Why Every Company Needs to Think and Operate Like a Labs Team

The Core Thesis

Lenny opens with a premise that Josh Woodward largely endorses: because the underlying technology is changing so quickly, every company is being forced to operate like a labs team. The capabilities of AI models shift constantly, so organizations must get very good at experimenting with the latest models and discovering what is newly possible before competitors or startups do. Josh agrees with the premise but cautions that “the devil’s in the details.” In his view, every startup is already a lab in some sense, and the real challenge is bringing that methodology, energy, and excitement into bigger and medium-sized teams, then building an environment that nurtures both ideas and the people who can thrive in it.

Where the Best Ideas Come From

Josh is clear that great ideas rarely emerge from design sprints or scheduled brainstorms. “You can’t microwave good ideas.” Instead, they tend to surface when people are tinkering in the hallway, over a weekend, or during a break, often as side projects they cannot stop themselves from building. Three lessons stand out from his answer:

• Stay on alert. Always be watching for something that catches your eye.
• Ideas do not come from conventional places. Notebook, Flow, AI Studio, and Google Beam all began as small groups of intensely curious people who would not stop.
• Focus on the people and their passion projects more than the idea itself, and create the chemistry that lets those projects grow and be funneled forward.

The “Almost Possible” Framework

A central tool in Google Labs is a running list of things that are “almost possible.” The team tracks capabilities that are on the verge of becoming feasible, and when one crosses over, Josh describes it as “almost like a phase shift,” like a substance moving from solid to gas. At that moment the team tries to assemble a group around it and move quickly. He pairs this with a thought experiment about where a domain is heading, such as the future of software development, creativity, work, or entertainment. For immersive entertainment, the team reasons about what latency, language coverage, and modality a model would need before certain experiences become viable. Josh also stresses looking well beyond the Bay Area for behavioral trends, mining other cities and classrooms for signals.

He acknowledges that the team makes 82 predictions in one internal doc, all beginning with “we believe” or “we predict.” He admits that making even one prediction is dangerous and that most will be wrong, perhaps nine out of ten. The point is to have a point of view and to chase the future, living in it a little.

Recognizing Signals and Product-Market Fit

Josh admits that finding product-market fit is “way more art than science.” Teams know how to run experiments, make data-driven decisions, and set OKRs, but in early days a few people often have conviction about a future they cannot fully explain. For early prototypes, the metric he watches is people’s eyes: “That is the metric.” He is not looking at DAU over MAU or D7 retention at that stage. If someone leans in and their eyes light up, that is the signal. Josh notes that the best feedback comes when a product hits a nerve in someone’s life, a pain point they immediately want solved. He also warns against falling in love with a product solution rather than the underlying problem, since it often takes three, four, or five pivots before something works.

When something does work, such as Nano Banana going viral on Gemini or the NotebookLM podcasts, the team swarms the opportunity, which he describes as “both so exhausting and so fun.”

Knowing When to Kill an Idea

Josh’s view here is that “the team usually knows before the leader knows.” Signs include passion running out and having tried everything without success. He emphasizes creating a culture where people can ask uncomfortable questions and say that something is not working. He cites a recent Gemini feature that the team was excited about, but early data showed it lacked the spark. A product manager argued against launching it, and Josh publicly thanked her. He says he trusts the team because they are the ones pushing every day, and if they say a project is not going, the answer is to pivot.

Labs Within a Massive Product Organization

The conversation explores why Josh oversees both Google Labs, which focuses on zero-to-one experiments, and the Gemini app, which serves well over a billion users. Josh says the people drawn to both share a zero-to-one heartbeat. The shared mantra is “users first, Google second, and our product third,” which keeps the teams focused on whether an end user is loving the product. Labs is now entirely AI-focused, though the team explores adjacent hardware such as Google Beam, the 3D video communication project.

Consumer AI, Personal Assistants, and Data Access

On consumer AI, Josh rejects the notion that the space is closed. He points to entertainment and feeds as an open question, to messaging and chat as a space that is newly unfixed, and to scarce resources such as time, money, and shared real-world experiences as areas where new products might be built. He describes the future interface as moving away from modes and toggles toward a single prompt box where the user states what they want.

When Lenny raises the question of why Google has not connected its own data more fully into assistant experiences, Josh says some of this exists today through Gemini Spark. He acknowledges that users often ask why their Google data is not simply put to work, and notes reluctance among users to connect that data to competitor products. The vision he articulates is a Gemini that is personal, proactive, and powerful, which he frames under the broader concept of personal intelligence.

Overhyped and Underhyped in AI

Josh’s underhyped pick is discussion of principles and values. He argues that whether one works at Google or a startup, products “speak through” and encode their principles, and he wishes the industry discussed what kind of future it wants to build more often. His overhyped pick is model benchmarks. Most people will never care about an ELO rating, and the industry risks over-rotating on benchmarks while losing sight of the user value that products should deliver. Lenny adds that each model launch is hailed as AGI, yet in practice the differences are often subtle.

Skills Rising in Value

Josh identifies several attributes he is looking for:

• Unlearning rate: how fast someone can learn something and then walk away from it, based on what they see when testing a product. He values this above a pure learning rate or growth mindset.
• Explosive endurance: a phrase borrowed from a book about Roger Federer, describing the ability to go very hard in bursts while sustaining effort over time. Labs organizes work into seasons to support this.
• Trust and collaboration: even as agents multiply what one person can direct, working with people matters more. The ability to build and scale trust is a key signal.

He also notes that teams for new labs projects have shrunk from five to seven people to two to three, or perhaps four, because execution costs are falling. Strong contributors act like talented musicians who team up in different combinations, and the most valuable people can flex across roles.

Roles Are Not Actually Blurring Into One Builder

Josh pushes back on the hype that everyone is becoming a builder. Product managers adapt more readily because their work is more horizontal and they are accustomed to simulating other functions. However, Labs has seen strategy, operations, business development, and marketing people become strong builders. The key point is that people still have specialties, a kind of major and minor. He worries that if everyone is labeled a builder, the respect and expertise that come from specialties will erode. He advises new PMs to vibe code and prototype in service of becoming exceptional PMs rather than abandoning their core craft.

Explosive Endurance and Rhythms

Josh describes labs as operating on seasons, with twenty or thirty experiments running at once. Each team sets milestones and a rhythm, and he has learned to name the mode a team is in, making clear who is opting in or out and when a team is not in a sprint. He points to the period after Google I/O, when teams are encouraged to hack and rediscover what is new. The temptation is to see June as unproductive, but he argues that those are the moments when new seeds form. He also warns that having too little to do, stress, or pressure causes people to get bored and look elsewhere. The long-term aim is to sustain people for years, not to empty the tank in nine months.

Planning and Timelines

Labs typically plans on roughly a six-month horizon, aligned with the cadence of model releases. The best teams go from idea to a meaningful milestone in fifty to one hundred days, and the approach relies on unblocking work so people can cook. Josh describes this as rolling planning. He recounts being asked for a two-year Gemini roadmap and replying that he has a six-month roadmap and a multi-year vision, because a project he saw two days earlier could be built in forty-eight hours.

Frontier Teams and the Lab Lifecycle

In discussing Dan Shipper’s Lenny and Friends Summit talk about dedicated frontier teams, Josh agrees that every company needs a team at the frontier that subscribes to every model and tracks the latest viral developments. He warns, however, that history shows large-company labs often fizzle out after three to four years or fail to commercialize their work. The solution requires a team that can bridge into the broader company and recognize when something that looks like a toy will become essential in five years.

Josh also describes a lifecycle Labs uses internally: zero-to-one labs, one-to-ten labs, and ten-to-hundred labs. Each stage has different challenges and metrics. Zero-to-one focuses on eyeballs, while ten-to-hundred emphasizes conversion, retention curves, and scaling. Successful products can crowd out experimentation, so the company intentionally maintains distinct stages.

Setting Up an Internal Labs Team

Josh offers three tips:

1. Create a space where weird things can grow, with independence rather than nesting under an existing business unit. In some companies this means reporting directly to the CEO.
2. Define the lab’s purpose clearly. Labs’ mission is to make users and Google successful rather than itself, and leaders must decide whether the goal is graduating projects into existing products or inventing new categories. Labs began as a graduation lab but expanded as AI blurred categories.
3. Use the lab to challenge how the company works. Labs creates “good trouble,” such as piloting a builder job ladder before rolling it out across Google.

On hiring, Josh looks for people who cannot stop building, are intellectually curious, solve user problems rather than chase ego, and add energy to a room. He notes that the zero-to-one, one-to-ten, and ten-to-hundred stages each require somewhat different people.

He also cautions against common org mistakes: rushing to consolidate sub-teams too early rather than living in the messiness, and hiring too many people too fast. He cites a project that reached thirty engineers with a compelling vision but no product-market fit.

The Future of Products

Responding to a question from Logan Kilpatrick, Josh says he is torn over whether Google will have 10,000 products or five in five years. He suggests that the concept of a product may change, with bespoke experiences generated by models and supported by labs teams building agents behind common surfaces. He and Logan sometimes argue both sides as a labs exercise: one scenario with consolidation into super apps and another with fragmentation. He currently leans toward the view that super apps have worked in some places but can be too complex for many people.

Culture, Rituals, and Closing Thoughts

Josh closes by asking listeners for feedback on the products they build. He is interested in how AI can aggregate and act on feedback loops, and he suggests that the best PMs have insight into their users that AI should try to replicate or augment. The lightning round reveals Labs’ playful culture, including the TPU Harvester award, which is a miniature rake for people who optimize and reclaim compute, a golden Band-Aid for fixing paper cuts, and an LLM whisperer award involving a mechanical keyboard. He also describes a “thank you” ritual in which leaders are invited to a meeting with no agenda and the team takes turns explaining why they value each person. Josh frames the purpose of these rituals as encoding value in simple symbols that signal what matters.

• Tags:
• Product Management
• Gemini
• Innovation
• Product Strategy
• NotebookLM
• AI Product Development
• Team Culture
• Product-Market Fit
• Experimentation
• Google Labs
• Internal Labs
• Frontier Teams