Jeff Dean, a longtime senior Google executive, is departing to launch an AI research startup called Discovery Loop, joined by fellow Google veterans Sanjay Ghemawat, Quoc Le, and Oriol Vinyals, with Dean reportedly set to serve as CEO. The public benefit corporation aims to use AI to run thousands of simultaneous experiments, automating scientific research processes traditionally limited by slow human iteration. The company said its systems are designed to leverage massive computational scale to transform how quickly research can advance, and the founding team described discovery as AI’s next major frontier beyond simply answering questions. Discovery Loop has secured funding co-led by Radical Ventures and Khosla Ventures, with participation from Kleiner Perkins, Lightspeed, Doerr Capital, and Google parent Alphabet. Dean, who joined Google in 1999 as its 30th employee, said the goal is to automate experimental loops to produce both greater quantity and higher quality research, ultimately accelerating scientific breakthroughs.
As well as this, Google announced new agentic capabilities for its “Ask Maps” feature within Google Maps, enabling users to order food, book hotels, and purchase event tickets directly through the tool. The feature will let users place restaurant orders through partners including Square, Toast, and Uber Eats, compare hotel pricing and availability, and browse local entertainment options. Google is also rolling out Personal Intelligence to Ask Maps, allowing the tool to draw on a user’s Gmail and Calendar data to personalize responses, though the feature remains off by default. Additional updates include conversational memory, letting users reference earlier planning discussions, and a live transit widget tracking real-time delays. The agentic features are launching in the U.S., while Personal Intelligence and the transit widget will roll out globally where Ask Maps is available.
In a third development, AI lab Mirendil has signed a multiyear partnership with Google Cloud worth more than $100 million to secure computing capacity for research into self-improving AI systems, according to co-founder and CEO Behnam Neyshabur. The deal grants Mirendil access to Google’s TPUs, Nvidia GPUs, and managed training infrastructure. Neyshabur said the goal is to build AI capable of continuously improving its own knowledge and performance in fields such as Alzheimer’s research, medicine, and materials science. Co-founder Harsh Mehta said Google’s diverse hardware options allow the startup to match workloads efficiently across chip types, reducing costs. Amin Vahdat, Google’s SVP and chief technologist of AI and infrastructure, said advancing AI now depends on orchestrating entire intelligent systems rather than individual chip performance alone.
