Jeff Dean and Google researchers form Discovery Loop
Jeff Dean is stepping down from Google to launch Discovery Loop, an artificial intelligence startup focused on scientific research. He is joined by Google researchers Sanjay Ghemawat, Quoc Le and Oriol Vinyals, with Dean expected to serve as chief executive.
Discovery Loop is organised as a public benefit corporation. Its stated goal is to use AI systems and large-scale computing to initiate and iterate thousands of experiments simultaneously, partially automating the research process and expanding the scale of experimentation.
An AI venture focused on experimental loops
The company says science and engineering have traditionally depended on slow, sequential human iterations. Discovery Loop intends to develop AI systems that automate complete experimental loops, seeking to change the speed and efficiency of innovation through massive computational scale.
Its ambitions also include using AI to help create more powerful AI, a process described as recursive self-improvement. The founders argue that the next frontier is not only answering questions, but making discoveries by accelerating engineering and scientific work.
Founders bring Google infrastructure and AI experience
Dean joined Google in 1999 as its 30th employee. His work has included core search infrastructure, notably crawling, indexing and query-serving systems, as well as a leading role in the company’s early AI research and contributions to Gemini’s multimodal models.
Ghemawat is a senior fellow and engineer at Google. Le was a founding member of Google Brain, while Vinyals is a senior research scientist at Google DeepMind. Their departure follows a broader movement of leading researchers into independent AI ventures, including Google researchers joining Anthropic and OpenAI departures from Google toward Anthropic and OpenAI, as competition for advanced-model expertise continues.
Funding and implications for research teams
Alphabet has provided financial support to Discovery Loop. The startup says its initial funding round is co-led by Radical Ventures and Khosla Ventures, with Kleiner Perkins, Lightspeed and Doerr Capital participating.
For businesses and research organisations, the development underscores a practical question: which experimental workflows can be scaled with AI while retaining appropriate human oversight. Teams evaluating such systems will need to define experiments, review outputs and measure whether increased iteration produces useful results.