How Google Labs Builds AI Products from 0 to 1

TL;DR

Google Labs, Google's experimental arm relaunched about three years ago, ships new AI products in 50 to 100 days by staffing small four-to-five-person teams that solve one user's problem first. It celebrates hitting 10,000 weekly active users and has built notebook LM, the Mariner computer-use agent, the Gemini API, AI Studio, and VO.

Transcript

what I found too Building Products over the years is it's very common everyone talks about product Market fit you'll know it when you see it and all that which is true but at least for me I've always felt in the first part of Building Products you iterate a lot on the product and sometimes you forget to iterate on the market and finding the right M... Read More

Key Insights

• Google Labs is Google's experimental arm, relaunched about three years ago, tasked with building zero-to-one AI products people love across consumer, B2B, and developer categories while sitting outside the big traditional Google product areas.
• Writing paragraph-level prompts is becoming old fashioned for end users, according to Josh Woodward, who predicts people will increasingly deliver context through images, PDFs, voice, or video rather than long text boxes.
• Context itself is not going away because these models love context, but the type of context and the way users deliver it is changing very fast, which is where Google Labs is placing many of its bets.
• Google Labs moves from idea to end users' hands in 50 to 100 days, prioritizing speed because it operates during an AI platform shift moment where fast movement matters enormously.
• Big things start small is a core Labs principle: teams get excited about reaching 10,000 weekly active users, a milestone so small that other Google groups' dashboards do not even count that low.
• Labs attracts a mix of longtime Googlers, startup founders, and ex-founders, seeking creative people who see themselves as underdogs, and it maintains a doc called 'labs in a nutshell' listing 16 or 17 bullets on who thrives there.
• Project selection blends top-down and bottom-up: leadership identifies strategic areas on mission for Google (the right 'ponds to fish in'), then four-or-five-person teams choose the specific user problems to solve.
• The 'AI joystick' concept, coined by a Labs marketing person, describes giving users a feeling of steering and control over the AI, a pattern Woodward highlights in notebook LM where users bring their own sources.

Questions & Answers

Q: What is Google Labs and what does it do?

Google Labs is Google's experimental arm, described by Josh Woodward as a collection of builders trying to build new AI products that people love. Relaunched about three years ago after an earlier version went dormant, it works on all zero-to-one products, spanning consumer, B2B, and developer offerings. It sits outside the big traditional Google product areas but collaborates with teams like search and Chrome, while also drawing on Google DeepMind's research frontier.

Q: Why does Josh Woodward say writing prompts is becoming old fashioned?

Woodward believes people will look back and be surprised that everyone tried to write paragraph-level prompts into little boxes. While AI engineers still write beautiful multi-page prompts, everyday users do not have time for that and would otherwise need to be a kind of 'whisperer' to unlock a model's ability. Labs sees more traction when users bring their own assets, such as dragging in a PDF or image, to shortcut long paragraph writing.

Q: How fast does Google Labs ship new products?

Google Labs moves from an idea to end users' hands in 50 to 100 days. Woodward emphasizes that speed matters a lot, especially during an AI platform shift moment, and the team does all kinds of things to make that rapid pace happen. This fast-moving culture is a deliberate value, differentiating Labs from other Google groups that operate on multi-quarter road maps rather than racing to the next user milestone.

Q: What does Google Labs consider a success milestone for a new product?

Labs gets excited about reaching 10,000 weekly active users, treating it as a big moment worth celebrating when starting a new project. Woodward notes this number is so small that many other Google groups' dashboards do not even count that low, since some Google products serve billions of people. The philosophy is that big things start small, usually by solving for one user and one pain point first.

Q: What kind of person thrives at Google Labs?

Labs looks for very creative people who almost see themselves as underdogs and have a hustle to them. It attracts a mix of longtime Googlers, startup founders, and ex-founders, and prizes 'unicorns' who combine deep model expertise with consumer product sensibility. The team maintains a doc called 'labs in a nutshell,' whose favorite section, 'who thrives in labs,' lists 16 or 17 bullets. Labs also looks at signals like a candidate's GitHub history.

Q: How does Google Labs decide which projects to work on?

Project selection is a blend of top-down and bottom-up. From the top down, leadership identifies areas that are strategic and on mission for Google, such as the future of software development, which Woodward compares to choosing the right 'ponds to fish in.' Then the four-or-five-person teams come up with the specific user problems to solve, meeting in the middle. This can look a little chaotic, without multi-quarter road maps.

Q: What products has Google Labs built?

Google Labs has worked on the Gemini API, AI Studio, notebook LM, the Mariner computer-use agent, and VO, Google's imagery and video model. Woodward is especially excited about notebook LM, where users bring their own sources and the AI grips into that material, enabling creations like the podcasts released last year. He says much more is coming that follows this pattern of giving users control and a feeling of steering the AI.

Q: What is the 'AI joystick' concept mentioned in the video?

The 'AI joystick' is a term coined by one of the marketing people on the Labs team to describe giving users a sense of control, as if they are steering the AI. Woodward highlights this in the context of notebook LM, where people bring their own sources and feel they are directing the AI rather than just receiving output. He finds this interesting because it gives people a lot of control over the experience.

Summary & Key Takeaways

• Josh Woodward, VP of Google Labs, joins Sequoia Capital's Ravi Gupta and Sonya Huang to describe Labs as a collection of builders creating zero-to-one AI products. Relaunched about three years ago, it sits outside Google's big product areas while keeping one foot inside Google DeepMind's research frontier.

• Woodward argues that writing long paragraph prompts is old fashioned for everyday users. While AI engineers still write multi-page prompts, most people prefer to supply context through images, PDFs, voice, or video. Context stays essential, but how it is delivered is changing fast, and Labs is betting on that shift.

• Labs prizes speed, shipping from idea to users in 50 to 100 days, and celebrates small milestones like 10,000 weekly active users. It hires creative underdogs and unicorns skilled in both research and product, blending top-down strategy with team-chosen user problems. Products include notebook LM, Mariner, and VO.