Jeff Dean leaves Google to launch Discovery Loop AI startup

Jeff Dean is leaving Google to launch a new artificial-intelligence startup, joining a small group of senior researchers and engineers who helped build the company’s modern computing and AI stack. TechCrunch reports that Dean will serve as CEO of the new venture, Discovery Loop, and that co-founders include Sanjay Ghemawat, Quoc Le and Oriol Vinyals.

Discovery Loop is being set up as a public benefit corporation and says it wants to use AI to accelerate scientific research by automating more of the experimental cycle. The company’s pitch is that scientific progress is still constrained by slow, sequential human iteration; its proposed alternative is to use large-scale computation to initiate and refine many experiments in parallel, turning hypothesis, test and revision into a tighter loop. Dean told the New York Times, according to TechCrunch, that AI systems could automate parts of what has traditionally been a human-intensive process.

The move lands at a moment when large tech firms are simultaneously the main funders of frontier AI and the main bottleneck for researchers who want to build outside corporate roadmaps. Dean has been at Google since 1999 and was its 30th employee, with a career spanning search infrastructure—crawling, indexing and query-serving systems—and later leadership in the company’s AI efforts, including work tied to Gemini models. When figures with that institutional reach leave, they take with them not only technical credibility but the informal networks that decide what gets compute, what gets published, and what becomes a product.

TechCrunch says Discovery Loop has financial backing from multiple sources, including Alphabet, and that its initial funding round is co-led by Radical Ventures and Khosla Ventures, with participation from Kleiner Perkins, Lightspeed and Doerr Capital. That mix—corporate support alongside venture capital—suggests a structure designed to keep one foot inside the platform economy while trying to operate with the speed and focus of a startup.

The company’s stated ambition is to push AI beyond answering questions toward “making discoveries,” including interest in recursive self-improvement to build more powerful systems with less human iteration. In practice, that promise depends on access to compute, data and lab capacity—resources that are scarce and increasingly priced like strategic inputs.

Dean’s departure is a personnel story, but it is also a reminder of where the leverage sits. The researchers are leaving Google; the compute and capital that make their plans plausible are still largely controlled by the same few balance sheets.