Lenny's Podcast: Product | Career | Growth
Why every company now needs to think and operate like a lab team | Josh Woodward (Google Labs, Gemini, AI Studio)
This debrief was written and voiced by AI from the episode's audio.
🎧 AI Debrief
🎙️ Full Episode
Every company needs a frontier team now
Josh articulates the core thesis: companies must dedicate teams to experimenting with the latest AI models because what's possible changes weekly. Without this, organizations risk disruption.
Key Moments
Josh articulates the core thesis: companies must dedicate teams to experimenting with the latest AI models because what's possible changes weekly. Without this, organizations risk disruption.
Ideas rarely emerge from design sprints or scheduled brainstorms. Instead, they come from obsessed people building in hallways, on weekends, over breaks. You can't microwave good ideas.
Josh shares a concrete framework: maintain a list of what's almost possible, then watch for the moment something crosses into possible. That's your signal to launch a project.
Forget DAU metrics in early stages. When showing prototypes, watch if people's eyes light up and they lean in. That visceral signal beats data in detecting early PMF.
Teams sense failure before leaders do. Create psychological safety so teams can voice when projects aren't working. Trust them, and move on quickly. Josh shared an example of a PM convincing him to kill a feature because the data showed it wasn't resonating.
The most valuable skills now are how fast people can learn and walk away from ideas, ability to go hard but sustain pace long-term, and capacity to build trust across teams and with AI agents.
Most corporate labs fizzle after three to four years. The problems: lack of independence, inability to commercialize findings, or failure to bridge breakthroughs back to the core business.
One: create independence so weird ideas can survive. Two: clarify your mission (graduate features or invent categories?). Three: use labs to challenge how the company works, not just products.
Look for people who can't stop building, intellectually curious, energy-givers. On interviews, ask how they've let go of ideas. People comfortable in unknown caves signal zero-to-one capability.
Josh suggests the question itself might become obsolete. Instead of counting products, we might think about bespoke experiences powered by AI agents, delivered through common surfaces.
Search Episode
Summary
Josh Woodward, head of Google Labs, argues that every company must now operate like a lab team because AI capabilities are changing so rapidly. The core thesis is that organizations need dedicated frontier teams experimenting with the latest models to discover new possibilities before competitors do. Woodward shares hard-won lessons from running one of the world's longest-standing labs: the best ideas rarely come from design sprints but emerge organically from obsessed people building on weekends; product-market fit is detected through observation (watching people's eyes light up) rather than metrics; and the hardest part is knowing when to kill ideas—teams usually sense failure before leaders do. He emphasizes that labs must have structural independence, a clear mission (whether graduating features or inventing new categories), and a culture where people prioritize users over their own ego. Woodward also discusses how roles are blurring but specialties still matter, the importance of "explosive endurance" to sustain pace over time, and why principles and vision matter more than model benchmarks in building products people actually want.
Key Takeaways
Every company needs a frontier team explicitly tasked with experimenting at the edge of what's possible with AI, staying on top of the latest releases and capabilities before competitors do.
highThe best product ideas emerge from obsessed individuals building in their spare time, not from formal brainstorms or design sprints. Watch for people who can't stop themselves from building.
highProduct-market fit in early stages is detected by watching people's eyes and body language during prototype demos, not by OKRs or DAU metrics. If their eyes light up and they lean in, you have a signal.
highTeams know when an idea isn't working before leaders do. Create psychological safety so teams can voice when projects are failing, then trust them and pivot quickly.
highLabs must have structural independence and cannot nest under existing business units, or they will be starved of resources and slowly absorbed into legacy priorities.
highThe most valuable emerging skill is 'unlearning rate'—how fast someone can learn something, test it, and walk away based on feedback. This matters more than traditional learning velocity in a fast-changing landscape.
mediumRoles are blurring but specialties still matter. The best builders have a major (their core expertise) and minor (adjacent skills), and should develop their major first, not treat everything equally.
mediumNotable Quotes
“You can't microwave good ideas.”
— Josh Woodward
When discussing where breakthrough ideas come from. Ideas don't emerge from scheduled brainstorms but from sustained obsession and exploration. This is a memorable way to reinforce that shortcuts don't work for innovation.
“When you're showing people early prototypes, you're looking at people's eyes. That is the metric.”
— Josh Woodward
On how to detect product-market fit. Rather than relying on data or analytics in early stages, observe emotional and physical responses—pupil dilation, leaning in, eye contact. It's visceral and accurate.
“The team usually knows before the leader knows.”
— Josh Woodward
On recognizing when to kill a failing project. Teams working daily on a product sense loss of momentum before executives do. Leaders should create safety for teams to voice these concerns.
“Most people in the world are never ever going to care about an Elo rating.”
— Josh Woodward
On why model benchmarks are overhyped. A direct, memorable critique of the AI industry's focus on abstract metrics over real user value.
“Labs creates good trouble at Google.”
— Josh Woodward
On the role of frontier teams. They don't just build products; they challenge organizational assumptions and processes, piloting new ways of working before broader rollout.
Predictions
Super apps will consolidate somewhat, but over-consolidation creates complexity that drives people away. There will likely be a middle ground with a few main surfaces and more bespoke, agent-driven experiences behind them.
— Josh Woodward · Confidence: medium · 5 years
Entertainment and immersive experiences powered by AI will unlock new consumer opportunities in ways not yet clear. The form factor (chatbot, agent, etc.) is still being invented.
— Josh Woodward · Confidence: medium · 2-3 years
Personal intelligence features—AI that is personal, proactive, and powerful—will become Google's next major product push, integrating Gmail, Calendar, and Docs data seamlessly.
— Josh Woodward · Confidence: high · 6-12 months
Messaging and chat spaces, once considered solved, are being reinvented with AI. New interaction patterns will emerge inside and outside Google.
— Josh Woodward · Confidence: medium · 1-2 years
Action Items
If you lead a company or team, establish a dedicated frontier team with explicit permission to experiment with the latest AI models and tools. Give them independence from existing business units and make it their job to surface emerging possibilities.
Josh emphasizes that labs must be structurally independent or they will fail. This is the most actionable step for larger organizations to avoid disruption.
Build a culture where people can voice when ideas aren't working without fear. Celebrate the PMs or leads who kill projects early with conviction.
Psychological safety around failure and honest feedback is what separates successful labs from ones that waste resources on zombie projects.
Create a 'what is almost possible' list and track it obsessively. When something crosses from almost possible to possible, it becomes a project candidate. Monitor your industry, X, model releases, and user behavior shifts.
This is a concrete framework Josh shared for staying at the frontier without waiting for inspiration to strike.
Hire for unlearning rate and explosive endurance, not just learning rate. Ask interview candidates how they've let go of ideas, pivoted, and sustained effort over long projects.
These are emerging skills that predict success in fast-changing environments better than traditional proxies.
Clarify your lab's mission early: Are you graduating innovations into existing products, or inventing new categories? Communicate this so the team optimizes correctly.
Misalignment on what winning looks like causes labs to fail. Different mission = different hiring, metrics, and timeline expectations.
Create lightweight rituals and traditions (like Josh's TPU Harvester award, LLM Whisperer award) that encode the behaviors you want to reinforce. Celebrate the unglamorous work that enables breakthrough products.
These rituals make values concrete and memorable. They also surprise and delight high performers in a way generic recognition doesn't.