Why every company now needs to think and operate like a lab team | Josh Woodward (VP Google Labs)

Summary · 5 min read instead of 1 h 2 m

Josh Woodward leads Google Labs, the Gemini app, and AI Studio, and his job is to find, experiment with, and scale new AI products. Labs, which Rachitsky calls the longest-lasting and biggest labs team in the world, has produced NotebookLM, Flow, Project Genie and Google Beam.

Woodward agrees with Rachitsky's thesis that fast-changing models force every company to operate like a labs team, but says the hard part is building an environment where such teams survive, since most corporate labs fizzle within three to four years or fail to commercialize their work. He says good ideas come from small obsessed groups rather than design sprints, and that Labs keeps a list of what is almost possible so it can act when something crosses over. Early product-market fit, in his view, shows in people's eyes rather than in usage metrics, and teams usually know before leaders when to kill a project. He values unlearning rate, explosive endurance and the ability to build trust, and argues that roles are not collapsing into a single builder job.

Minute by minute

1. 5:34

   Good ideas rarely come from sprints. Woodward says the best ideas have rarely come from design sprints or scheduled ideation, but from people hacking on weekends or over breaks. "You can't microwave good ideas," he says. Projects such as Notebook, Flow, AI Studio and Google Beam started with very small, curious groups obsessed with a side project, so leaders should create space for that and be on alert to catch it.

2. 7:55

   Track what is almost possible. Labs keeps a list of things that are almost possible and moves a team onto one when it crosses over, which Woodward likens to a phase shift. A document holds 82 predictions about the future, knowing most will be wrong. One area he is watching is immersive entertainment, and he pushes teams to look for behavior changes outside the Bay Area.

3. 11:45

   NotebookLM began with a surprising demo. Woodward looks for something that surprises him, then checks it against user behavior and whether some group will pay because they cannot live without it. He recalls first hearing a NotebookLM podcast in which two AI hosts made British Parliament debates interesting. He also cites Whisk, a precursor to Flow, and notes that many exciting ideas still produce products that are not good.

4. 14:32

   Product-market fit shows in people's eyes. Woodward says early product-market fit is more art than science and sometimes rests on a few people's conviction. When showing prototypes, he watches whether people's eyes light up or they lean in, saying "that is the metric" rather than daily or weekly usage figures. He urges teams to fall in love with the problem, not the product, because it takes three, four or five pivots to get somewhere.

5. 17:50

   Teams know before leaders when to quit. Woodward says the team usually knows a project is failing before the leader does, as passion runs out and nothing works. Leaders should make it safe to say so. He describes a Gemini feature the team was excited about whose PM recommended not launching after early data disappointed, and says he replied all to thank her.

6. 19:29

   Labs and Gemini share a zero-to-one culture. Although Gemini has over a billion users and Labs celebrates 10,000 users, Woodward says both attract the same zero-to-one builders who love AI. Both follow a priority order of users first, Google second and their own product third. He says Labs is now entirely AI-driven, including adjacent hardware such as Google Beam.

7. 22:15

   Consumer AI is wide open again. Woodward sees opportunity in entertainment, in a messaging space that once seemed solved, and in products addressing scarce things like time, money and shared real-world experiences. Asked why Google lags rivals that tap Gmail and Calendar data, he points to Gemini Spark, says the industry is dropping modes and toggles for a single prompt box, and describes a vision of a personal, proactive, powerful Gemini.

8. 27:08

   Principles underhyped, benchmarks overhyped. Woodward says companies speak through their products, which encode their principles, and wishes the industry discussed what future it wants to build. He calls model benchmarks overhyped, arguing most people will never care about an ELO rating and that the industry over-rotates to speeds and feeds instead of usefulness.

9. 31:32

   Unlearning rate, endurance and trust rising. Woodward looks for people's unlearning rate, meaning how fast they drop what they learned when tests disagree. He borrows explosive endurance from a book about Roger Federer to describe going hard while pacing oneself. He also values scaling trust, since Labs teams have shrunk from five to seven people to two to three, and expects more people teaming up on more products.

10. 35:39

    Roles are not merging into builders. Woodward says PMs adapt to other functions fastest, perhaps because they already simulate lawyers or sales in meetings, but strong builders come from strategy, ops and marketing too. He calls the idea that everyone becomes a single builder role overhyped, saying people keep a specialty like a college major, and tells junior PMs to vibe code in service of becoming exceptional PMs.

11. 38:54

    Name the season the team is in. Labs runs work in seasons, explicitly declaring when a team is in an all-out push, such as before a major model launch, and when it is not. After Google I/O, June is for hacking and rediscovery, which Woodward says plants seeds for the next season. Gemini keeps a six-month roadmap within a multi-year vision, and the best teams reach a meaningful milestone in 50 to 100 days.

12. 43:24

    Corporate labs usually fail at commercialization. Asked about Dan Shipper's frontier-team idea, Woodward agrees companies need people at the frontier, but says most corporate labs fizzle after three to four years or never commercialize their output. Labs must bridge to the wider company. Internally, Labs distinguishes zero-to-one, one-to-10 and 10-to-100 stages, which optimize for different things, from eyeballs to conversion.

13. 47:27

    Give the lab independence and a clear game. Woodward's tips: give the lab independence, possibly reporting to the CEO; define whether it graduates ideas into existing products or invents new categories; and use it to create "good trouble" in how the company works, as Labs did by piloting a builder job ladder. He warns against consolidating teams too early and hiring too fast, recalling a project with 30 engineers and zero product-market fit.

14. 53:52

    Products may become bespoke experiences. On Logan Kilpatrick's question of whether Google will have 10,000 products or five in five years, Woodward says he keeps switching but now doubts products will be thought of the same way, leaning toward bespoke, model-made experiences. He also wants AI to improve product feedback loops, and describes awards like a rake for TPU harvesters and a golden band-aid for fixing paper cuts.

Published Oct 11, 2026.