10 years ago this week, Astro Teller was asked by Google’s Larry Page and Sergey Brin to run a great experiment. To build a “moonshot factory”. They called it “X”.

Astro (his real name is Eric but his friends thought his spiky hair reminded them of astroturf) was born in Cambridge, England then grew up just outside Chicago, Illinois. He had great pedigree, his grandparents include French economist and mathematician Gérard Debreu and Hungarian-born American theoretical physicist Edward Teller.

Astro followed a similar path, studying computer science at Stanford, and then gained a PhD in AI. He wrote a novel, Exegesis, when he was 27 about an AI program that develops consciousness and will and begins to correspond with its creator. His early career saw him co-found BodyMedia, a maker of wearable devices that measure sleep, perspiration and calories burned.

At X, he became known as Captain of Moonshots. This is Astro’s story of a decade of moonshots.

So, why build a Moonshot Factory?

In 2010 he was recruited by Google to create “something far beyond an innovation lab” ideas were initially fuzzy as what X was going to do, but they knew they wanted to create an organisation that could invent and launch breakthrough technologies that would make the world a radically better place.

They also recognised that X could play an important role in Google’s future development, reaching beyond its core search engine business and beating what many called the innovator’s dilemma, to create the future business whilst still focused on today’s.

Teller was motivated to create an organisation that could repeatedly generate breakthrough innovations filled with hyper-creative people and weird, out-there thinking. He says “humanity is choosing to keep much of its potential off the table, underemployed, and under-utilised. Many people go to work at jobs that aren’t designed to be fulfilling, and many more don’t get the chance to contribute at all.”

“Meanwhile, many powerful organisations are forced to focus on their own profitability and short term goals at the expense of everything else,  leaving the status quo intact, or at best only making the world incrementally better. Yet the massive problems facing us this century need the widest array of minds, the wildest imaginations, and enormous commitments of time, resources, and attention.”

X itself is a prototype solution to this need. In particular Teller hopes to prove that “good for the world” can be financially rewarding, so that more organisations are inspired to operate like this.

Searching for radical creativity with real impact has not been easy. He says “what I’ve come to realise is that our main cultural battle is against fear and the strong gravitational pull toward conventional ways of thinking and behaving. All of us have been conditioned for years not to fail, not to be vulnerable, and to minimise risk.”

Teller says that 7 big lessons have emerged:

#1 Stop thinking you can predict the future. Nobody is much better than random at knowing what businesses and products will succeed in the long run. Instead, try audacious things and figure out as quickly as you can when you’re wrong. Sometimes it takes dozens of iterations; one team at X is working on improving how people hear, and they explored 35 different ideas before they found the one we’re going to pour fuel on.

#2 Take a long term view — it changes everything. Work on hard problems whose answers are 5–10 years over the horizon. This gives space to explore and experiment and learn more deeply. Taking the long view also enables us to think through the implications not just the applications of what we’re building. Take, for example, the self-driving car concept, which could easily have been trivialised as a smart driving app, but instead become something more profound.

#3 Carve out space for creative and weird souls. Willy Wonka had to build a chocolate factory to house the Oompa Loompas, because they struggled to survive in the wild. And innovators and dreamers often can’t thrive in typical organizations. Their constant “what ifs” and “why does it have to be this ways” can be irritating for organizations trying to lock in an execution plan and meet quarterly targets.

#4 Dream like a child, test like a grownup. You can get way further than you think just by being optimistic enough to try. You’d be surprised how many people find motivation and joy in a ridiculously hard problem, especially if it has a magical-sounding solution. This is why we’ve often said that 10X can actually be easier than 10%. Before it seemed possible, teams at X believed in self-flying delivery drones, and now they are becoming real.

#5 Look for holy s**t moments — they can get you further than clear objectives. Most projects don’t have quantified goals because they want to be free to find the intriguing paths that lead to the biggest surprises. But it’s hard to resist the allure of a clear milestone because humans dislike uncertainty. To counteract this, teams are asked to try a bunch of experiments and come back when they’ve found something that makes everyone say “holy s**t.”

#6 Cultivate the ability to be passionately dispassionate. While nobody can predict the future, X is trying to invent it efficiently, to maximize the impact of our effort and resources. So it’s absolutely necessary to be intellectually honest and kill things that are pretty good (even  loved) in order move on to better  opportunities. One of the hardest was a beautiful idea to create carbon-neutral fuel from seawater. It worked, but not at a realistic cost.

#7 Create fearless teams of chaos pilots. The lone inventor having a eureka moment is largely a myth; innovation comes from great teams. This doesn’t mean innovation by committee or consensus. X deliberately assembles teams from a wide range of backgrounds, cultures, and communities, so we can generate more creative ideas and ask hard questions. Former rocket scientists work alongside concert pianists and puppeteers, and marine biologists mingle with physicists and machine learning experts.

10 years into X, many companies and products have emerged that carry the X DNA, including successful “graduate” businesses including:

Waymo … transforming mobility with self-driving cars

Every year 1.25 million people around the world die from car accidents. Of these accidents, an estimated 94% are caused by human error. A common culprit? Human inattention, made worse in recent years by the rise of mobile devices.

This begs the question: what if cars could drive themselves safely from point A to point B? X was home to the Google Self-Driving Car project, which had the goal of developing technology that could transform mobility for millions of people, whether by reducing road deaths caused by human error, reclaiming the billions of hours wasted in traffic, or bringing everyday destinations within reach for those unable to drive. In 2016, the project graduated from X to become Waymo.

In December 2016, with over two million miles of self-driving experience on the roads — the equivalent of 300 years of human driving — the Self-Driving Car project graduated from X. Today it’s Waymo, a self-driving technology company with a clear mission: to make it safe and easy for people and things to move around. Waymo’s first public trial is currently underway in Phoenix, AZ where the team continues to deliver on their quest to improve road safety and mobility for everyone.

Loon … expanding Internet connectivity with stratospheric balloons

The Internet has transformed the way the world communicates, does business, learns, governs, and exchanges ideas, but not everyone can harness the benefits and advantages it provides. Right now, billions of people across the globe still do not have Internet access. They are completely left out of a digital revolution that could improve their finances, education, and health.

Project Loon is a radical approach to expanding Internet connectivity. Instead of trying to extend the Internet from the ground, Loon takes to the sky via a network of balloons, traveling along the edge of space, to expand Internet connectivity to rural areas, fill coverage gaps, and improve network resilience in the event of a disaster.

Loon has delivered connectivity to communities where the communications infrastructure has been damaged or wiped out. Loon partnered with Telefonica over many months in 2017 to provide basic Internet connectivity to tens of thousands of people across Peru who were displaced due to extreme rains and flooding. The Loon team also worked closely with AT&T and T-Mobile to bring the Internet to more than 200,00 people in Puerto Rico after Hurricane Maria made landfall.

Verily … inventing new technologies and services to help people live healthier lives

Starting in 2012, X invested in several projects that had the potential to transform the detection, prevention, and management of disease. They were inspired by a common insight: technological advances like machine learning, the miniaturization of sensors, and wearable devices were likely to lead to innovations that could someday help health professionals be more proactive in their treatment of disease. Originally known as Google Life Sciences, this collection of projects ultimately graduated from X in 2015 to become the Alphabet company Verily Life Sciences.

The team at Verily wants to help address some of the most daunting challenges in healthcare. They combine expertise in science, engineering, and medicine with a startup spirit, and they collaborate closely with partners across the healthcare and life sciences industries who are as eager as they are to challenge convention and relentlessly pursue fresh breakthroughs against common diseases and other global health problems.

Glass … helping people work faster and safer with smart glasses

Workers with hands-on jobs often have to look away from what they’re doing to access information. This shifting of focus is time consuming and can also create distractions that lead to mistakes. Manufacturing and field workers, for example, often have to step away from the work at hand to consult manuals or guides while fixing machinery, and doctors often have to divert their attention from patients to transcribe notes or look up records.

Glass Enterprise Edition is a hands-free, wearable computing device that intuitively fits into the user’s workflow to help them remain engaged and focused on their work. Glass can easily clip onto glasses or safety shields and puts a display in the upper right corner of a user’s field of view to allow them to focus on their work, while simultaneously connect to a deeper world of information.

From giving hands-on workers information where and when they need it to providing extra expertise with a “you-see-what-I-see” video feature, Glass is helping many businesses work better, safer, faster. After two years of development at the moonshot factory the Glass team moved from X back to Google to scale their efforts and make the newest version of their device available to hands on workers everywhere.

Malta … storing renewable energy in molten salt

Wind and solar power are abundant, clean, and increasingly inexpensive energy sources. However, they’re not always available when the demand for power is greatest.

If wind and solar farms are producing more energy than the electric grid needs, the energy goes to waste. In California, up to 30% of solar energy cannot be used when it’s produced. Worse, if electricity demand spikes during periods when the sun isn’t shining or the wind isn’t blowing, utilities will often fire up “peaker plants” to bring extra power online quickly. These are usually powered by fossil fuels and emit large amounts of CO2 relative to ordinary power plants.

Malta is building a grid-scale energy storage technology that stores electricity from renewable energy sources as heat inside large tanks of high temperature molten salt and as cold in large tanks of chilled liquid. The system can discharge electricity back to the grid when energy demand is high – effectively “time shifting” energy from when it’s produced to when it’s most needed.

Malta, Inc is now an independent company. It plans to build and sell industrial-grade, grid-scale energy storage solutions that can be located anywhere in the world. These storage solutions will collect and store large quantities of energy to dispatch quickly as electricity on demand. The team will develop a megawatt-scale pilot plant to prove the technology at commercial scale, and are working with partners that have the expertise to help them build, operate and connect a pilot to the grid.

Dandelion … reducing heating costs and carbon emissions with geothermal energy

In the U.S., buildings account for 39% of all carbon emissions, with the majority of these emissions coming from the combustion of fossil fuels for heating and cooling. In the Northeast of the country, heating and cooling is particularly carbon-intensive due to the high use of fuel oil and propane gas during the cold months. Use of these conventional fuels also has an unfortunate side effect for homeowners — they feel the financial hit if fuel prices rise during a long cold winter.

Dandelion seeks to make it easier and more affordable to heat and cool homes with a clean, free, abundant, and renewable resource: geothermal energy. Dandelion uses high-performance equipment and a proprietary, low-cost installation process that allows homeowners to save money and help the environment by moving away from conventional heating and cooling methods.

After two years of utilizing X’s prototyping techniques and labs to help develop the drilling technology, Dandelion graduated to become an independent company. Dandelion offers high-performance equipment and a proprietary, low-cost installation process that allow homeowners to save money by switching away from conventional heating fuels. The team is currently signing up customers in New York state and is excited to advance geothermal energy as a clean and abundant choice that can help the planet.

Astro Teller wrote a great blog post “Inside Our Moonshot Factory” in 2016:

Ever since we started as Google[x] in 2010, X has had a single mission: to invent and launch “moonshot” technologies that we hope could someday make the world a radically better place. We developed a simple blueprint to help us find ideas that could deliver 10x impact, not just incremental improvement over the status quo: an X project must solve a problem that affects millions or billions of people; it has to have an audacious, sci-fi sounding technology; and there has to be at least a glimmer of hope that it’s actually achievable in the next 5–10 years.

Over the years, we’ve gotten better at finding and developing moonshots that satisfy these criteria, and we’ve learned a lot along the way. What makes this really difficult is that innovation is a delicate thing. You can’t over-process it; you have to respect weird creativity and serendipitous discovery. But if someday you’re going to harvest the fruits of that innovation in the form of revenue and profits, you need to find just the right amount of structure. That’s why we’re trying to build X into a “moonshot factory”.

At X, we’re trying to do much more than just research dazzling new tools and scientific approaches. We’re trying to create the processes and culture that will help us systematize innovation. (We’re not the first to try this, of course; we can trace our inspiration at least as far back as Thomas Edison.) We hope to find repeatable methods for delivering real impact to the world, by developing technology products that provide the foundations of large, sustainable new businesses.

What I’ll share today are some of the factory processes we’ve developed over the years to keep ourselves in the sweet spot between high-risk/idealistic (where most research lives), and safe-bet/pragmatic (where most big companies live).

Responsibly irresponsible

We embrace failure as a powerful tool for learning and, counter-intuitively, making progress; this is one of our most important values. Now, you might be a bit concerned about someone sitting in the CEO chair telling you they’re setting out to fail most of the time. But to make progress toward any audacious idea, you have to make mistakes — you have to seek out frequent, messy, instructive failure that shows you what to do (or not do) next. You can read more in my SXSW talk here and my TED talk here about how this philosophy shapes X’s technology development, and in this post I’ll explore it in the context of our organizational processes.

In short, we try to steer X to be “responsibly irresponsible” as we develop new products. We couple radical, 10x thinking that throws previous conventions to the wind with the discipline to identify risks early, learn cheaply and quickly what’s wrong with our ideas, and face brutal honesty about where we’re succeeding and where we’re not. We now have a system that fuels our optimism while grounding us — just enough — in reality.

Finding people who fall in love with problems

At X, we love new technologies. But technology is the tool, not the end game. If we catch ourselves spending a lot of time refining a new technology and saying, “This could be great for lots of things…someday” without an idea of how to make it so, that’s usually a bad sign. Instead, we want to fall in love with a problem, and aim to understand it so deeply that it becomes easier to find fresh new approaches.

We seek people who are “T-shaped”: they have enormous intellectual flexibility with deep expertise in a particular field, and they can also collaborate easily across diverse domains. (Here’s more about our T-shaped team.) These are the kinds of people who love problems and joyfully apply techniques from one arena to another, like bringing the science of Doritos bags, condoms, and sausage casing to preventing Loon balloons from leaking.

While each moonshot has a core team, we try to keep that team as small as possible, and supplement them with people from a permanent roster of in-house experts, many of whom specialize in helping moonshots make contact with the real world so they can learn as quickly as possible what physics and public opinion have to teach them. This includes things like mechanical engineering, user experience research and design, and public policy. We have a Design Kitchen and several hardware labs that teams can tap for design sprints, rapid prototyping, and failure analysis. These centralized resources make it easy for us to shift people around and quickly help teams of five act like teams of fifty. This also helps us pass along hard-won wisdom and factory lore, while giving us maximum flexibility for when projects fail or go into a reset-and-rebuild mode.

Thinking about X as a portfolio

Being a “corporate lab” is a difficult balancing act: place big bets on the future, but don’t spook the people giving you the money. As an Other Bet (one of the Alphabet divisions that’s not Google), we want to be good stewards of the resources invested here and deliver a good return so that we’re trusted to keep the factory open for years to come.

We look for opportunities to balance X’s overall portfolio sensibly, and aim for diversity: a mix of hardware and software, a mix of industries and problems, a mix of ideas that will take more (closer to 10 years) or less (closer to 5 years) time to have an impact. We have clear budgets and limitations; we can aspire to creating significant growth for Alphabet without significantly growing ourselves.

People often ask us how many projects live at X, and the answer is that it moves around and depends on how you count. We have hundreds of ideas that last for a few hours, dozens for a few weeks, handfuls for months. That’s the nature of moonshot-taking: we weed out ideas as we learn more about the size of their potential or their risk. So the number of explorations under our roof will vary at any given time.

Moving through the factory

We now have a clear process for how ideas move through the factory. They start with our Rapid Evaluation (aka “Rapid Eval”) team, where we investigate the seedlings of technology and science breakthroughs that might offer the core ingredients and inspiration for a moonshot.

There are two stages of Rapid Eval investigations. In the first stage, investigators get a few weeks and a few thousand dollars to try to understand a nascent moonshot’s biggest risks; this kills many dozens of ideas quickly. The second stage is “Extended Investigations.” A couple of team members are given a few months and a bit more money to build prototypes, running at the hardest and riskiest parts of the technology to deeply understand the problem they’re trying to solve. As they build their technology prototypes, they must also build a solid techno-economic analysis that proves this idea could survive in the real world as a large business with a real market and real applications. By design, only a handful of ideas survive this process.

Then — and this is a new stage developed in the summer of 2015 and led by Obi Felten — we have the X Foundry. We developed Foundry because an idea that’s only a few months old may still carry a lot of risk, including risks that we don’t fully understand yet. So we wanted to create a special environment that lasts about a year or so where we could keep teams small and nimble (usually with less than 10 people, and often fewer than five) and focused on the riskiest elements. In this stage we want to learn that an idea can eventually be turned into a product and business, or that it can’t. We look to develop confidence that it can exist comfortably in the real world — not in a rosy vision of how the world should be, but in a reality-tested plan that can make money in a reasonable amount of time.

We expect Foundry to have a handful of projects at any one time. We generally expect that half of the projects will be killed, and half will survive. We might send some back to the Rapid Eval team for re-imagining. And if we decide a product has the potential to become a large business soon, we hire a general manager and set it up as an fully-fledged X project like Project Wing and Project Loon are today.

Actively killing our ideas

One of our most valuable cultural habits is our willingness to kill our ideas. We’ve always conditioned Xers to run at their hardest problems and biggest risks first, and as I wrote earlier this year, we’ve developed a number of techniques for making people feel psychologically safe even as they declare that something’s not working and they want to walk away. This ensures we don’t end up with a roster of wasteful projects, and we quickly can move on to new ideas that are more likely to turn into successful moonshot businesses.

Our teams start each day assuming failure is the norm, and we manage moonshots-in-progress on that basis. For example, we don’t just hand out more resources when teams show us that they’re doing well. It’s very easy to work on simple stuff and show some progress, but that doesn’t teach you the hard things that you need to understand deeply to solve the problem in the long run. So we ask teams at X to identify their biggest risks, and we make those into the team’s milestones; knocking down a major risk will unlock more budget or permission to grow the team a little. For example, an early milestone for the Loon team, which is developing balloon-powered Internet, was to have balloons that could last for three months. They’ve achieved that, so now one of their milestones is to lower the total cost of the balloons.

Our teams start each day assuming failure is the norm, and we manage moonshots-in-progress on that basis.

Kill signals are another one of our important tools. These are metrics that a team agrees on when the project starts, which, if you reach them (or in many cases, fail to reach them in a certain timeframe), indicate that, “We should walk away from this project now.” This makes it easy for teams to remain intellectually honest and make tough decisions when they’re faced with difficult results. A recent investigation we called Foghorn proved the value of having kill signals.

Foghorn had the potential to dramatically cut greenhouse gases from transportation and slow the warming of the planet: we produced carbon-neutral fuel out of plain old seawater. Amazing! But awesome tech has to be possible at a reasonable cost. We stared at the techno-economic analysis we’d developed: we couldn’t see how this technique could produce fuel in the next 5–10 years at a cost that’s competitive with gasoline at the pump. Our optimistic selves started to make excuses: “A breakthrough could be just around the corner!” “Isn’t persistence the key to any great innovation?” But the team’s pragmatism kicked in: they couldn’t deny that they’d hit their kill signals, so they made the decision to shelve their work.

Graduating moonshots from the factory

X is designed to be a protected space for long-term thinking, prototyping, and de-risking. Our strength is building the bridge from idea to proven concept. We’ve gotten good at pushing forward things people think are crazy to the point of feasible product prototype. For example, self-driving cars have hit the tipping point of “not if but when,” we licensed our smart contact lens to Novartis, and we moved Google Brain from academic idea to ready for commercial use, and it now powers many Google products.

Once a team is ready to polish products or scale operations, they’re ready to graduate from X. All of our life sciences projects became Alphabet company Verily in 2015. The Genie team, which developed smart software for designing green buildings, blazed a graduation path to the outside world: they’re now the independent venture-funded startup Flux.io. We also graduate early-stage teams as we see that they can bring a lot of value to another part of Alphabet. That’s why Tango, Insight, GCam, Watch, Glass, and Google Brain, among others, graduated to Google. We celebrate these graduations with mortarboards and diplomas, and then put X’s resources back to work on other moonshot ideas. On the whole, our X graduates have had a high survival rate, which means we’re doing a good job of de-risking (and killing) ideas while they’re here.

The world needs more moonshots

We’re by no means done with our factory, but we think it’s working pretty well so far. At six years in, I’m a firm believer that it’s possible to make progress against previously intractable problems by being committed over the long term, and committed to finding answers that are 10x better, not 10% better. I believe it’s possible to manage long-term bets responsibly by creating a culture where people cheerfully run at the hardest things first, kill their work, and head back to the drawing board to find the next great idea. I hope we can continue to get enough right at X that we inspire other people to try for radical new approaches to the problems they care about most. From climate change to transportation, from protecting our oceans and forests to improving access to food and water, there’s no shortage of problems in the world: what we need is more moonshots.