Jisoo · December 8, 2025
How to Start Google by Paul Graham [Translation]
(This is a talk I gave to 14- and 15-year-old students about what people who might someday want to start a startup should do now. These days, many schools think they should teach students something about startups. This is what I wanted to tell them.)
Most of you probably think that once you go out into the so-called 'real world,' you will eventually have to get some kind of job. But that is not necessarily true. Today I want to talk about one way you might go through your whole life without ever having a job.
The method is very simple: start your own company. Of course, this is not a trick for avoiding work. In fact, if you start your own company, you will usually work much more than you would at an ordinary job. But you can avoid many of the irritating things that come with a job, such as doing what a boss tells you to do.
It is much more fun to work on something you genuinely want to do, rather than someone else's project. You can also make more money. This is the most typical path to becoming rich. If you look at the lists of the world's richest people in the media, most of them are people who started their own companies.
Starting your own company covers a very wide range, from opening a neighborhood barbershop to starting Google. I want to talk about the most extreme end of that spectrum: starting Google.
Companies at the Google end of this spectrum are usually called 'startups' at first. The reason I can say I know this world well is that my wife Jessica and I created a kind of startup factory called Y Combinator. Since 2005, Y Combinator has invested in more than 4,000 startups, and over the past 19 years we have watched countless people start companies up close. So we know fairly precisely what it takes to start a startup.
When I said I would 'teach you how to start Google,' some people may have thought it was a joke. You may have thought, 'How could we possibly start Google?' But in fact, the people who started Google were thinking the same thing before they began. If you had told Larry Page and Sergey Brin that the company they were about to start would one day be worth more than a trillion dollars, they probably would have frozen in place.
When you start a startup, there is very little you know for certain. There is only one thing: it feels worth trying right now. No one knows whether it will grow into a multibillion-dollar company or quietly disappear. So when I talk about 'how to start Google,' I mean how to reach the point where you can start a company with odds similar to Google's when it first began.
How can you get from where you are now to that point? You need three things. First, become sufficiently skilled at some technology. Second, have an idea for what to build. Third, find cofounders to do it with.
First, let us look at technology. How do you get good at a technology? And which technology should you choose? The answer to both questions is actually the same: work on your own projects. Do not try to predict in advance which technology will be most promising, whether it is gene editing, LLMs, rockets, or something else. No one can predict that. Just work on whatever is most interesting to you right now. When you are interested in something, you put in a level of energy that cannot be compared with work you do only because you 'have to.'
If you still feel lost, start with programming. For the past 30 years, programming has almost always been at the center of startups, and the next 10 years are unlikely to be very different.
If you are already taking computer science classes at school, you might think, 'But we are already learning programming.' Honestly, that is not enough. What you learn in class and actually writing code are completely different things. It is common to see people who do well in computer science courses but can barely program. Some people even graduate with computer science degrees from top universities and still cannot code. That is why tech companies give coding tests to everyone, regardless of school or grades. They know very well that a transcript does not prove real ability.
If you really want to become good at programming, there is only one way: work on your own projects. That is the fastest way. Imagine you are making a game. You want to add a feature to the game, but you do not know how. The speed at which you learn in that moment cannot be compared to learning something in class.
Of course, programming is not everything. By 'technology' here, I mean almost anything that can be described with the words 'make' or 'build.' Welding is a technology, making clothes is a technology, and making videos is one too. The important criterion is clear: are you making, or only consuming? Are you making games, or just playing them? That line is the boundary.
Apple founder Steve Jobs learned calligraphy as a teenager, the beautiful lettering you see in medieval manuscripts. At the time, no one, not even Jobs himself, thought it would help his career. He did it simply because it was fun. But the experience later played a decisive role. The Macintosh, which made Apple huge, arrived at the moment when computers could first display typography that looked like print, and at that point Apple overwhelmed its competitors. One reason was that Steve was one of the rare people in the computer industry who truly understood graphic design.
Projects do not have to be serious. It is fine if they are light. What matters is that you, the maker, are excited by them. In fact, 90% of programmers begin by making games. They and their friends like games, so they build what they want to use. And that is exactly what you should be doing at 15 if you may someday want to start a startup.
A project does not have to be just one thing either. It is actually better to try several. Steve Jobs did not learn only calligraphy. He also learned electronics, and that turned out to be a much bigger asset. Ultimately, the key is to follow what genuinely interests you.
This is the first requirement for starting a startup: how to become skilled at some technology. The method is no different from violin or soccer. It is practice. If you want to start a startup at 22 and you begin writing code now, you will have practiced for at least seven years by the time you start. And seven years is enough to become quite good at almost anything.
Now let us assume you are 22 and have become pretty good at a technology. The next question is this: where do startup ideas come from? This may seem like the hardest part. You might think, 'I am good at programming, but how am I supposed to think of an idea like Google?'
In reality, however, it is the opposite. Once you become sufficiently familiar with a technology, ideas become easier to see. When you understand a technology deeply, the empty spaces in the world begin to appear like faint dotted lines: things missing from the technology itself, and problems that technology could easily fix but that remain neglected. Each one is a seed of a startup.
Near my house, there is a store with a notice saying the door does not close properly. The sign has been there for years. To the people at the store, the sticking door probably feels like a natural phenomenon. So they just put up a sign telling customers to be careful and leave it that way. But if a carpenter saw it, he would think, 'Why don't they just shave down the part that sticks?'
Once you become skilled at programming, all the 'missing software' in the world starts to stand out as clearly as a sticking door does to a carpenter. Here is a real example. Until the end of the 20th century, American universities published paper directories containing students' names and contact information. Once you hear the name of these directories, you will immediately know which startup story I am about to tell. They were called 'facebooks,' because there were photos next to students' names.
When Mark Zuckerberg entered Harvard in 2002, the school still had not moved this facebook online. Individual dorms already had online facebooks, but there was no version covering the entire university. The administration was seriously discussing the issue in meeting after meeting, and perhaps it might have solved it 10 years later. Most students did not see it as much of a problem. But Mark was a programmer. He looked at the situation and thought, 'This makes no sense. I could fix this overnight. Let people upload their own photos, gather the data in one place, and make it a site for the whole school.' So he actually did that. And literally almost overnight, thousands of users appeared.
Of course, Facebook at that point was not yet a startup. It was just a project. To emphasize again: projects are not only the best way to learn a technology; they are also the best soil in which startup ideas are born.
Facebook is not the only example. Apple and Google both began as projects. Apple did not start from the idea of creating a company. Steve Wozniak simply wanted to build a computer of his own. Then Steve Jobs said, 'Could we sell this to other people?' and it became a company. Strictly speaking, what they first sold was not computers, but computer designs. Looking back, it was a rather rough beginning.
Google was the same. Larry and Sergey did not initially set out to create a company either. They simply wanted to make search better. Search engines before Google did not sort results by importance. If you searched for 'rugby,' they just listed every webpage containing that word. In 1997, the web was still small, so this approach worked to some extent; there were only about 20 or 30 pages. But the web was growing exponentially, which meant this kind of search was breaking exponentially. Most users merely felt, 'I have to dig through too many results to find what I want.' It was a sticking door. But Larry and Sergey, like Mark, were programmers. They looked at the situation and thought, 'This is strange. Some pages about rugby are clearly more important than others. Why not show those first?'
Looking back now, it is obvious that this was a great startup idea. But at the time, it was not obvious at all. That is always how it is. If starting Apple, Google, or Facebook had been obviously a good idea from the beginning, someone else would already have done it. So the best startups usually come from projects that did not begin with the goal of making a startup. The founders were not trying to start a company; they were simply moving their hands in a direction they found interesting. And if you are young and skilled at a technology, your unconscious sense of what is interesting is far more accurate than your conscious calculation of what might become a good business.
So what really matters for young founders is building something for themselves and their friends. The most common mistake is building something for some unknown 'other people' out there. By contrast, if you can build something that you and your friends truly want to use--not just something they use out of loyalty, but something they would genuinely miss if it disappeared--then it almost certainly contains the seed of a good startup idea. It may not look like a startup right now, and you may not be able to think of a revenue model. That is fine. The method will come later.
A startup idea really needs only one thing: something my friends actually want. That is all. Once you become sufficiently skilled at a technology, such ideas are never hard to find. You begin to see sticking doors everywhere.
Now let us talk about the third and final element you need: a cofounder. The ideal startup usually begins with two or three founders. So where do you find cofounders? You may already guess what I am going to say. Yes: projects. You meet cofounders by working on projects together. Two things matter: whether the person is good at what they do, and whether you work well together. You cannot know that from words alone; you can only know it by actually making something together.
Here I probably have to say something you may not want to hear. Schoolwork really matters. It does not matter whether it is a subject you memorize for tests or a class where people talk at length about literature. The reason is clear. You need to do well in school to get into a good college, and if you want to start a startup, it is advantageous to go to the best college you can. That is where good cofounders are. And the excellent colleagues you will later work with gather there too. When Larry and Sergey started Google, the first thing they did was hire the smartest people they knew at Stanford, and that became a decisive strength.
This is clear empirically as well. If you look at the places that produce especially many successful startups, they overlap almost exactly with the list of the hardest colleges to get into.
I do not think this is because of the prestige of the college name. Nor is it because the quality of education is especially better. The point is simple: the fact that they are hard to get into. To get into MIT or Cambridge, you have to be quite smart and persistent. So people with similar traits naturally gather there.
Of course, you do not have to meet your cofounders at college. The founders of Twitch met when they were seven years old, and Patrick and John Collison, the founders of Stripe, had known each other since birth. Even so, the place where cofounders most often come from is still college. And where cofounders are, ideas are too. The best ideas ultimately emerge among people who have worked on projects together.
So the list of things you need to do from where you are now to reach the point where you can start a startup is surprisingly simple. First, become skilled at a technology. The way to do that is to work on your own projects. Second, do as well as you can in school and get into a good college. That is where cofounders and ideas are.
That is all. Build, and study. Those are the only two things you need to do.
→ How to Start Google
Thanks for reading Commits! Subscribe for free to receive new posts and support my work.