He tries 1,000 ideas a decade & kills nearly all—that's how X built Waymo & Google Brain. | Astro Teller, CEO of Moonshots at X (Alphabet)

Alphabet's X kills 1,000 ideas to find one Waymo

Alphabet's X is a factory for moonshots. It tries a thousand ideas a decade, funds them for weeks or months, and kills nearly all of them. The scarce skill is burning down the riskiest assumption fast, which X calls training the monkey first. Waymo came from that process, and X's job is to bridge the gap between basic science and a venture-investable company.

Takeaways

• X tries 1,000 ideas for a few hits. X looks at 100 to 200 ideas a year, giving them code names. The goal is the one or two percent that become once-in-a-generation businesses for Alphabet, like Waymo and Google Brain.
• Train the monkey before building the pedestal. Teller's test asks whether the project's hard part is the risky assumption. Building a 10-foot pedestal for a Shakespeare-reciting monkey is half the work but burns no risk, so Xers ask what's the monkey.
• Teams write their own kill criteria. Past investigation, teams write what must be true in a year. If they cannot tell that story, the default is to kill the project. Teller does not write the criteria so the team owns them.
• A moonshot must be a testable hypothesis. X defines it as a huge world problem and a science-fiction-sounding solution. The technology must make the idea testable. Google Glass failed because X could not prove that if it could make the product, it would win.
• Waymo reframed from cars to safest driver. Started around 2009 or 2010 when LIDAR and radar were ready enough, the case was a million deaths and a trillion dollars wasted yearly. The framing moved from selling cars to building the world's safest driver.
• Moonshot market fit needs incumbent rapport. X spent five or six years fitting with electric grid operators before building. Selling to incumbents who resist change can yield a billion dollars a year without changing the industry.

How X staffs early projects

• No one is full time on a project at the start. People carry a major and a minor because projects often stop.
• X keeps world experts in electrochemistry and 3D printing on staff. A project can borrow five or ten percent of an expert's time without hiring anyone.
• A project with three or four people full time can act like a ten-person company because of the resources it can borrow around X.
• X does not track whose idea it was and discourages talk about who founded an idea, keeping the process bottom-up.

What happens when a project graduates

• The team is given founders' shares, and a lot of the company goes to them.