized when the loop finds something valuable, and who gets to decide.
Why the Stock Dropped 4%
Here is my read on the market reaction. Alphabet shares fell about 4 percent after the announcement because investors finally treat AI talent as a material business issue. In the old software era, one famous engineer leaving was a culture story. In the AI era, it is a balance-sheet story, because a handful of researchers can move billions of dollars of perceived capability in a single afternoon.
The leadership shuffle adds context. Demis Hassabis steps up to chair Alphabet's AI unit and serve as chief scientist of the parent company, while Koray Kavukcuoglu formally takes the reins at Google DeepMind. It is a friendly reshuffle, but the direction of travel is unmistakable: the people who built the modern AI era are increasingly choosing to build the next one outside the walls of the biggest labs.
Sundar Pichai put the best face on it. "Over 27 years, Jeff and Sanjay helped to drive some of the most significant technology transitions, from our early search infrastructure to the neural networks that helped create the modern AI era," he said, adding that Google will work with Discovery Loop as a founding investor and Cloud partner. Keep your friends close, and your brilliant former employees closer.
I do not think this is the death of Big AI. Google still has extraordinary depth, and the reported $1.5 billion talks for the AI coding startup Mechanize show the company is still spending aggressively to buy the future it can no longer assume will stay in-house. But the takeaway for me is stubborn and simple. The era when a handful of giant companies could assume the best minds would stay forever is over. The pitch deck with the receipts won, and now it is a competitor.
Comments