Jeff Dean leaves Google after 27 years for 4-person AI startup

Jeff Dean, Google's former chief scientist, left the company this month after 27 years to found Discovery Loop, a four-person AI startup automating scientific research, a move that erased roughly $190 billion from Alphabet's market value in a single day.

"They have a plan for making the Gemini models awesome," Dean said at a Stanford University talk on August 7, his first public appearance since departing, while appearing to choke back tears about leaving the company.

Discovery Loop, structured as a public benefit company, aims to automate the full cycle of scientific experimentation — proposing, implementing, running, and evaluating experiments. The startup targets portions of the 14 National Academy of Engineering Grand Challenges, ranging from "Reverse Engineer the Brain" to "Prevent Nuclear Terror." Dean is in talks to raise $1 billion at a roughly $10 billion valuation, with Radical Ventures and Khosla Ventures leading the initial round alongside Lightspeed, Kleiner Perkins, and Doerr Capital. Alphabet is a founding investor and cloud partner, according to Google CEO Sundar Pichai.

The departure highlights a structural shift in AI talent economics. Dean argued that the growing availability of cloud compute lets small teams "rely on" providers like Google to build infrastructure, making it possible for a handful of people to raise significant capital without building their own platforms. "Sometimes a focused small company with everyone focused on exactly that mission" can outperform larger organizations, he said.

Cloud Compute Levels the Playing Field

Dean's move reflects a broader trend of AI-powered startups scaling with only a handful of employees. Modern startups are proving that AI agents can automate processes that once required multiple staff, allowing tiny teams to compete with much larger incumbents. Dean said working on a smaller team lets his startup cut through "slight distractions" that hamper progress in large organizations.

"We're super excited as a small company about working in a really focused way on science and engineering automation," he said.

Skim 10 Papers, Not One

At the Asian American Scholar Forum's 2026 Frontier & Pioneer Symposium, Dean offered practical guidance for young people entering tech. "I often tell students it's better to skim 10 papers than to read one in detail because you then get 10 points in your cloud of what might be possible," he said. "Or, even skim 100 abstracts because what you want to be able to do is connect important ideas that have not yet been connected."

Dean also advised targeting problems with a five-year horizon. A problem that could take 20 years to solve is too ambitious without a clear attack plan, while a two-year problem may be too obvious for a major breakthrough. "The perfect shape of a problem that you want to work on in a reasonably long-term manner is like five years or something," he said. "Try lots of things that might not work. Some of them will."

Dean remains bullish on AI's potential despite concerns about unemployment and inequality. "By building models that are really good at understanding many different domains of science and engineering you can get Ph.D.-level expertise in a model across many different domains," he said. The 58-year-old, who earned his PhD in computer science from the University of Washington in 1996, acknowledged the venture carries risk. "It's a little nerve-wracking," he said. "But at the same time it's also exciting."

For investors, Dean's exit from Google after 27 years — including stints as head of Google AI from 2018 to 2023 and chief scientist from 2023 to 2026 — raises questions about talent retention at Alphabet as competition for AI researchers intensifies. Alphabet shares fell 3.9 percent on the news, while Discovery Loop's $10 billion valuation target would place it among the most valuable AI startups before shipping a product.

This article is for informational purposes only and does not constitute investment advice.