cd /news/artificial-intelligence/radical-ventures-lists-jeff-dean-as-… · home topics artificial-intelligence article
[ARTICLE · art-116085] src=runtimewire.com ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

Radical Ventures lists Jeff Dean as Discovery Loop's CEO

Radical Ventures' portfolio page now lists Jeff Dean as CEO of Discovery Loop, an AI research venture he co-founded with Sanjay Ghemawat, Quoc Le, and Oriol Vinyals, resolving the ambiguity from its August 5 launch when Dean said 'I think I'm the CEO.' Discovery Loop's own site still presents the four founders without executive titles. The venture aims to build AI systems that propose and run experiments, starting with machine-learning research before expanding to fields like chip design and drug discovery.

read5 min views1 publishedAug 31, 2026
Radical Ventures lists Jeff Dean as Discovery Loop's CEO
Image: Runtimewire (auto-discovered)

The investor's live portfolio page identifies Dean as co-founder and CEO, while Discovery Loop's own site still presents four founders without executive titles.

By RuntimeWire Staff · Published

Primary source: X

Why it matters #

Radical Ventures' live portfolio listing supplies a public CEO title for Jeff Dean while Discovery Loop's own site continues to present all four founders without individual executive roles.

Radical Ventures now lists Jeff Dean as Discovery Loop's CEO, while Discovery Loop's own site still presents four founders without assigning executive titles. Dean founded Discovery Loop, an AI research venture, with Sanjay Ghemawat, Quoc Le and Oriol Vinyals.

Radical Ventures' designation addresses an ambiguity left by Discovery Loop's August 5 launch. When WIRED asked who would run Discovery Loop, Dean d before answering: "I think I'm the CEO". He added that everyone had pointed at him. Radical Ventures, which co-led the initial financing, now gives him that title on its portfolio page.

The title ends the launch-day ambiguity as four veteran Google researchers move from a pitch deck to hiring, infrastructure and experiments. The four founders have spent decades working together, with shared work extending from Google's early distributed systems to AI models and computing infrastructure.

Dean joined Google in 1999 and left after nearly 27 years. He and Ghemawat helped create MapReduce, Bigtable and Spanner, systems that shaped how Google handled computation and data at scale. Dean later led Google Brain and served as Google's chief scientist.

Dean studied computer science and economics at the University of Minnesota, completed a doctorate at the University of Washington and worked on software modeling the AIDS pandemic for the World Health Organization before joining Google.

The founders are treating research as a systems problem

Discovery Loop says it is building AI systems that propose experiments, run them, evaluate the results and choose what to try next. Dean and his cofounders plan to begin with machine-learning research, using Discovery Loop's technology to improve its own models and infrastructure before moving into fields such as chip design, biology, drug discovery and materials science.

The immediate plan remains a buildout rather than a released product. Discovery Loop's website describes systems under development and says the founders are assembling a lean, in-person group. It presents no commercial offering or published scientific result. Dean wrote at launch that the immediate work included finding office space, recruiting the founding staff and constructing the infrastructure required to run large numbers of experiments.

Discovery Loop is trying to build computational infrastructure, AI models, experiment orchestration and eventually products for researchers. Dean's systems background fits that work. A CEO would also have to decide which components Discovery Loop builds, which models it depends on and when an internal research system is ready for outside users.

Ghemawat gives Discovery Loop another systems architect with decades of experience working beside Dean. Le, a founding member of Google Brain, worked on sequence-to-sequence learning, neural architecture search and automated machine-learning methods. Vinyals served as a Google DeepMind vice president of research and technical lead for Gemini, with work including AlphaStar and sequence-to-sequence learning.

The four founders describe themselves as longtime friends as well as collaborators. Investors are backing technical records that would be difficult for an early-stage product to match.

Dean has capital and compute before Discovery Loop has a product

Radical Ventures and Khosla Ventures co-led Discovery Loop's initial funding, with Lightspeed, Kleiner Perkins, Doerr Capital and Alphabet participating. Axios reported that a source described the first round as being in the "hundreds of millions" of dollars; Discovery Loop has not published a round figure or valuation.

Alphabet's involvement extends beyond its investment. Google said it would continue working with Discovery Loop as a founding investor and Cloud partner, including collaboration on a research framework for machine-learning systems and related infrastructure. WIRED reported that the arrangement would provide compute power for Discovery Loop's first year.

The arrangement softens Dean's break from Google after one of the search group's most consequential technical departures. Alphabet remains an investor, Google remains a Cloud partner, and the companies retain a research relationship. Dean gets room to build outside Google without immediately recreating every part of its computing base.

RuntimeWire covered Discovery Loop's launch on August 5 and later examined the investor group behind Dean's plan.

Automated research still has a verification problem

Discovery Loop enters a field where prototypes have already shown both the appeal and fragility of automated experimentation. Sakana AI's AI Scientist, for example, has demonstrated a workflow that generates research ideas, runs experiments and writes papers.

An independent evaluation of that system reported that 42% of its experiments failed because of coding errors. The evaluators also found weak novelty judgments, poorly supported manuscripts and hallucinated numerical results, even as they credited the system's speed and low cost. The study does not evaluate Discovery Loop. It shows a problem any automated experimentation system must address: running thousands of experiments can also multiply errors, duplicated ideas and misleading measurements.

Discovery Loop's founders have extensive experience building infrastructure and AI systems. Discovery Loop will also need methods for choosing useful questions, designing evaluations that measure the intended result and rejecting outputs that look plausible while being wrong.

Radical Ventures' live portfolio listing identifies Dean as Discovery Loop's co-founder and CEO. Discovery Loop's own site continues to identify Dean, Ghemawat, Le and Vinyals as its founding team without publishing individual executive titles.

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @radical ventures 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/radical-ventures-lis…] indexed:0 read:5min 2026-08-31 ·