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

Lenny opens with a blunt claim about organizational design rather than technology: because the underlying models change quickly, what a company can build changes quickly, so the ability to probe the frontier stops being a specialty and becomes a core competency. In his framing the required skills are not coding speed but the loop around it: finding ideas, running many experiments, deciding what to double down on, and building teams that can work that way without pulling the rest of the company into every shiny distraction. The competitive clock is explicit, with startups named as the thing that will eat your lunch if you are slow.

My thesis is that every company needs to start thinking and operating like a Labs team.

Josh Woodward agrees, but immediately moves the difficulty from the slogan to the implementation, and his qualification is the most useful part of the exchange. He observes that a startup is already a lab in some sense, which means the real problem is not inventing the method but transplanting it into medium-sized and large teams. That reframing also splits the work in two: cultivating ideas, and cultivating the people who can thrive in that environment. Lenny accepts this and treats the rest of the conversation as an elaboration of those details.

The staffing test Josh describes is behavioral rather than credential-based. The frontier group is a defined set of people whose actual job is to be early, and his shorthand for that is consumption habits: the people who hold many separate paid subscriptions to the newest tools. Lenny restates the same expectation from the company side, that everyone should become good at trying a pile of things with the latest models and discovering what is now possible ahead of competitors. Read together, the two statements imply a budget and a mandate, not a hackathon.

The structural warning is where Josh is most specific and least optimistic. He argues that a lab needs room for odd, unplanned growth, and that the historical pattern for large companies is a lab that is ineffective and fades after roughly three to four years, usually because it was tucked under an existing business unit. Independence, in his account, is not a status symbol but the condition that determines survival. This is a claim drawn from observation of corporate history rather than a controlled comparison, and he does not offer a threshold for how much independence is enough.

You've got to try to create a space where weird things can grow.

Asked for patterns behind Labs' best ideas, Josh refuses the premise of a repeatable venue. He says good ideas rarely originate in design sprints or in calendar blocks reserved for generating the next idea, and instead surface when people are already hacking, often on weekends or over breaks, then bring back a video of something they built because they could not stop themselves. Lenny notices the omission explicitly: no brainstorm meeting appears anywhere in the answer.

You can't microwave good ideas.

The practical consequence Josh draws is a shift of attention from ideas to people. His third lesson is that the interesting unit is not the concept but the curious person and their side projects, plus the chemistry that forms around them. He backs this with the origin of products he names directly, Notebook, Flow, AI Studio and Google Beam, all of which he says began with very small groups of obsessed people rather than with a formal ideation process. That is a strong claim about origins, though it is retrospective and filtered through the products that survived.

The one mechanism that does look designed is a maintained list of what the team calls almost possible. The discipline is to stay at the frontier, watch which experiences are just out of reach, and notice the moment a capability crosses over, which Josh describes as a phase shift. When something on the list becomes real, the response is to assemble a team around it quickly. He declines, at this point, to say what is currently on the list, so the method is described but not demonstrated with live examples.

Technology alone is explicitly insufficient in his account. The second half of the filter is an interesting user problem, and the third is whether the combination is good enough that people will pay. He also generalizes the method to founders pitching investors: track the almost-possible, then move when the constraint breaks. The sequence matters because it inverts the usual pitch order, starting from a capability and then searching for a problem worth attaching it to.

To make that search concrete, Josh describes thought experiments about the future of software development, creativity, work and knowledge, and says entertainment is currently under close watch. The technical triggers he lists are latency, language coverage and modality, each treated as a threshold that unlocks new product shapes. Alongside the technology, he says Labs studies behavior and trends, and deliberately looks beyond the Bay Area to other cities and classrooms as places to mine for behavioral change.