When your résumé includes building the world's largest search engine, most lucrative ad platform, and the custom chips that power AI itself, your Series A pitch deck isn't a pitch—it's a victory lap with a hand out.
The Summary
- Jeff Dean, Google's chief scientist for 27 years, is leaving with three fellow AI legends to launch Discovery Loop, a public benefit corporation focused on automating the experimental loop in science and engineering
- The pitch deck lists founder achievements including building Google Search, Gmail, Ads, Gemini models, and TPUs—products that collectively touch billions of users and generate hundreds of billions in revenue
- The company aims to automate hypothesis-to-experiment-to-results cycles at scale, enabling "parallel execution of thousands of experiments" to solve problems traditional research can't crack
- Former Google product leader Bilawal Sidhu called it "one of the most stacked pitch decks ever made"—and he's not wrong
The Signal
Jeff Dean isn't just Google's chief scientist. He's the engineer who built the infrastructure that made Google possible. MapReduce. BigTable. TensorFlow. He's taking the band with him: co-founders Sanjay Ghemawat, Quoc Le, and Oriol Vinyals collectively worked on Search, Ads, Gmail, Gemini, and Google's custom TPU chips. This isn't four smart people leaving to start something. This is the departure of institutional knowledge you can't replace.
Discovery Loop is structured as a public benefit corporation with a mission to "automate machine learning, science, and engineering to accelerate discoveries and progress." The core idea: scientists and engineers spend most of their time on the grunt work—setting up experiments, running them, parsing results, then doing it again. What if agents did that loop for you, thousands of times in parallel, while you stayed at the hypothesis level?
"The loop part of the name comes from automating the process by which scientists propose an experiment, implement and run it, and evaluate results."
The pitch deck Dean shared includes a slide listing what the founding team built:
- Google Search (the world's largest search engine)
- Google Ads (one of the most profitable businesses ever created)
- Gmail (1.8 billion active users)
- Gemini (Google's frontier AI models)
- TPUs (the chips that trained those models)
Dean told followers on X that the deck had been shared with "a handful of VC firms". Translation: this isn't a roadshow. This is a selection process. When your slide deck is a list of the internet's critical infrastructure, you don't pitch investors. They pitch you.
The automation angle is the real story. Right now, AI agents handle narrow tasks—write code, summarize documents, book meetings. Discovery Loop is aiming higher: automating the entire experimental method in science and engineering. Propose a hypothesis. Let the system design the experiment, run it in parallel across thousands of parameter sets, analyze the results, and propose the next round. The scientist stays in the loop but exits the drudgery.
The Implication
If Discovery Loop works, it doesn't just speed up research. It changes what research means. The bottleneck in most fields isn't ideas. It's iteration speed. Pharma takes a decade to test one drug candidate. Materials science runs one experiment at a time. Climate models take weeks to train. Automating that loop at scale collapses timelines from years to weeks.
Watch what happens to corporate R&D budgets. If four ex-Googlers can automate scientific discovery with agents, every major lab will try to build or buy the same capability. This is the agent economy moving past productivity theater into actual scientific leverage. The people who figure out how to direct these systems, not run the experiments themselves, will be the ones who matter. Dean just made the first serious bet that we're already there.