Jeff Dean spent nearly three decades inside Google. He shaped the company’s search engine, built foundational systems for its advertising empire and steered its artificial intelligence efforts through the rise of deep learning. On August 5, he walked away. Along with three longtime collaborators, Dean launched Discovery Loop, a startup that wants to turn AI into an autonomous researcher capable of running thousands of experiments in parallel.
The move sent ripples across Silicon Valley. Within days investors began circling. Business Insider reported this week that Dean is in talks for roughly $1 billion in new financing at a valuation near $10 billion. No product exists yet. No customers. No revenue. The number reflects belief in the team more than any tangible progress.
Discovery Loop operates as a public benefit corporation. Its stated mission centers on automating machine learning, science and engineering. The founders believe current research moves too slowly, constrained by human iteration. Their approach replaces sequential experiments with massive parallel loops. Propose an idea. Implement it. Evaluate results. Repeat at scale.
“We think there is opportunity for AI to more fully automate what has traditionally been a very human-intensive experimental loop,” Dean told The New York Times. “You will get both a higher quantity and a higher quality of experiments, and that will lead to scientific breakthroughs and advances.”
The four founders read like a greatest-hits list from Google’s AI history. Sanjay Ghemawat joined Dean in 1999 and co-developed core distributed computing technologies that still power much of the company’s infrastructure. Quoc Le helped launch Google Brain and drove advances in large-scale neural networks and AutoML systems. Oriol Vinyals contributed key work on sequence modeling and reinforcement learning at DeepMind while serving as technical lead on Gemini.
Dean himself co-founded Google Brain, oversaw development of TensorFlow and the company’s custom TPUs, and played a central role in Gemini. Their combined resumes include foundational contributions to Google Search, Ads, Gmail and the modern AI era. A pitch deck the team shared on X highlighted these achievements. One former Google product leader called it “one of the most stacked pitch decks ever made.”
Investors needed little convincing. Radical Ventures and Khosla Ventures co-led the initial round. Lightspeed, Kleiner Perkins and Doerr Capital joined. Alphabet, Google’s parent, signed on as a founding investor and cloud partner. Sundar Pichai, Google’s CEO, publicly wished the team well while noting continued collaboration.
Vinod Khosla met the founders on a Saturday to avoid signaling Google’s loss. He didn’t ask for many details. “With this team, I wouldn’t need to know what they were doing before I backed them,” Khosla said in Wired. “It’s the ultimate superstar team.” He added that the venture flips the usual relationship. “Humans have been using AI to do research, not using AI to be a researcher. The fundamental thing is that AI is the researcher.”
Jordan Jacobs, managing partner at Radical Ventures, will join the board. “Building something to solve all those problems is a powerful idea,” he told Wired. “These people have been doing this kind of work in the past so they know what they’re doing.”
The idea for Discovery Loop came together quickly. The four had been exploring similar concepts separately before realizing overlap. They sketched a simple three-slide deck without relying on AI tools. Meetings happened at Dean’s house. By late July the company existed. Dean, somewhat sheepishly, took the CEO title after the others pointed at him.
Goals stretch across domains. The team wants to automate discovery in biology, drug development, materials science and chip design. Early focus falls on improving AI itself. Dean has expressed particular interest in discovering new transformer architectures. Oriol Vinyals emphasized the challenge of generating novel ideas. “One of the things that we’ll be obviously very focused on is how these models come up with new ideas to try,” he said in Wired. “That’s not something that currently they’re super strong at.”
Quoc Le echoed excitement around machine learning automation. The founders see potential for small teams to outpace giant research organizations once automation deepens. Humans would co-develop ideas at first. Full autonomy remains the long-term target.
Google’s reaction mixed pragmatism with nostalgia. In an official statement Pichai noted that Dean and Ghemawat “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.” The company will work with Discovery Loop on research frameworks and infrastructure. Alphabet’s investment keeps a foot in the door.
Yet the departure highlights tension. Dean’s influence had somewhat diminished inside Google as other executives, including Sergey Brin, asserted greater control over AI strategy, The New York Times reported. Oriol Vinyals cited organizational inertia. “In a large organization there is always a lot of inertia you have to overcome to make any radical changes,” he explained in Wired. “We want to build something different.”
Dean and his co-founders are not alone. Talent has poured out of Google and DeepMind in recent years. David Silver, a former DeepMind researcher, raised $1.1 billion this year for Ineffable Intelligence at a $5.1 billion valuation. Another pair of ex-DeepMind scientists reportedly targeted a $25 billion pre-money valuation for their new lab. Sakana AI, co-founded by former Google researchers, reached $2.65 billion last year.
These exits reflect a broader market dynamic. Top AI minds command enormous sums. Companies pay tens or hundreds of millions simply to retain individual researchers. Dean’s group brings collective institutional knowledge that no single hire could match.
The joint statement from the founding team captured their ambition. “The next great frontier for AI is to go beyond answering questions and to begin making discoveries,” they said, according to TechCrunch. “By fundamentally accelerating how engineering and scientific discovery are conducted, we can deliver the benefits of transformative technologies to the world far sooner.”
They also described the historical bottleneck. “While science and engineering have tremendously advanced society over past centuries, progress has traditionally relied on slow, sequential human iterations, creating a significant bottleneck. Discovery Loop is developing advanced AI systems that leverage massive computational scale to fundamentally transform the speed and efficiency of innovation by automating complete experimental loops.”
Skeptics might point to the gap between vision and execution. Running thousands of meaningful experiments requires not just compute but high-quality data, robust evaluation frameworks and ways to translate results into actionable breakthroughs. The team has built such systems before inside Google. Whether they can replicate that success independently, without the company’s resources and data moat, remains unproven.
Yet the $10 billion valuation talks suggest many investors have already placed their bets. The price echoes recent sky-high deals for AI talent but stands out for a pre-product company. Details could still shift. Dean’s representatives declined to comment on fundraising specifics.
Discovery Loop’s focus on public benefit status signals longer-term thinking. The founders appear conscious of societal impact as they pursue automated discovery. If successful, their systems could speed drug development, optimize chip architectures or uncover new materials at rates previously unimaginable.
And the timing feels deliberate. AI capabilities have advanced enough to make recursive improvement plausible. Models can now propose hypotheses, generate code for experiments and analyze outcomes in ways that were science fiction a decade ago. The question is how far that loop can be pushed before hitting hard limits.
Dean has hinted at optimism. In a talk that secretly previewed the startup’s ideas, he described the automated scientific method. Propose. Implement. Evaluate. Scale the loop. The results, he suggested, could unlock advances across multiple fields simultaneously.
Google, for its part, continues reorganizing. Demis Hassabis stepped back from day-to-day leadership at DeepMind to become chairman while retaining a chief scientist title. Other internal shifts suggest the company is adapting to life after its most visible AI architect.
The broader industry watches closely. Dean’s departure adds to a pattern of top researchers striking out on their own, often with former employer’s blessing and capital. It underscores how talent has become the scarcest resource in AI development.
Whether Discovery Loop delivers on its ambitious promises won’t be clear for years. The technology required sits at the frontier. But the team has a track record of turning ambitious ideas into production systems that reshape entire industries. Investors are wagering that history will repeat, this time outside Google’s walls.
So far the market agrees. A near-$10 billion valuation before writing much code speaks volumes about confidence in what this particular group might achieve. The real test lies ahead, in the experiments they run and the discoveries that follow.
View Comments (0)