Paul Graham - Essays

If it's hard to change something so simple as a name, imagine how hard it is to garbage-collect an idea. A name only has one point of attachment into your head. An idea for a company gets woven into your thoughts. So you must consciously discount for that. Plunge in, by all means, but remember later to look at your idea in the harsh light of morning and ask: is this something people will pay for? Is this, of all the things we could make, the thing people will pay most for?

The second mistake we made with Artix is also very common. Putting galleries on the Web seemed cool.

One of the most valuable things my father taught me is an old Yorkshire saying: where there's muck, there's brass. Meaning that unpleasant work pays. And more to the point here, vice versa. Work people like doesn't pay well, for reasons of supply and demand. The most extreme case is developing programming languages, which doesn't pay at all, because people like it so much they do it for free.

When we started Artix, I was still ambivalent about business. I wanted to keep one foot in the art world. Big, big, mistake. Going into business is like a hang-glider launch-- you'd better do it wholeheartedly, or not at all. The purpose of a company, and a startup especially, is to make money. You can't have divided loyalties.

Which is not to say that you have to do the most disgusting sort of work, like spamming, or starting a company whose only purpose is patent litigation. What I mean is, if you're starting a company that will do something cool, the aim had better be to make money and maybe be cool, not to be cool and maybe make money.

It's hard enough to make money that you can't do it by accident. Unless it's your first priority, it's unlikely to happen at all.

When I probe our motives with Artix, I see a third mistake: timidity. If you'd proposed at the time that we go into the e-commerce business, we'd have found the idea terrifying. Surely a field like that would be dominated by fearsome startups with five million dollars of VC money each. Whereas we felt pretty sure that we could hold our own in the slightly less competitive business of generating Web sites for art galleries.

We erred ridiculously far on the side of safety. As it turns out, VC-backed startups are not that fearsome. They're too busy trying to spend all that money to get software written. In 1995, the e-commerce business was very competitive as measured in press releases, but not as measured in software. And really it never was. The big fish like Open Market (rest their souls) were just consulting companies pretending to be product companies [5], and the offerings at our end of the market were a couple hundred lines of Perl scripts. Or could have been implemented as a couple hundred lines of Perl; in fact they were probably tens of thousands of lines of C++ or Java. Once we actually took the plunge into e-commerce, it turned out to be surprisingly easy to compete.

So why were we afraid? We felt we were good at programming, but we lacked confidence in our ability to do a mysterious, undifferentiated thing we called "business." In fact there is no such thing as "business." There's selling, promotion, figuring out what people want, deciding how much to charge, customer support, paying your bills, getting customers to pay you, getting incorporated, raising money, and so on. And the combination is not as hard as it seems, because some tasks (like raising money and getting incorporated) are an O(1) pain in the ass, whether you're big or small, and others (like selling and promotion) depend more on energy and imagination than any kind of special training.

Artix was like a hyena, content to survive on carrion because we were afraid of the lions. Except the lions turned out not to have any teeth, and the business of putting galleries online barely qualified as carrion.

Sum up all these sources of error, and it's no wonder we had such a bad idea for a company. We did the first thing we thought of; we were ambivalent about being in business at all; and we deliberately chose an impoverished market to avoid competition.

Looking at the applications for the Summer Founders Program, I see signs of all three. But the first is by far the biggest problem. Most of the groups applying have not stopped to ask: of all the things we could do, is this the one with the best chance of making money?

If they'd already been through their Artix phase, they'd have learned to ask that. After the reception we got from art dealers, we were ready to. This time, we thought, let's make something people want.

Reading the Wall Street Journal for a week should give anyone ideas for two or three new startups. The articles are full of descriptions of problems that need to be solved. But most of the applicants don't seem to have looked far for ideas.

We expected the most common proposal to be for multiplayer games. We were not far off: this was the second most common. The most common was some combination of a blog, a calendar, a dating site, and Friendster. Maybe there is some new killer app to be discovered here, but it seems perverse to go poking around in this fog when there are valuable, unsolved problems lying about in the open for anyone to see. Why did no one propose a new scheme for micropayments? An ambitious project, perhaps, but I can't believe we've considered every alternative. And newspapers and magazines are (literally) dying for a solution.

Why did so few applicants really think about what customers want? I think the problem with many, as with people in their early twenties generally, is that they've been trained their whole lives to jump through predefined hoops. They've spent 15-20 years solving problems other people have set for them. And how much time deciding what problems would be good to solve? Two or three course projects? They're good at solving problems, but bad at choosing them.

But that, I'm convinced, is just the effect of training. Or more precisely, the effect of grading. To make grading efficient, everyone has to solve the same problem, and that means it has to be decided in advance. It would be great if schools taught students how to choose problems as well as how to solve them, but I don't know how you'd run such a class in practice.

The good news is, choosing problems is something that can be learned. I know that from experience. Hackers can learn to make things customers want. [6]

This is a controversial view. One expert on "entrepreneurship" told me that any startup had to include business people, because only they could focus on what customers wanted. I'll probably alienate this guy forever by quoting him, but I have to risk it, because his email was such a perfect example of this view:

80% of MIT spinoffs succeed provided they have at least one management person in the team at the start. The business person represents the "voice of the customer" and that's what keeps the engineers and product development on track.

This is, in my opinion, a crock. Hackers are perfectly capable of hearing the voice of the customer without a business person to amplify the signal for them. Larry Page and Sergey Brin were grad students in computer science, which presumably makes them "engineers." Do you suppose Google is only good because they had some business guy whispering in their ears what customers wanted? It seems to me the business guys who did the most for Google were the ones who obligingly flew Altavista into a hillside just as Google was getting started.

The hard part about figuring out what customers want is figuring out that you need to figure it out. But that's something you can learn quickly. It's like seeing the other interpretation of an ambiguous picture. As soon as someone tells you there's a rabbit as well as a duck, it's hard not to see it.