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

1. Key Themes

Every company needs a "frontier team," but the real failure mode is commercialization, not invention

Lenny's premise is that the underlying tech changes so fast that every company must get good at testing the latest models before competitors do. Woodward agrees, but warns that the hard part is bridging the lab to the core business. He cites the historical record: "if you just look at the data, a large company who starts a lab, the lab is usually ineffective and fizzles out after about 3 to 4 years. Um, or they have duration, but the company can never commercialize the things coming out of the lab." 00:44:08 His prescription is a team that lives at the frontier but can also tell the main company "that thing looks like a toy right now, that's going to be essential to our future in 5 years." 00:44:08

Great ideas can't be scheduled: they come from obsessed people with side projects

Woodward says the best Labs products (NotebookLM, Flow, AI Studio, Google Beam) started with tiny groups of obsessives, not process. "They very rarely, at least in my experience, have come from design sprints." 00:05:58 "They usually come when people are swimming, in the hallway, maybe on a weekend, often over a break." 00:06:26 Then: "You can't microwave good ideas." 00:00:49 The leadership job is to "create a space where that can happen and then kind of be on alert to like catch it and try to like, you know, funnel it forward." 00:06:53

Track "what's almost possible" and act when something crosses the threshold

Labs keeps an explicit list of capabilities that are nearly feasible, then pounces when one flips. "We talk about what's almost possible. We just have a list, actually, where we're trying to track those things. And when one crosses over, it's almost like a phase shift... you're like, 'Ooh, that just became possible.'" 00:08:00 They pair that with point-of-view forecasting: "we have, um, 82 predictions. They all begin with, 'We believe the future is', or, 'We predict.'... you almost have to have a point of view about the future, knowing that nine out of 10 or maybe 9.9 out of 10 you're going to be wrong." 00:11:19

Product-market fit at the 0-to-1 stage is read in people's eyes, not dashboards

"When you're showing people early prototypes... you're looking at people's eyes. And like, that is the metric. It's like, uh, it is not a DAU, it is not a DAU over MAU, it's not a D7, it's none of that." 00:01:10 The stage-appropriate metrics shift as products grow: "in 0 to 1, you're looking at eyeballs... And maybe 10 to 100, you're looking at like waterfalls, and CPAs." 00:46:26 Labs even formalizes this as "0 to 1 Labs, 1 to 10 Labs, and 10 to 100." 00:46:44

Fall in love with the problem, not the product, because it takes 3-5 pivots

"What's the problem we're falling in love with, not the product, because what I found is it takes 3, 4, 5 pivots before you either got something or you're just like, 'Ta, that's it. Didn't work.'" 00:15:58 He adds that when "people fall in love with an actual product solution first... it's going to, at least in my experience, lead to a bad time." 00:16:35

The team knows before the leader that an idea should die

"The team usually knows before the leader knows. And the signs are like, does the passion start to run out?" 00:01:19 Woodward gave a live example: a Gemini feature the team was excited about tested poorly, and the PM said "I don't think we should launch this. It's not good." His response was to "reply all" with thanks. 00:18:24 The cultural lesson is to create an environment where people can raise uncomfortable truths.

Skills in rising demand: unlearning rate, explosive endurance, and trust-building

Woodward looks for "mispriced signals": "What's their unlearning rate? Like, how fast can they learn something and then walk away from it?" 00:31:53 He also wants "explosive endurance" (from a Roger Federer book): people who can "go really hard... but they can do it over a while." 00:31:53 And he thinks collaboration rises in value: "people that can build trust, that can scale trust, that underpins any kind of great collaboration." 00:31:53

Teams are shrinking and becoming "free agents jamming," but specialties still matter

Labs teams went from "five to seven people. Now it's about two to three people, uh, maybe four." 00:34:29 He uses a music metaphor: "super talented musicians who team up with different people and then they play different songs together. The songs are the products." 00:34:29 But he rejects the "everyone is just a builder" narrative: "people still have their specialty, it's like their major or minor in college." 00:36:15

Planning horizons have collapsed to rolling ~6 months, with 50-100 day milestones

"We usually are thinking... somewhere on the order of about 6 months, and usually I say that because that's that almost possible window where... you assume every lab is cranking out some new train- pre-training model a couple of times a year." 00:42:45 Best Labs teams "go from like idea to some meaningful milestone in like 50 to 100 days." 00:42:45 On Gemini, a request for a 2-year roadmap got: "we have a 6-month roadmap on this team." 00:41:46

Product simplification is the direction: one prompt box, not modes and toggles

On why Google's assistant isn't yet doing everything Lenny wants, Woodward says "the lesson even from the last couple of weeks is forget the modes and toggles. Just like, one prompt box, tell it what you want, and it just does stuff for you." 00:25:19 He frames Gemini's direction as "personal, proactive, powerful," and says a lot is "coming soon."

2. Contrarian Perspectives

Model benchmarks and Elo ratings are overhyped

Despite running Gemini, Woodward says what's overhyped is "model benchmarks... most people in the world are never, ever, ever going to care about an Elo rating. They may not even know how to pronounce Elo." 00:28:28 He argues the industry "over-rotate[s] to that, and kind of lose[s] sight of like, why is this useful to someone?" 00:28:51 A model lab leader calling its own scoreboard a distraction is notable.

"Everyone is a builder" is overhyped; specialties become more important, not less

Against the popular narrative of blurring roles: "I think right now, um, people are maybe overhyping like, everybody's a builder and it's just this new job function and everything's kind of the same... there's a risk if everybody becomes a builder that you lose a lot of respect and appreciation, and obviously the expertise that comes from those specialties." 00:36:15 He still advises new PMs to vibe-code "in service of becoming an exceptional, extraordinary PM. And like, don't lose that major." 00:38:11

Collaboration with humans gets more valuable as agents get better

Conventional wisdom says agents reduce the need to work with people. Woodward says "as much as coding and other things are becoming... trending towards like, you being able to direct lots of agents and not having to work with as many people, I still feel like collaboration and like how you work with people matters even more now." 00:31:53 His reasoning: lower execution costs mean more products and more team reshuffling, so trust-building speed becomes the bottleneck.

Maybe Google won't have 10,000 products, or five, because "products" may not be the unit

Asked whether Google will have 10,000 products or five in five years: "I don't know if we're even going to think of products the same way... are there almost like experiences or things that are going to be more bespoke that the models might make?" 00:54:23 He admits he flips his answer and that Labs is running experiments "to prove both sides of this debate." 00:54:23 He also notes that super-apps "can be very complex" and that this "assum[es] humans are going to be the ones tapping most of these experiences." 00:54:23

Hiring fast is a vanity metric, and consolidating teams too early is a mistake

"You hire too many people too fast... it can be so addicting to like, sometimes look at like, 'Oh, our team doubled or tripled in the last quarter or 6 months,' or whatever, and these are just vanity metrics that are really bad most of the time." 00:52:30 His own example: "we got to like 30 engineers and I was so happy, and we had zero product-market fit, but a very inspiring vision." 00:52:30 On org design, "sometimes you kind of have to live in the messiness a little bit and almost like see what emerges" rather than rushing to consolidate. 00:52:41

3. Companies Identified

NotebookLM (Google Labs)

Google's AI notebook product, known for its AI-generated podcasts, which grew out of Labs. Mentioned as the canonical example of an idea recognized from a "wow" moment. Woodward: "I remember still the first time I heard the first like NotebookLM podcast... it was the proceedings of the British Parliament debates. Um, not known for its riveting material, but these two AI hosts were talking about it in a way that was was literally interesting." 00:13:28 Also cited as a product-market-fit moment: "when the NotebookLM podcasts have gone... you're literally just swarming it and holding on for dear life." 00:16:35

Google Flow and Whisk (Google Labs)

Creative AI tools from Labs. Whisk was a predecessor letting users "whisk together multiple images and then start to animate them." Woodward: "it was a level of control and creative kind of channeling that had never been possible." 00:14:00 Flow is also listed among products that have "kind of grown out of Labs."

Nano Banana (Gemini)

Google's image generation/editing capability inside Gemini. Cited as a product-market-fit moment: "when Nano Banana went viral on Gemini..." 00:16:35

Google AI Studio

Google's developer-facing model playground, now one of the larger Labs-originated products. Listed among projects started by "very small groups of people who are just obsessed." 00:06:53

Google Beam

A Labs project Woodward calls "this incredible kind of magic mirror, they kind of can create like a 3D version... where you can kind of talk to someone almost like they're a hologram." 00:21:40 He frames it as adjacent to AI hardware.

Stitch (Google Labs)

Named alongside Flow as a Labs project that has had viral moments. 00:16:35

Gemini app and Gemini Spark

Google's flagship assistant, "over a billion users." On connecting personal data, Woodward says "if you're in Gemini and Gemini Spark, you can actually connect all that stuff and it's quite powerful." 00:25:19 The direction is "personal intelligence," a Gemini that is "personal, proactive, powerful." 00:26:22

WorkOS (sponsor)

Enterprise-readiness infrastructure (SSO, SCIM, RBAC, audit logs). The ad narrator says it powers "OpenAI, Anthropic, Cursor, Replit, Sierra, Clay," and calls it "essentially Stripe for enterprise features." 00:05:13

DX (sponsor)

Developer-productivity platform for measuring AI adoption impact. Ad: "hundreds of enterprises, including Snowflake, Sony, and BNY, use the DX platform to measure AI's impact on developer productivity." 00:31:06

Anthropic

Mentioned by Lenny as an example that even the most frontier companies have a Labs team: "even Anthropic has a Labs team. Like, the most cutting edge company... they're like, 'Okay, but we also need a Labs team to actually watch the frontier frontier.'" 00:46:12

OpenAI

Referenced via Lenny's earlier conversation with Tara from OpenAI about deciding what to build now that "we could just build everything." 00:30:17

Facebook

Mentioned by Lenny as one of the "juggernauts" that once made consumer feel dead. 00:22:22

Google AI Mode

Cited by Lenny as an example of Google's distribution advantage: "once something happens... it's going to be very easy to grow it and convince people to use it, similar... within like AI mode." 00:26:51

Muse

Referenced by Lenny as a recent launch showing Google's distribution leverage. 00:26:51

Lenny's Most Replayed Moments (YouTube channel)

Lenny's new channel for the most rewatched segments of long-form episodes. 00:01:56

4. People Identified

Josh Woodward

Head of Google Labs, the Gemini app, and AI Studio; 16+ years at Google. Lenny: "Josh runs the longest-lasting and biggest labs team in the world." 00:01:56 His operating signature is user-first prioritization: "users first, Google second, and like our product third." 00:20:51

Logan Kilpatrick

Head of AI Studio at Google and former podcast guest; suggested questions for the episode. Woodward: "He's great." 00:54:08 His "10,000 products or five?" question is a running debate with Woodward. 00:54:23

Dan Shipper

Gave a talk at the Lenny and Friends summit advocating a dedicated frontier team. Lenny: "His advice is to have a frontier team that's dedicated to testing out the new stuff. And, basically, it's their job, try all the new stuff, report on what we need to pay attention to so that the rest of the company can not have to be distracted." 00:43:38 Woodward endorsed it, adding his lab-history caveat.

Roger Federer

Tennis great whose "explosive endurance" Woodward uses as a hiring and team-pacing metaphor. Woodward: "Federer had explosive endurance. It's just this beautiful turn of phrase." 00:31:53

Tara (OpenAI)

Lenny's earlier guest; their conversation about deciding what to build, not just what can be built, was echoed here. 00:30:42

Steven Johnson, Sophie Miller, Jeff Wang

Credited by Lenny for suggesting topics and questions for the conversation. 00:02:16

The unnamed Gemini PM

A PM who recommended killing a Gemini feature after early user testing. Woodward: "I was so proud of the PM on the team, because she was like, 'I don't think we should launch this. It's not good.'" 00:19:06

5. Operating Insights

Name the mode: declare when a team is in "explosive" sprint and when it isn't

Woodward's best leadership lesson on pacing: "maybe the best lesson I've learned as a leader is almost to sort of name that, declare that. You know, people opt in or opt out of that. Um, but also be very clear when a team is not in that mode." 00:39:07 He builds a rhythm around model launches: "We've got a big model coming. These five products are super important to showcase this model... it is explosive time." 00:39:07

Schedule a deliberate "hack month" after big launches to seed the next cycle

After Google I/O, Labs teams are told to just build: "June is like, 'Hey, just go hack. Go rediscover what's new.'" He warns the instinct is to see it as unproductive, but "those are where all the new seeds are coming from for like the next season." 00:39:07

Use the lab to pilot company-wide change, such as job ladders

"Labs sometimes creates good trouble at Google... We were one of the first teams that were like, 'We're seeing the job roles blur. We have to have a builder job ladder... we should use Labs as the place to pilot it before we roll it out across Google.'" 00:48:02

Use symbolic awards to encode what the team values

Labs gives physical awards tied to strategic behaviors: a rake for the "TPU Harvester" who reclaims compute ("people that optimize"), a golden band-aid for fixing product paper cuts, and giant ears plus a mechanical keyboard for the "LLM Whisperer." 00:58:33 Woodward: "we're trying to find like, ways that kind of like... encode the value in these like, dorky symbols." 00:59:08

Run a "good meeting" appreciation ritual

A calendar invite with 4-5 leaders and no agenda, where each person says why they appreciate the recipient. Woodward opens with "This is not a bad meeting, this is a good meeting." 01:00:52 It is a retention tool for the most valuable people.

6. Overlooked Insights

Gemini's dominant product risk is user reluctance to hand personal data to competitors, not Google's inability to build

Almost in passing, Woodward reveals the demand signal: "one of the main things uh people will ping me about or others is like, 'Oh man, Google, I've got all my stuff with you. Just make the thing work.' Um, and there's a lot of reluctance actually to be connecting it to a lot of other sort of competitor products." 00:25:50 This hints that users trusting their email, calendar, and docs to Google is a durable moat for a personal agent, and that third-party agent startups built on Gmail/Calendar data face a distribution and trust ceiling Google can overcome once it ships.

The "0-to-1 crowd-out" problem: successful products starve the experimental ones

Woodward notes in a single aside that as Labs graduated hits like Notebook, Flow, and AI Studio, "there's a risk they crowd out the 0 to 1 stuff." 00:46:26 This is the same mechanism that kills innovation labs elsewhere. Success creates demands on talent, attention, and compute (compute scarcity is why they reward "TPU Harvesters"), and the fix was to explicitly stage Labs as 0-to-1, 1-to-10, and 10-to-100 with distinct metrics. Any company copying the "lab" model needs to plan for its own hits cannibalizing its pipeline.