diff --git "a/haystack_padding.json" "b/haystack_padding.json"
--- "a/haystack_padding.json"
+++ "b/haystack_padding.json"
@@ -1,10 +1,10 @@
-{"context_length":8192,"context_depth":"n\/a","secret_code":"n\/a","copy":0,"context":"_(I wrote this article to help myself understand exactly what McCarthy discovered. You don't need to know this stuff to program in Lisp, but it should be helpful to anyone who wants to understand the essence of Lisp \ufffd both in the sense of its origins and its semantic core. The fact that it has such a core is one of Lisp's distinguishing features, and the reason why, unlike other languages, Lisp has dialects.)_\n\nIn 1960, [John McCarthy](http:\/\/www-formal.stanford.edu\/jmc\/index.html) published a remarkable paper in which he did for programming something like what Euclid did for geometry. He showed how, given a handful of simple operators and a notation for functions, you can build a whole programming language. He called this language Lisp, for \"List Processing,\" because one of his key ideas was to use a simple data structure called a _list_ for both code and data.\n\nIt's worth understanding what McCarthy discovered, not just as a landmark in the history of computers, but as a model for what programming is tending to become in our own time. It seems to me that there have been two really clean, consistent models of programming so far: the C model and the Lisp model. These two seem points of high ground, with swampy lowlands between them. As computers have grown more powerful, the new languages being developed have been [moving steadily](diff.html) toward the Lisp model. A popular recipe for new programming languages in the past 20 years has been to take the C model of computing and add to it, piecemeal, parts taken from the Lisp model, like runtime typing and garbage collection.\n\nIn this article I'm going to try to explain in the simplest possible terms what McCarthy discovered. The point is not just to learn about an interesting theoretical result someone figured out forty years ago, but to show where languages are heading. The unusual thing about Lisp \ufffd in fact, the defining quality of Lisp \ufffd is that it can be written in itself. To understand what McCarthy meant by this, we're going to retrace his steps, with his mathematical notation translated into running Common Lisp code.\n\n[Complete Article (Postscript)](https:\/\/sep.yimg.com\/ty\/cdn\/paulgraham\/jmc.ps?t=1595850613&)\n\n[What Made Lisp Different](diff.html)\n\n[The Code](https:\/\/sep.yimg.com\/ty\/cdn\/paulgraham\/jmc.lisp?t=1595850613&)_(This essay is derived from an invited talk at ICFP 2004.)_\n\nI had a front row seat for the Internet Bubble, because I worked at Yahoo during 1998 and 1999. One day, when the stock was trading around $200, I sat down and calculated what I thought the price should be. The answer I got was $12. I went to the next cubicle and told my friend Trevor. \"Twelve!\" he said. He tried to sound indignant, but he didn't quite manage it. He knew as well as I did that our valuation was crazy.\n\nYahoo was a special case. It was not just our price to earnings ratio that was bogus. Half our earnings were too. Not in the Enron way, of course. The finance guys seemed scrupulous about reporting earnings. What made our earnings bogus was that Yahoo was, in effect, the center of a Ponzi scheme. Investors looked at Yahoo's earnings and said to themselves, here is proof that Internet companies can make money. So they invested in new startups that promised to be the next Yahoo. And as soon as these startups got the money, what did they do with it? Buy millions of dollars worth of advertising on Yahoo to promote their brand. Result: a capital investment in a startup this quarter shows up as Yahoo earnings next quarter\u2014stimulating another round of investments in startups.\n\nAs in a Ponzi scheme, what seemed to be the returns of this system were simply the latest round of investments in it. What made it not a Ponzi scheme was that it was unintentional. At least, I think it was. The venture capital business is pretty incestuous, and there were presumably people in a position, if not to create this situation, to realize what was happening and to milk it.\n\nA year later the game was up. Starting in January 2000, Yahoo's stock price began to crash, ultimately losing 95% of its value.\n\nNotice, though, that even with all the fat trimmed off its market cap, Yahoo was still worth a lot. Even at the morning-after valuations of March and April 2001, the people at Yahoo had managed to create a company worth about $8 billion in just six years.\n\nThe fact is, despite all the nonsense we heard during the Bubble about the \"new economy,\" there was a core of truth. You need that to get a really big bubble: you need to have something solid at the center, so that even smart people are sucked in. (Isaac Newton and Jonathan Swift both lost money in the South Sea Bubble of 1720.)\n\nNow the pendulum has swung the other way. Now anything that became fashionable during the Bubble is ipso facto unfashionable. But that's a mistake\u2014an even bigger mistake than believing what everyone was saying in 1999. Over the long term, what the Bubble got right will be more important than what it got wrong.\n\n**1\\. Retail VC**\n\nAfter the excesses of the Bubble, it's now considered dubious to take companies public before they have earnings. But there is nothing intrinsically wrong with that idea. Taking a company public at an early stage is simply retail VC: instead of going to venture capital firms for the last round of funding, you go to the public markets.\n\nBy the end of the Bubble, companies going public with no earnings were being derided as \"concept stocks,\" as if it were inherently stupid to invest in them. But investing in concepts isn't stupid; it's what VCs do, and the best of them are far from stupid.\n\nThe stock of a company that doesn't yet have earnings is worth _something._ It may take a while for the market to learn how to value such companies, just as it had to learn to value common stocks in the early 20th century. But markets are good at solving that kind of problem. I wouldn't be surprised if the market ultimately did a better job than VCs do now.\n\nGoing public early will not be the right plan for every company. And it can of course be disruptive\u2014by distracting the management, or by making the early employees suddenly rich. But just as the market will learn how to value startups, startups will learn how to minimize the damage of going public.\n\n**2\\. The Internet**\n\nThe Internet genuinely is a big deal. That was one reason even smart people were fooled by the Bubble. Obviously it was going to have a huge effect. Enough of an effect to triple the value of Nasdaq companies in two years? No, as it turned out. But it was hard to say for certain at the time. \\[1\\]\n\nThe same thing happened during the Mississippi and South Sea Bubbles. What drove them was the invention of organized public finance (the South Sea Company, despite its name, was really a competitor of the Bank of England). And that did turn out to be a big deal, in the long run.\n\nRecognizing an important trend turns out to be easier than figuring out how to profit from it. The mistake investors always seem to make is to take the trend too literally. Since the Internet was the big new thing, investors supposed that the more Internettish the company, the better. Hence such parodies as Pets.Com.\n\nIn fact most of the money to be made from big trends is made indirectly. It was not the railroads themselves that made the most money during the railroad boom, but the companies on either side, like Carnegie's steelworks, which made the rails, and Standard Oil, which used railroads to get oil to the East Coast, where it could be shipped to Europe.\n\nI think the Internet will have great effects, and that what we've seen so far is nothing compared to what's coming. But most of the winners will only indirectly be Internet companies; for every Google there will be ten JetBlues.\n\n**3\\. Choices**\n\nWhy will the Internet have great effects? The general argument is that new forms of communication always do. They happen rarely (till industrial times there were just speech, writing, and printing), but when they do, they always cause a big splash.\n\nThe specific argument, or one of them, is the Internet gives us more choices."}
-{"context_length":8192,"context_depth":"n\/a","secret_code":"n\/a","copy":0,"context":"_(This essay is derived from an invited talk at ICFP 2004.)_\n\nI had a front row seat for the Internet Bubble, because I worked at Yahoo during 1998 and 1999. One day, when the stock was trading around $200, I sat down and calculated what I thought the price should be. The answer I got was $12. I went to the next cubicle and told my friend Trevor. \"Twelve!\" he said. He tried to sound indignant, but he didn't quite manage it. He knew as well as I did that our valuation was crazy.\n\nYahoo was a special case. It was not just our price to earnings ratio that was bogus. Half our earnings were too. Not in the Enron way, of course. The finance guys seemed scrupulous about reporting earnings. What made our earnings bogus was that Yahoo was, in effect, the center of a Ponzi scheme. Investors looked at Yahoo's earnings and said to themselves, here is proof that Internet companies can make money. So they invested in new startups that promised to be the next Yahoo. And as soon as these startups got the money, what did they do with it? Buy millions of dollars worth of advertising on Yahoo to promote their brand. Result: a capital investment in a startup this quarter shows up as Yahoo earnings next quarter\u2014stimulating another round of investments in startups.\n\nAs in a Ponzi scheme, what seemed to be the returns of this system were simply the latest round of investments in it. What made it not a Ponzi scheme was that it was unintentional. At least, I think it was. The venture capital business is pretty incestuous, and there were presumably people in a position, if not to create this situation, to realize what was happening and to milk it.\n\nA year later the game was up. Starting in January 2000, Yahoo's stock price began to crash, ultimately losing 95% of its value.\n\nNotice, though, that even with all the fat trimmed off its market cap, Yahoo was still worth a lot. Even at the morning-after valuations of March and April 2001, the people at Yahoo had managed to create a company worth about $8 billion in just six years.\n\nThe fact is, despite all the nonsense we heard during the Bubble about the \"new economy,\" there was a core of truth. You need that to get a really big bubble: you need to have something solid at the center, so that even smart people are sucked in. (Isaac Newton and Jonathan Swift both lost money in the South Sea Bubble of 1720.)\n\nNow the pendulum has swung the other way. Now anything that became fashionable during the Bubble is ipso facto unfashionable. But that's a mistake\u2014an even bigger mistake than believing what everyone was saying in 1999. Over the long term, what the Bubble got right will be more important than what it got wrong.\n\n**1\\. Retail VC**\n\nAfter the excesses of the Bubble, it's now considered dubious to take companies public before they have earnings. But there is nothing intrinsically wrong with that idea. Taking a company public at an early stage is simply retail VC: instead of going to venture capital firms for the last round of funding, you go to the public markets.\n\nBy the end of the Bubble, companies going public with no earnings were being derided as \"concept stocks,\" as if it were inherently stupid to invest in them. But investing in concepts isn't stupid; it's what VCs do, and the best of them are far from stupid.\n\nThe stock of a company that doesn't yet have earnings is worth _something._ It may take a while for the market to learn how to value such companies, just as it had to learn to value common stocks in the early 20th century. But markets are good at solving that kind of problem. I wouldn't be surprised if the market ultimately did a better job than VCs do now.\n\nGoing public early will not be the right plan for every company. And it can of course be disruptive\u2014by distracting the management, or by making the early employees suddenly rich. But just as the market will learn how to value startups, startups will learn how to minimize the damage of going public.\n\n**2\\. The Internet**\n\nThe Internet genuinely is a big deal. That was one reason even smart people were fooled by the Bubble. Obviously it was going to have a huge effect. Enough of an effect to triple the value of Nasdaq companies in two years? No, as it turned out. But it was hard to say for certain at the time. \\[1\\]\n\nThe same thing happened during the Mississippi and South Sea Bubbles. What drove them was the invention of organized public finance (the South Sea Company, despite its name, was really a competitor of the Bank of England). And that did turn out to be a big deal, in the long run.\n\nRecognizing an important trend turns out to be easier than figuring out how to profit from it. The mistake investors always seem to make is to take the trend too literally. Since the Internet was the big new thing, investors supposed that the more Internettish the company, the better. Hence such parodies as Pets.Com.\n\nIn fact most of the money to be made from big trends is made indirectly. It was not the railroads themselves that made the most money during the railroad boom, but the companies on either side, like Carnegie's steelworks, which made the rails, and Standard Oil, which used railroads to get oil to the East Coast, where it could be shipped to Europe.\n\nI think the Internet will have great effects, and that what we've seen so far is nothing compared to what's coming. But most of the winners will only indirectly be Internet companies; for every Google there will be ten JetBlues.\n\n**3\\. Choices**\n\nWhy will the Internet have great effects? The general argument is that new forms of communication always do. They happen rarely (till industrial times there were just speech, writing, and printing), but when they do, they always cause a big splash.\n\nThe specific argument, or one of them, is the Internet gives us more choices. In the \"old\" economy, the high cost of presenting information to people meant they had only a narrow range of options to choose from. The tiny, expensive pipeline to consumers was tellingly named \"the channel.\" Control the channel and you could feed them what you wanted, on your terms. And it was not just big corporations that depended on this principle. So, in their way, did labor unions, the traditional news media, and the art and literary establishments. Winning depended not on doing good work, but on gaining control of some bottleneck.\n\nThere are signs that this is changing. Google has over 82 million unique users a month and annual revenues of about three billion dollars. \\[2\\] And yet have you ever seen a Google ad? Something is going on here.\n\nAdmittedly, Google is an extreme case. It's very easy for people to switch to a new search engine. It costs little effort and no money to try a new one, and it's easy to see if the results are better. And so Google doesn't _have_ to advertise. In a business like theirs, being the best is enough.\n\nThe exciting thing about the Internet is that it's shifting everything in that direction. The hard part, if you want to win by making the best stuff, is the beginning. Eventually everyone will learn by word of mouth that you're the best, but how do you survive to that point? And it is in this crucial stage that the Internet has the most effect. First, the Internet lets anyone find you at almost zero cost. Second, it dramatically speeds up the rate at which reputation spreads by word of mouth. Together these mean that in many fields the rule will be: Build it, and they will come. Make something great and put it online. That is a big change from the recipe for winning in the past century.\n\n**4\\. Youth**\n\nThe aspect of the Internet Bubble that the press seemed most taken with was the youth of some of the startup founders. This too is a trend that will last. There is a huge standard deviation among 26 year olds. Some are fit only for entry level jobs, but others are ready to rule the world if they can find someone to handle the paperwork for them.\n\nA 26 year old may not be very good at managing people or dealing with the SEC. Those require experience. But those are also commodities, which can be handed off to some lieutenant."}
-{"context_length":8192,"context_depth":"n\/a","secret_code":"n\/a","copy":0,"context":"People who worry about the increasing gap between rich and poor generally look back on the mid twentieth century as a golden age. In those days we had a large number of high-paying union manufacturing jobs that boosted the median income. I wouldn't quite call the high-paying union job a myth, but I think people who dwell on it are reading too much into it.\n\nOddly enough, it was working with startups that made me realize where the high-paying union job came from. In a rapidly growing market, you don't worry too much about efficiency. It's more important to grow fast. If there's some mundane problem getting in your way, and there's a simple solution that's somewhat expensive, just take it and get on with more important things. EBay didn't win by paying less for servers than their competitors.\n\nDifficult though it may be to imagine now, manufacturing was a growth industry in the mid twentieth century. This was an era when small firms making everything from cars to candy were getting consolidated into a new kind of corporation with national reach and huge economies of scale. You had to grow fast or die. Workers were for these companies what servers are for an Internet startup. A reliable supply was more important than low cost.\n\nIf you looked in the head of a 1950s auto executive, the attitude must have been: sure, give 'em whatever they ask for, so long as the new model isn't delayed.\n\nIn other words, those workers were not paid what their work was worth. Circumstances being what they were, companies would have been stupid to insist on paying them so little.\n\nIf you want a less controversial example of this phenomenon, ask anyone who worked as a consultant building web sites during the Internet Bubble. In the late nineties you could get paid huge sums of money for building the most trivial things. And yet does anyone who was there have any expectation those days will ever return? I doubt it. Surely everyone realizes that was just a temporary aberration.\n\nThe era of labor unions seems to have been the same kind of aberration, just spread over a longer period, and mixed together with a lot of ideology that prevents people from viewing it with as cold an eye as they would something like consulting during the Bubble.\n\nBasically, unions were just Razorfish.\n\nPeople who think the labor movement was the creation of heroic union organizers have a problem to explain: why are unions shrinking now? The best they can do is fall back on the default explanation of people living in fallen civilizations. Our ancestors were giants. The workers of the early twentieth century must have had a moral courage that's lacking today.\n\nIn fact there's a simpler explanation. The early twentieth century was just a fast-growing startup overpaying for infrastructure. And we in the present are not a fallen people, who have abandoned whatever mysterious high-minded principles produced the high-paying union job. We simply live in a time when the fast-growing companies overspend on different things.If you have a US startup called X and you don't have x.com, you should probably change your name.\n\nThe reason is not just that people can't find you. For companies with mobile apps, especially, having the right domain name is not as critical as it used to be for getting users. The problem with not having the .com of your name is that it signals weakness. Unless you're so big that your reputation precedes you, a marginal domain suggests you're a marginal company. Whereas (as Stripe shows) having x.com signals strength even if it has no relation to what you do.\n\nEven good founders can be in denial about this. Their denial derives from two very powerful forces: identity, and lack of imagination.\n\nX is what we _are_, founders think. There's no other name as good. Both of which are false.\n\nYou can fix the first by stepping back from the problem. Imagine you'd called your company something else. If you had, surely you'd be just as attached to that name as you are to your current one. The idea of switching to your current name would seem repellent. \\[[1](#f1n)\\]\n\nThere's nothing intrinsically great about your current name. Nearly all your attachment to it comes from it being attached to you. \\[[2](#f1n)\\]\n\nThe way to neutralize the second source of denial, your inability to think of other potential names, is to acknowledge that you're bad at naming. Naming is a completely separate skill from those you need to be a good founder. You can be a great startup founder but hopeless at thinking of names for your company.\n\nOnce you acknowledge that, you stop believing there is nothing else you could be called. There are lots of other potential names that are as good or better; you just can't think of them.\n\nHow do you find them? One answer is the default way to solve problems you're bad at: find someone else who can think of names. But with company names there is another possible approach. It turns out almost any word or word pair that is not an obviously bad name is a sufficiently good one, and the number of such domains is so large that you can find plenty that are cheap or even untaken. So make a list and try to buy some. That's what [Stripe](http:\/\/www.quora.com\/How-did-Stripe-come-up-with-its-name?share=1) did. (Their search also turned up parse.com, which their friends at Parse took.)\n\nThe reason I know that naming companies is a distinct skill orthogonal to the others you need in a startup is that I happen to have it. Back when I was running YC and did more office hours with startups, I would often help them find new names. 80% of the time we could find at least one good name in a 20 minute office hour slot.\n\nNow when I do office hours I have to focus on more important questions, like what the company is doing. I tell them when they need to change their name. But I know the power of the forces that have them in their grip, so I know most won't listen. \\[[3](#f1n)\\]\n\nThere are of course examples of startups that have succeeded without having the .com of their name. There are startups that have succeeded despite any number of different mistakes. But this mistake is less excusable than most. It's something that can be fixed in a couple days if you have sufficient discipline to acknowledge the problem.\n\n100% of the top 20 YC companies by valuation have the .com of their name. 94% of the top 50 do. But only 66% of companies in the current batch have the .com of their name. Which suggests there are lessons ahead for most of the rest, one way or another.\n\n**Notes**\n\n\\[1\\] Incidentally, this thought experiment works for [nationality and religion](identity.html) too.\n\n\\[2\\] The liking you have for a name that has become part of your identity manifests itself not directly, which would be easy to discount, but as a collection of specious beliefs about its intrinsic qualities. (This too is true of nationality and religion as well.)\n\n\\[3\\] Sometimes founders know it's a problem that they don't have the .com of their name, but delusion strikes a step later in the belief that they'll be able to buy it despite having no evidence it's for sale. Don't believe a domain is for sale unless the owner has already told you an asking price.\n\n**Thanks** to Sam Altman, Jessica Livingston, and Geoff Ralston for reading drafts of this.A couple months ago I got an email from a recruiter asking if I was interested in being a \"technologist in residence\" at a new venture capital fund. I think the idea was to play Karl Rove to the VCs' George Bush.\n\nI considered it for about four seconds. Work for a VC fund? Ick.\n\nOne of my most vivid memories from our startup is going to visit Greylock, the famous Boston VCs. They were the most arrogant people I've met in my life. And I've met a lot of arrogant people. \\[1\\]\n\nI'm not alone in feeling this way, of course. Even a VC friend of mine dislikes VCs. \"Assholes,\" he says.\n\nBut lately I've been learning more about how the VC world works, and a few days ago it hit me that there's a reason VCs are the way they are. It's not so much that the business attracts jerks, or even that the power they wield corrupts them."}
-{"context_length":8192,"context_depth":"n\/a","secret_code":"n\/a","copy":0,"context":"The biggest component in most investors' opinion of you is the opinion of other investors. Which is of course a recipe for exponential growth. When one investor wants to invest in you, that makes other investors want to, which makes others want to, and so on.\n\nSometimes inexperienced founders mistakenly conclude that manipulating these forces is the essence of fundraising. They hear stories about stampedes to invest in successful startups, and think it's therefore the mark of a successful startup to have this happen. But actually the two are not that highly correlated. Lots of startups that cause stampedes end up flaming out (in extreme cases, partly as a result of the stampede), and lots of very successful startups were only moderately popular with investors the first time they raised money.\n\nSo the point of this essay is not to explain how to create a stampede, but merely to explain the forces that generate them. These forces are always at work to some degree in fundraising, and they can cause surprising situations. If you understand them, you can at least avoid being surprised.\n\nOne reason investors like you more when other investors like you is that you actually become a better investment. Raising money decreases the risk of failure. Indeed, although investors hate it, you are for this reason justified in raising your valuation for later investors. The investors who invested when you had no money were taking more risk, and are entitled to higher returns. Plus a company that has raised money is literally more valuable. After you raise the first million dollars, the company is at least a million dollars more valuable, because it's the same company as before, plus it has a million dollars in the bank. \\[[1](#f1n)\\]\n\nBeware, though, because later investors so hate to have the price raised on them that they resist even this self-evident reasoning. Only raise the price on an investor you're comfortable with losing, because some will angrily refuse. \\[[2](#f2n)\\]\n\nThe second reason investors like you more when you've had some success at fundraising is that it makes you more confident, and an investors' opinion of [you](convince.html) is the foundation of their opinion of your company. Founders are often surprised how quickly investors seem to know when they start to succeed at raising money. And while there are in fact lots of ways for such information to spread among investors, the main vector is probably the founders themselves. Though they're often clueless about technology, most investors are pretty good at reading people. When fundraising is going well, investors are quick to sense it in your increased confidence. (This is one case where the average founder's inability to remain poker-faced works to your advantage.)\n\nBut frankly the most important reason investors like you more when you've started to raise money is that they're bad at judging startups. Judging startups is hard even for the best investors. The mediocre ones might as well be flipping coins. So when mediocre investors see that lots of other people want to invest in you, they assume there must be a reason. This leads to the phenomenon known in the Valley as the \"hot deal,\" where you have more interest from investors than you can handle.\n\nThe best investors aren't influenced much by the opinion of other investors. It would only dilute their own judgment to average it together with other people's. But they are indirectly influenced in the practical sense that interest from other investors imposes a deadline. This is the fourth way in which offers beget offers. If you start to get far along the track toward an offer with one firm, it will sometimes provoke other firms, even good ones, to make up their minds, lest they lose the deal.\n\nUnless you're a wizard at negotiation (and if you're not sure, you're not) be very careful about exaggerating this to push a good investor to decide. Founders try this sort of thing all the time, and investors are very sensitive to it. If anything oversensitive. But you're safe so long as you're telling the truth. If you're getting far along with investor B, but you'd rather raise money from investor A, you can tell investor A that this is happening. There's no manipulation in that. You're genuinely in a bind, because you really would rather raise money from A, but you can't safely reject an offer from B when it's still uncertain what A will decide.\n\nDo not, however, tell A who B is. VCs will sometimes ask which other VCs you're talking to, but you should never tell them. Angels you can sometimes tell about other angels, because angels cooperate more with one another. But if VCs ask, just point out that they wouldn't want you telling other firms about your conversations, and you feel obliged to do the same for any firm you talk to. If they push you, point out that you're inexperienced at fundraising \u2014 which is always a safe card to play \u2014 and you feel you have to be extra cautious. \\[[3](#f3n)\\]\n\nWhile few startups will experience a stampede of interest, almost all will at least initially experience the other side of this phenomenon, where the herd remains clumped together at a distance. The fact that investors are so much influenced by other investors' opinions means you always start out in something of a hole. So don't be demoralized by how hard it is to get the first commitment, because much of the difficulty comes from this external force. The second will be easier.\n\n**Notes**\n\n\\[1\\] An accountant might say that a company that has raised a million dollars is no richer if it's convertible debt, but in practice money raised as convertible debt is little different from money raised in an equity round.\n\n\\[2\\] Founders are often surprised by this, but investors can get very emotional. Or rather indignant; that's the main emotion I've observed; but it is very common, to the point where it sometimes causes investors to act against their own interests. I know of one investor who invested in a startup at a $15 million valuation cap. Earlier he'd had an opportunity to invest at a $5 million cap, but he refused because a friend who invested earlier had been able to invest at a $3 million cap.\n\n\\[3\\] If an investor pushes you hard to tell them about your conversations with other investors, is this someone you want as an investor?\n\n**Thanks** to Paul Buchheit, Jessica Livingston, Geoff Ralston, and Garry Tan for reading drafts of this.In high school I decided I was going to study philosophy in college. I had several motives, some more honorable than others. One of the less honorable was to shock people. College was regarded as job training where I grew up, so studying philosophy seemed an impressively impractical thing to do. Sort of like slashing holes in your clothes or putting a safety pin through your ear, which were other forms of impressive impracticality then just coming into fashion.\n\nBut I had some more honest motives as well. I thought studying philosophy would be a shortcut straight to wisdom. All the people majoring in other things would just end up with a bunch of domain knowledge. I would be learning what was really what.\n\nI'd tried to read a few philosophy books. Not recent ones; you wouldn't find those in our high school library. But I tried to read Plato and Aristotle. I doubt I believed I understood them, but they sounded like they were talking about something important. I assumed I'd learn what in college.\n\nThe summer before senior year I took some college classes. I learned a lot in the calculus class, but I didn't learn much in Philosophy 101. And yet my plan to study philosophy remained intact. It was my fault I hadn't learned anything. I hadn't read the books we were assigned carefully enough. I'd give Berkeley's _Principles of Human Knowledge_ another shot in college. Anything so admired and so difficult to read must have something in it, if one could only figure out what.\n\nTwenty-six years later, I still don't understand Berkeley. I have a nice edition of his collected works. Will I ever read it? Seems unlikely."}
-{"context_length":8192,"context_depth":"n\/a","secret_code":"n\/a","copy":0,"context":"I don't think Apple realizes how badly the App Store approval process is broken. Or rather, I don't think they realize how much it matters that it's broken.\n\nThe way Apple runs the App Store has harmed their reputation with programmers more than anything else they've ever done. Their reputation with programmers used to be great. It used to be the most common complaint you heard about Apple was that their fans admired them too uncritically. The App Store has changed that. Now a lot of programmers have started to see Apple as evil.\n\nHow much of the goodwill Apple once had with programmers have they lost over the App Store? A third? Half? And that's just so far. The App Store is an ongoing karma leak.\n\n\\* \\* \\*\n\nHow did Apple get into this mess? Their fundamental problem is that they don't understand software.\n\nThey treat iPhone apps the way they treat the music they sell through iTunes. Apple is the channel; they own the user; if you want to reach users, you do it on their terms. The record labels agreed, reluctantly. But this model doesn't work for software. It doesn't work for an intermediary to own the user. The software business learned that in the early 1980s, when companies like VisiCorp showed that although the words \"software\" and \"publisher\" fit together, the underlying concepts don't. Software isn't like music or books. It's too complicated for a third party to act as an intermediary between developer and user. And yet that's what Apple is trying to be with the App Store: a software publisher. And a particularly overreaching one at that, with fussy tastes and a rigidly enforced house style.\n\nIf software publishing didn't work in 1980, it works even less now that software development has evolved from a small number of big releases to a constant stream of small ones. But Apple doesn't understand that either. Their model of product development derives from hardware. They work on something till they think it's finished, then they release it. You have to do that with hardware, but because software is so easy to change, its design can benefit from evolution. The standard way to develop applications now is to launch fast and iterate. Which means it's a disaster to have long, random delays each time you release a new version.\n\nApparently Apple's attitude is that developers should be more careful when they submit a new version to the App Store. They would say that. But powerful as they are, they're not powerful enough to turn back the evolution of technology. Programmers don't use launch-fast-and-iterate out of laziness. They use it because it yields the best results. By obstructing that process, Apple is making them do bad work, and programmers hate that as much as Apple would.\n\nHow would Apple like it if when they discovered a serious bug in OS\u00a0X, instead of releasing a software update immediately, they had to submit their code to an intermediary who sat on it for a month and then rejected it because it contained an icon they didn't like?\n\nBy breaking software development, Apple gets the opposite of what they intended: the version of an app currently available in the App Store tends to be an old and buggy one. One developer told me:\n\n> As a result of their process, the App Store is full of half-baked applications. I make a new version almost every day that I release to beta users. The version on the App Store feels old and crappy. I'm sure that a lot of developers feel this way: One emotion is \"I'm not really proud about what's in the App Store\", and it's combined with the emotion \"Really, it's Apple's fault.\"\n\nAnother wrote:\n\n> I believe that they think their approval process helps users by ensuring quality. In reality, bugs like ours get through all the time and then it can take 4-8 weeks to get that bug fix approved, leaving users to think that iPhone apps sometimes just don't work. Worse for Apple, these apps work just fine on other platforms that have immediate approval processes.\n\nActually I suppose Apple has a third misconception: that all the complaints about App Store approvals are not a serious problem. They must hear developers complaining. But partners and suppliers are always complaining. It would be a bad sign if they weren't; it would mean you were being too easy on them. Meanwhile the iPhone is selling better than ever. So why do they need to fix anything?\n\nThey get away with maltreating developers, in the short term, because they make such great hardware. I just bought a new 27\" iMac a couple days ago. It's fabulous. The screen's too shiny, and the disk is surprisingly loud, but it's so beautiful that you can't make yourself care.\n\nSo I bought it, but I bought it, for the first time, with misgivings. I felt the way I'd feel buying something made in a country with a bad human rights record. That was new. In the past when I bought things from Apple it was an unalloyed pleasure. Oh boy! They make such great stuff. This time it felt like a Faustian bargain. They make such great stuff, but they're such assholes. Do I really want to support this company?\n\n\\* \\* \\*\n\nShould Apple care what people like me think? What difference does it make if they alienate a small minority of their users?\n\nThere are a couple reasons they should care. One is that these users are the people they want as employees. If your company seems evil, the best programmers won't work for you. That hurt Microsoft a lot starting in the 90s. Programmers started to feel sheepish about working there. It seemed like selling out. When people from Microsoft were talking to other programmers and they mentioned where they worked, there were a lot of self-deprecating jokes about having gone over to the dark side. But the real problem for Microsoft wasn't the embarrassment of the people they hired. It was the people they never got. And you know who got them? Google and Apple. If Microsoft was the Empire, they were the Rebel Alliance. And it's largely because they got more of the best people that Google and Apple are doing so much better than Microsoft today.\n\nWhy are programmers so fussy about their employers' morals? Partly because they can afford to be. The best programmers can work wherever they want. They don't have to work for a company they have qualms about.\n\nBut the other reason programmers are fussy, I think, is that evil begets stupidity. An organization that wins by exercising power starts to lose the ability to win by doing better work. And it's not fun for a smart person to work in a place where the best ideas aren't the ones that win. I think the reason Google embraced \"Don't be evil\" so eagerly was not so much to impress the outside world as to inoculate themselves against arrogance. \\[[1](#f1n)\\]\n\nThat has worked for Google so far. They've become more bureaucratic, but otherwise they seem to have held true to their original principles. With Apple that seems less the case. When you look at the famous [1984 ad](http:\/\/www.uriahcarpenter.info\/1984.html) now, it's easier to imagine Apple as the dictator on the screen than the woman with the hammer. \\[[2](#f2n)\\] In fact, if you read the dictator's speech it sounds uncannily like a prophecy of the App Store.\n\n> We have triumphed over the unprincipled dissemination of facts.  \n>   \n> We have created, for the first time in all history, a garden of pure ideology, where each worker may bloom secure from the pests of contradictory and confusing truths.\n\nThe other reason Apple should care what programmers think of them is that when you sell a platform, developers make or break you. If anyone should know this, Apple should. VisiCalc made the Apple II.\n\nAnd programmers build applications for the platforms they use. Most applications\u2014most startups, probably\u2014grow out of personal projects. Apple itself did. Apple made microcomputers because that's what Steve Wozniak wanted for himself. He couldn't have afforded a minicomputer. \\[[3](#f3n)\\] Microsoft likewise started out making interpreters for little microcomputers because Bill Gates and Paul Allen were interested in using them. It's a rare startup that doesn't build something the founders use."}
-{"context_length":8192,"context_depth":"n\/a","secret_code":"n\/a","copy":0,"context":"_(This article describes the spam-filtering techniques used in the spamproof web-based mail reader we built to exercise [Arc](arc.html). An improved algorithm is described in [Better Bayesian Filtering](better.html).)_\n\nI think it's possible to stop spam, and that content-based filters are the way to do it. The Achilles heel of the spammers is their message. They can circumvent any other barrier you set up. They have so far, at least. But they have to deliver their message, whatever it is. If we can write software that recognizes their messages, there is no way they can get around that.\n\n\\_ \\_ \\_\n\nTo the recipient, spam is easily recognizable. If you hired someone to read your mail and discard the spam, they would have little trouble doing it. How much do we have to do, short of AI, to automate this process?\n\nI think we will be able to solve the problem with fairly simple algorithms. In fact, I've found that you can filter present-day spam acceptably well using nothing more than a Bayesian combination of the spam probabilities of individual words. Using a slightly tweaked (as described below) Bayesian filter, we now miss less than 5 per 1000 spams, with 0 false positives.\n\nThe statistical approach is not usually the first one people try when they write spam filters. Most hackers' first instinct is to try to write software that recognizes individual properties of spam. You look at spams and you think, the gall of these guys to try sending me mail that begins \"Dear Friend\" or has a subject line that's all uppercase and ends in eight exclamation points. I can filter out that stuff with about one line of code.\n\nAnd so you do, and in the beginning it works. A few simple rules will take a big bite out of your incoming spam. Merely looking for the word \"click\" will catch 79.7% of the emails in my spam corpus, with only 1.2% false positives.\n\nI spent about six months writing software that looked for individual spam features before I tried the statistical approach. What I found was that recognizing that last few percent of spams got very hard, and that as I made the filters stricter I got more false positives.\n\nFalse positives are innocent emails that get mistakenly identified as spams. For most users, missing legitimate email is an order of magnitude worse than receiving spam, so a filter that yields false positives is like an acne cure that carries a risk of death to the patient.\n\nThe more spam a user gets, the less likely he'll be to notice one innocent mail sitting in his spam folder. And strangely enough, the better your spam filters get, the more dangerous false positives become, because when the filters are really good, users will be more likely to ignore everything they catch.\n\nI don't know why I avoided trying the statistical approach for so long. I think it was because I got addicted to trying to identify spam features myself, as if I were playing some kind of competitive game with the spammers. (Nonhackers don't often realize this, but most hackers are very competitive.) When I did try statistical analysis, I found immediately that it was much cleverer than I had been. It discovered, of course, that terms like \"virtumundo\" and \"teens\" were good indicators of spam. But it also discovered that \"per\" and \"FL\" and \"ff0000\" are good indicators of spam. In fact, \"ff0000\" (html for bright red) turns out to be as good an indicator of spam as any pornographic term.\n\n\\_ \\_ \\_\n\nHere's a sketch of how I do statistical filtering. I start with one corpus of spam and one of nonspam mail. At the moment each one has about 4000 messages in it. I scan the entire text, including headers and embedded html and javascript, of each message in each corpus. I currently consider alphanumeric characters, dashes, apostrophes, and dollar signs to be part of tokens, and everything else to be a token separator. (There is probably room for improvement here.) I ignore tokens that are all digits, and I also ignore html comments, not even considering them as token separators.\n\nI count the number of times each token (ignoring case, currently) occurs in each corpus. At this stage I end up with two large hash tables, one for each corpus, mapping tokens to number of occurrences.\n\nNext I create a third hash table, this time mapping each token to the probability that an email containing it is a spam, which I calculate as follows \\[1\\]: (let ((g (\\* 2 (or (gethash word good) 0))) (b (or (gethash word bad) 0))) (unless (< (+ g b) 5) (max .01 (min .99 (float (\/ (min 1 (\/ b nbad)) (+ (min 1 (\/ g ngood)) (min 1 (\/ b nbad))))))))) where word is the token whose probability we're calculating, good and bad are the hash tables I created in the first step, and ngood and nbad are the number of nonspam and spam messages respectively.\n\nI explained this as code to show a couple of important details. I want to bias the probabilities slightly to avoid false positives, and by trial and error I've found that a good way to do it is to double all the numbers in good. This helps to distinguish between words that occasionally do occur in legitimate email and words that almost never do. I only consider words that occur more than five times in total (actually, because of the doubling, occurring three times in nonspam mail would be enough). And then there is the question of what probability to assign to words that occur in one corpus but not the other. Again by trial and error I chose .01 and .99. There may be room for tuning here, but as the corpus grows such tuning will happen automatically anyway.\n\nThe especially observant will notice that while I consider each corpus to be a single long stream of text for purposes of counting occurrences, I use the number of emails in each, rather than their combined length, as the divisor in calculating spam probabilities. This adds another slight bias to protect against false positives.\n\nWhen new mail arrives, it is scanned into tokens, and the most interesting fifteen tokens, where interesting is measured by how far their spam probability is from a neutral .5, are used to calculate the probability that the mail is spam. If probs is a list of the fifteen individual probabilities, you calculate the [combined](naivebayes.html) probability thus: (let ((prod (apply #'\\* probs))) (\/ prod (+ prod (apply #'\\* (mapcar #'(lambda (x) (- 1 x)) probs))))) One question that arises in practice is what probability to assign to a word you've never seen, i.e. one that doesn't occur in the hash table of word probabilities. I've found, again by trial and error, that .4 is a good number to use. If you've never seen a word before, it is probably fairly innocent; spam words tend to be all too familiar.\n\nThere are examples of this algorithm being applied to actual emails in an appendix at the end.\n\nI treat mail as spam if the algorithm above gives it a probability of more than .9 of being spam. But in practice it would not matter much where I put this threshold, because few probabilities end up in the middle of the range.\n\n\\_ \\_ \\_\n\nOne great advantage of the statistical approach is that you don't have to read so many spams. Over the past six months, I've read literally thousands of spams, and it is really kind of demoralizing. Norbert Wiener said if you compete with slaves you become a slave, and there is something similarly degrading about competing with spammers. To recognize individual spam features you have to try to get into the mind of the spammer, and frankly I want to spend as little time inside the minds of spammers as possible.\n\nBut the real advantage of the Bayesian approach, of course, is that you know what you're measuring. Feature-recognizing filters like SpamAssassin assign a spam \"score\" to email. The Bayesian approach assigns an actual probability. The problem with a \"score\" is that no one knows what it means. The user doesn't know what it means, but worse still, neither does the developer of the filter."}
-{"context_length":8192,"context_depth":"n\/a","secret_code":"n\/a","copy":0,"context":"_(This essay is derived from a talk at Google.)_\n\nA few weeks ago I found to my surprise that I'd been granted four [patents](http:\/\/paulgraham.infogami.com\/blog\/morepatents). This was all the more surprising because I'd only applied for three. The patents aren't mine, of course. They were assigned to Viaweb, and became Yahoo's when they bought us. But the news set me thinking about the question of software patents generally.\n\nPatents are a hard problem. I've had to advise most of the startups we've funded about them, and despite years of experience I'm still not always sure I'm giving the right advice.\n\nOne thing I do feel pretty certain of is that if you're against software patents, you're against patents in general. Gradually our machines consist more and more of software. Things that used to be done with levers and cams and gears are now done with loops and trees and closures. There's nothing special about physical embodiments of control systems that should make them patentable, and the software equivalent not.\n\nUnfortunately, patent law is inconsistent on this point. Patent law in most countries says that algorithms aren't patentable. This rule is left over from a time when \"algorithm\" meant something like the Sieve of Eratosthenes. In 1800, people could not see as readily as we can that a great many patents on mechanical objects were really patents on the algorithms they embodied.\n\nPatent lawyers still have to pretend that's what they're doing when they patent algorithms. You must not use the word \"algorithm\" in the title of a patent application, just as you must not use the word \"essays\" in the title of a book. If you want to patent an algorithm, you have to frame it as a computer system executing that algorithm. Then it's mechanical; phew. The default euphemism for algorithm is \"system and method.\" Try a patent search for that phrase and see how many results you get.\n\nSince software patents are no different from hardware patents, people who say \"software patents are evil\" are saying simply \"patents are evil.\" So why do so many people complain about software patents specifically?\n\nI think the problem is more with the patent office than the concept of software patents. Whenever software meets government, bad things happen, because software changes fast and government changes slow. The patent office has been overwhelmed by both the volume and the novelty of applications for software patents, and as a result they've made a lot of mistakes.\n\nThe most common is to grant patents that shouldn't be granted. To be patentable, an invention has to be more than new. It also has to be non-obvious. And this, especially, is where the USPTO has been dropping the ball. Slashdot has an icon that expresses the problem vividly: a knife and fork with the words \"patent pending\" superimposed.\n\nThe scary thing is, this is the _only_ icon they have for patent stories. Slashdot readers now take it for granted that a story about a patent will be about a bogus patent. That's how bad the problem has become.\n\nThe problem with Amazon's notorious one-click patent, for example, is not that it's a software patent, but that it's obvious. Any online store that kept people's shipping addresses would have implemented this. The reason Amazon did it first was not that they were especially smart, but because they were one of the earliest sites with enough clout to force customers to log in before they could buy something. \\[[1](#f1n)\\]\n\nWe, as hackers, know the USPTO is letting people patent the knives and forks of our world. The problem is, the USPTO are not hackers. They're probably good at judging new inventions for casting steel or grinding lenses, but they don't understand software yet.\n\nAt this point an optimist would be tempted to add \"but they will eventually.\" Unfortunately that might not be true. The problem with software patents is an instance of a more general one: the patent office takes a while to understand new technology. If so, this problem will only get worse, because the rate of technological change seems to be increasing. In thirty years, the patent office may understand the sort of things we now patent as software, but there will be other new types of inventions they understand even less.\n\nApplying for a patent is a negotiation. You generally apply for a broader patent than you think you'll be granted, and the examiners reply by throwing out some of your claims and granting others. So I don't really blame Amazon for applying for the one-click patent. The big mistake was the patent office's, for not insisting on something narrower, with real technical content. By granting such an over-broad patent, the USPTO in effect slept with Amazon on the first date. Was Amazon supposed to say no?\n\nWhere Amazon went over to the dark side was not in applying for the patent, but in enforcing it. A lot of companies (Microsoft, for example) have been granted large numbers of preposterously over-broad patents, but they keep them mainly for defensive purposes. Like nuclear weapons, the main role of big companies' patent portfolios is to threaten anyone who attacks them with a counter-suit. Amazon's suit against Barnes & Noble was thus the equivalent of a nuclear first strike.\n\nThat suit probably hurt Amazon more than it helped them. Barnes & Noble was a lame site; Amazon would have crushed them anyway. To attack a rival they could have ignored, Amazon put a lasting black mark on their own reputation. Even now I think if you asked hackers to free-associate about Amazon, the one-click patent would turn up in the first ten topics.\n\nGoogle clearly doesn't feel that merely holding patents is evil. They've applied for a lot of them. Are they hypocrites? Are patents evil?\n\nThere are really two variants of that question, and people answering it often aren't clear in their own minds which they're answering. There's a narrow variant: is it bad, given the current legal system, to apply for patents? and also a broader one: is it bad that the current legal system allows patents?\n\nThese are separate questions. For example, in preindustrial societies like medieval Europe, when someone attacked you, you didn't call the police. There were no police. When attacked, you were supposed to fight back, and there were conventions about how to do it. Was this wrong? That's two questions: was it wrong to take justice into your own hands, and was it wrong that you had to? We tend to say yes to the second, but no to the first. If no one else will defend you, you have to defend yourself. \\[[2](#f2n)\\]\n\nThe situation with patents is similar. Business is a kind of ritualized warfare. Indeed, it evolved from actual warfare: most early traders switched on the fly from merchants to pirates depending on how strong you seemed. In business there are certain rules describing how companies may and may not compete with one another, and someone deciding that they're going to play by their own rules is missing the point. Saying \"I'm not going to apply for patents just because everyone else does\" is not like saying \"I'm not going to lie just because everyone else does.\" It's more like saying \"I'm not going to use TCP\/IP just because everyone else does.\" Oh yes you are.\n\nA closer comparison might be someone seeing a hockey game for the first time, realizing with shock that the players were _deliberately_ bumping into one another, and deciding that one would on no account be so rude when playing hockey oneself.\n\nHockey allows checking. It's part of the game. If your team refuses to do it, you simply lose. So it is in business. Under the present rules, patents are part of the game.\n\nWhat does that mean in practice? We tell the startups we fund not to worry about infringing patents, because startups rarely get sued for patent infringement. There are only two reasons someone might sue you: for money, or to prevent you from competing with them. Startups are too poor to be worth suing for money. And in practice they don't seem to get sued much by competitors, either."}
-{"context_length":8192,"context_depth":"n\/a","secret_code":"n\/a","copy":0,"context":"The Segway hasn't delivered on its initial promise, to put it mildly. There are several reasons why, but one is that people don't want to be seen riding them. Someone riding a Segway looks like a dork.\n\nMy friend Trevor Blackwell built [his own Segway](http:\/\/tlb.org\/#scooter), which we called the Segwell. He also built a one-wheeled version, [the Eunicycle](http:\/\/tlb.org\/#eunicycle), which looks exactly like a regular unicycle till you realize the rider isn't pedaling. He has ridden them both to downtown Mountain View to get coffee. When he rides the Eunicycle, people smile at him. But when he rides the Segwell, they shout abuse from their cars: \"Too lazy to walk, ya fuckin homo?\"\n\nWhy do Segways provoke this reaction? The reason you look like a dork riding a Segway is that you look _smug_. You don't seem to be working hard enough.\n\nSomeone riding a motorcycle isn't working any harder. But because he's sitting astride it, he seems to be making an effort. When you're riding a Segway you're just standing there. And someone who's being whisked along while seeming to do no work \u2014 someone in a sedan chair, for example \u2014 can't help but look smug.\n\nTry this thought experiment and it becomes clear: imagine something that worked like the Segway, but that you rode with one foot in front of the other, like a skateboard. That wouldn't seem nearly as uncool.\n\nSo there may be a way to capture more of the market Segway hoped to reach: make a version that doesn't look so easy for the rider. It would also be helpful if the styling was in the tradition of skateboards or bicycles rather than medical devices.\n\nCuriously enough, what got Segway into this problem was that the company was itself a kind of Segway. It was too easy for them; they were too successful raising money. If they'd had to grow the company gradually, by iterating through several versions they sold to real users, they'd have learned pretty quickly that people looked stupid riding them. Instead they had enough to work in secret. They had focus groups aplenty, I'm sure, but they didn't have the people yelling insults out of cars. So they never realized they were zooming confidently down a blind alley.One of the biggest things holding people back from doing great work is the fear of making something lame. And this fear is not an irrational one. Many great projects go through a stage early on where they don't seem very impressive, even to their creators. You have to push through this stage to reach the great work that lies beyond. But many people don't. Most people don't even reach the stage of making something they're embarrassed by, let alone continue past it. They're too frightened even to start.\n\nImagine if we could turn off the fear of making something lame. Imagine how much more we'd do.\n\nIs there any hope of turning it off? I think so. I think the habits at work here are not very deeply rooted.\n\nMaking new things is itself a new thing for us as a species. It has always happened, but till the last few centuries it happened so slowly as to be invisible to individual humans. And since we didn't need customs for dealing with new ideas, we didn't develop any.\n\nWe just don't have enough experience with early versions of ambitious projects to know how to respond to them. We judge them as we would judge more finished work, or less ambitious projects. We don't realize they're a special case.\n\nOr at least, most of us don't. One reason I'm confident we can do better is that it's already starting to happen. There are already a few places that are living in the future in this respect. Silicon Valley is one of them: an unknown person working on a strange-sounding idea won't automatically be dismissed the way they would back home. In Silicon Valley, people have learned how dangerous that is.\n\nThe right way to deal with new ideas is to treat them as a challenge to your imagination \ufffd not just to have lower standards, but to [switch polarity](altair.html) entirely, from listing the reasons an idea won't work to trying to think of ways it could. That's what I do when I meet people with new ideas. I've become quite good at it, but I've had a lot of practice. Being a partner at Y Combinator means being practically immersed in strange-sounding ideas proposed by unknown people. Every six months you get thousands of new ones thrown at you and have to sort through them, knowing that in a world with a power-law distribution of outcomes, it will be painfully obvious if you miss the needle in this haystack. Optimism becomes urgent.\n\nBut I'm hopeful that, with time, this kind of optimism can become widespread enough that it becomes a social custom, not just a trick used by a few specialists. It is after all an extremely lucrative trick, and those tend to spread quickly.\n\nOf course, inexperience is not the only reason people are too harsh on early versions of ambitious projects. They also do it to seem clever. And in a field where the new ideas are risky, like startups, those who dismiss them are in fact more likely to be right. Just not when their predictions are [weighted by outcome](swan.html).\n\nBut there is another more sinister reason people dismiss new ideas. If you try something ambitious, many of those around you will hope, consciously or unconsciously, that you'll fail. They worry that if you try something ambitious and succeed, it will put you above them. In some countries this is not just an individual failing but part of the national culture.\n\nI wouldn't claim that people in Silicon Valley overcome these impulses because they're morally better. \\[[1](#f1n)\\] The reason many hope you'll succeed is that they hope to rise with you. For investors this incentive is particularly explicit. They want you to succeed because they hope you'll make them rich in the process. But many other people you meet can hope to benefit in some way from your success. At the very least they'll be able to say, when you're famous, that they've known you since way back.\n\nBut even if Silicon Valley's encouraging attitude is rooted in self-interest, it has over time actually grown into a sort of benevolence. Encouraging startups has been practiced for so long that it has become a custom. Now it just seems that that's what one does with startups.\n\nMaybe Silicon Valley is too optimistic. Maybe it's too easily fooled by impostors. Many less optimistic journalists want to believe that. But the lists of impostors they cite are suspiciously short, and plagued with asterisks. \\[[2](#f2n)\\] If you use revenue as the test, Silicon Valley's optimism seems better tuned than the rest of the world's. And because it works, it will spread.\n\nThere's a lot more to new ideas than new startup ideas, of course. The fear of making something lame holds people back in every field. But Silicon Valley shows how quickly customs can evolve to support new ideas. And that in turn proves that dismissing new ideas is not so deeply rooted in human nature that it can't be unlearnt.\n\n\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\n\nUnfortunately, if you want to do new things, you'll face a force more powerful than other people's skepticism: your own skepticism. You too will judge your early work too harshly. How do you avoid that?\n\nThis is a difficult problem, because you don't want to completely eliminate your horror of making something lame. That's what steers you toward doing good work. You just want to turn it off temporarily, the way a painkiller temporarily turns off pain.\n\nPeople have already discovered several techniques that work. Hardy mentions two in _A Mathematician's Apology_:\n\n> Good work is not done by \"humble\" men. It is one of the first duties of a professor, for example, in any subject, to exaggerate a little both the importance of his subject and his importance in it.\n\nIf you overestimate the importance of what you're working on, that will compensate for your mistakenly harsh judgment of your initial results."}
-{"context_length":8192,"context_depth":"n\/a","secret_code":"n\/a","copy":0,"context":"I recently told applicants to Y Combinator that the best advice I could give for getting in, per word, was\n\n> Explain what you've learned from users.\n\nThat tests a lot of things: whether you're paying attention to users, how well you understand them, and even how much they need what you're making.\n\nAfterward I asked myself the same question. What have I learned from YC's users, the startups we've funded?\n\nThe first thing that came to mind was that most startups have the same problems. No two have exactly the same problems, but it's surprising how much the problems remain the same, regardless of what they're making. Once you've advised 100 startups all doing different things, you rarely encounter problems you haven't seen before.\n\nThis fact is one of the things that makes YC work. But I didn't know it when we started YC. I only had a few data points: our own startup, and those started by friends. It was a surprise to me how often the same problems recur in different forms. Many later stage investors might never realize this, because later stage investors might not advise 100 startups in their whole career, but a YC partner will get this much experience in the first year or two.\n\nThat's one advantage of funding large numbers of early stage companies rather than smaller numbers of later-stage ones. You get a lot of data. Not just because you're looking at more companies, but also because more goes wrong.\n\nBut knowing (nearly) all the problems startups can encounter doesn't mean that advising them can be automated, or reduced to a formula. There's no substitute for individual office hours with a YC partner. Each startup is unique, which means they have to be advised by specific partners who know them well. \\[[1](#f1n)\\]\n\nWe learned that the hard way, in the notorious \"batch that broke YC\" in the summer of 2012. Up till that point we treated the partners as a pool. When a startup requested office hours, they got the next available slot posted by any partner. That meant every partner had to know every startup. This worked fine up to 60 startups, but when the batch grew to 80, everything broke. The founders probably didn't realize anything was wrong, but the partners were confused and unhappy because halfway through the batch they still didn't know all the companies yet. \\[[2](#f2n)\\]\n\nAt first I was puzzled. How could things be fine at 60 startups and broken at 80? It was only a third more. Then I realized what had happened. We were using an _O(n2)_ algorithm. So of course it blew up.\n\nThe solution we adopted was the classic one in these situations. We sharded the batch into smaller groups of startups, each overseen by a dedicated group of partners. That fixed the problem, and has worked fine ever since. But the batch that broke YC was a powerful demonstration of how individualized the process of advising startups has to be.\n\nAnother related surprise is how bad founders can be at realizing what their problems are. Founders will sometimes come in to talk about some problem, and we'll discover another much bigger one in the course of the conversation. For example (and this case is all too common), founders will come in to talk about the difficulties they're having raising money, and after digging into their situation, it turns out the reason is that the company is doing badly, and investors can tell. Or founders will come in worried that they still haven't cracked the problem of user acquisition, and the reason turns out to be that their product isn't good enough. There have been times when I've asked \"Would you use this yourself, if you hadn't built it?\" and the founders, on thinking about it, said \"No.\" Well, there's the reason you're having trouble getting users.\n\nOften founders know what their problems are, but not their relative importance. \\[[3](#f3n)\\] They'll come in to talk about three problems they're worrying about. One is of moderate importance, one doesn't matter at all, and one will kill the company if it isn't addressed immediately. It's like watching one of those horror movies where the heroine is deeply upset that her boyfriend cheated on her, and only mildly curious about the door that's mysteriously ajar. You want to say: never mind about your boyfriend, think about that door! Fortunately in office hours you can. So while startups still die with some regularity, it's rarely because they wandered into a room containing a murderer. The YC partners can warn them where the murderers are.\n\nNot that founders listen. That was another big surprise: how often founders don't listen to us. A couple weeks ago I talked to a partner who had been working for YC for a couple batches and was starting to see the pattern. \"They come back a year later,\" she said, \"and say 'We wish we'd listened to you.'\"\n\nIt took me a long time to figure out why founders don't listen. At first I thought it was mere stubbornness. That's part of the reason, but another and probably more important reason is that so much about startups is [counterintuitive](before.html). And when you tell someone something counterintuitive, what it sounds to them is wrong. So the reason founders don't listen to us is that they don't _believe_ us. At least not till experience teaches them otherwise. \\[[4](#f4n)\\]\n\nThe reason startups are so counterintuitive is that they're so different from most people's other experiences. No one knows what it's like except those who've done it. Which is why YC partners should usually have been founders themselves. But strangely enough, the counterintuitiveness of startups turns out to be another of the things that make YC work. If it weren't counterintuitive, founders wouldn't need our advice about how to do it.\n\nFocus is doubly important for early stage startups, because not only do they have a hundred different problems, they don't have anyone to work on them except the founders. If the founders focus on things that don't matter, there's no one focusing on the things that do. So the essence of what happens at YC is to figure out which problems matter most, then cook up ideas for solving them \u2014 ideally at a resolution of a week or less \u2014 and then try those ideas and measure how well they worked. The focus is on action, with measurable, near-term results.\n\nThis doesn't imply that founders should rush forward regardless of the consequences. If you correct course at a high enough frequency, you can be simultaneously decisive at a micro scale and tentative at a macro scale. The result is a somewhat winding path, but executed very rapidly, like the path a running back takes downfield. And in practice there's less backtracking than you might expect. Founders usually guess right about which direction to run in, especially if they have someone experienced like a YC partner to bounce their hypotheses off. And when they guess wrong, they notice fast, because they'll talk about the results at office hours the next week. \\[[5](#f5n)\\]\n\nA small improvement in navigational ability can make you a lot faster, because it has a double effect: the path is shorter, and you can travel faster along it when you're more certain it's the right one. That's where a lot of YC's value lies, in helping founders get an extra increment of focus that lets them move faster. And since moving fast is the essence of a startup, YC in effect makes startups more startup-like.\n\nSpeed defines startups. Focus enables speed. YC improves focus.\n\nWhy are founders uncertain about what to do? Partly because startups almost by definition are doing something new, which means no one knows how to do it yet, or in most cases even what \"it\" is. Partly because startups are so counterintuitive generally. And partly because many founders, especially young and ambitious ones, have been trained to win the wrong way. That took me years to figure out. The educational system in most countries trains you to win by [hacking the test](lesson.html) instead of actually doing whatever it's supposed to measure. But that stops working when you start a startup. So part of what YC does is to retrain founders to stop trying to hack the test."}
-{"context_length":8192,"context_depth":"n\/a","secret_code":"n\/a","copy":0,"context":"Here's a simple trick for getting more people to read what you write: write in spoken language.\n\nSomething comes over most people when they start writing. They write in a different language than they'd use if they were talking to a friend. The sentence structure and even the words are different. No one uses \"pen\" as a verb in spoken English. You'd feel like an idiot using \"pen\" instead of \"write\" in a conversation with a friend.\n\nThe last straw for me was a sentence I read a couple days ago:\n\n> The mercurial Spaniard himself declared: \"After Altamira, all is decadence.\"\n\nIt's from Neil Oliver's _A History of Ancient Britain_. I feel bad making an example of this book, because it's no worse than lots of others. But just imagine calling Picasso \"the mercurial Spaniard\" when talking to a friend. Even one sentence of this would raise eyebrows in conversation. And yet people write whole books of it.\n\nOk, so written and spoken language are different. Does that make written language worse?\n\nIf you want people to read and understand what you write, yes. Written language is more complex, which makes it more work to read. It's also more formal and distant, which gives the reader's attention permission to drift. But perhaps worst of all, the complex sentences and fancy words give you, the writer, the false impression that you're saying more than you actually are.\n\nYou don't need complex sentences to express complex ideas. When specialists in some abstruse topic talk to one another about ideas in their field, they don't use sentences any more complex than they do when talking about what to have for lunch. They use different words, certainly. But even those they use no more than necessary. And in my experience, the harder the subject, the more informally experts speak. Partly, I think, because they have less to prove, and partly because the harder the ideas you're talking about, the less you can afford to let language get in the way.\n\nInformal language is the athletic clothing of ideas.\n\nI'm not saying spoken language always works best. Poetry is as much music as text, so you can say things you wouldn't say in conversation. And there are a handful of writers who can get away with using fancy language in prose. And then of course there are cases where writers don't want to make it easy to understand what they're saying\u2014in corporate announcements of bad news, for example, or at the more [bogus](https:\/\/scholar.google.com\/scholar?hl=en&as_sdt=1,5&q=transgression+narrative+postmodern+gender) end of the humanities. But for nearly everyone else, spoken language is better.\n\nIt seems to be hard for most people to write in spoken language. So perhaps the best solution is to write your first draft the way you usually would, then afterward look at each sentence and ask \"Is this the way I'd say this if I were talking to a friend?\" If it isn't, imagine what you would say, and use that instead. After a while this filter will start to operate as you write. When you write something you wouldn't say, you'll hear the clank as it hits the page.\n\nBefore I publish a new essay, I read it out loud and fix everything that doesn't sound like conversation. I even fix bits that are phonetically awkward; I don't know if that's necessary, but it doesn't cost much.\n\nThis trick may not always be enough. I've seen writing so far removed from spoken language that it couldn't be fixed sentence by sentence. For cases like that there's a more drastic solution. After writing the first draft, try explaining to a friend what you just wrote. Then replace the draft with what you said to your friend.\n\nPeople often tell me how much my essays sound like me talking. The fact that this seems worthy of comment shows how rarely people manage to write in spoken language. Otherwise everyone's writing would sound like them talking.\n\nIf you simply manage to write in spoken language, you'll be ahead of 95% of writers. And it's so easy to do: just don't let a sentence through unless it's the way you'd say it to a friend.\n\n**Thanks** to Patrick Collison and Jessica Livingston for reading drafts of this.Technology tends to separate normal from natural. Our bodies weren't designed to eat the foods that people in rich countries eat, or to get so little exercise. There may be a similar problem with the way we work: a normal job may be as bad for us intellectually as white flour or sugar is for us physically.\n\nI began to suspect this after spending several years working with startup founders. I've now worked with over 200 of them, and I've noticed a definite difference between programmers working on their own startups and those working for large organizations. I wouldn't say founders seem happier, necessarily; starting a startup can be very stressful. Maybe the best way to put it is to say that they're happier in the sense that your body is happier during a long run than sitting on a sofa eating doughnuts.\n\nThough they're statistically abnormal, startup founders seem to be working in a way that's more natural for humans.\n\nI was in Africa last year and saw a lot of animals in the wild that I'd only seen in zoos before. It was remarkable how different they seemed. Particularly lions. Lions in the wild seem about ten times more alive. They're like different animals. I suspect that working for oneself feels better to humans in much the same way that living in the wild must feel better to a wide-ranging predator like a lion. Life in a zoo is easier, but it isn't the life they were designed for.\n\n**Trees**\n\nWhat's so unnatural about working for a big company? The root of the problem is that humans weren't meant to work in such large groups.\n\nAnother thing you notice when you see animals in the wild is that each species thrives in groups of a certain size. A herd of impalas might have 100 adults; baboons maybe 20; lions rarely 10. Humans also seem designed to work in groups, and what I've read about hunter-gatherers accords with research on organizations and my own experience to suggest roughly what the ideal size is: groups of 8 work well; by 20 they're getting hard to manage; and a group of 50 is really unwieldy. \\[[1](#f1n)\\]\n\nWhatever the upper limit is, we are clearly not meant to work in groups of several hundred. And yet\u2014for reasons having more to do with technology than human nature\u2014a great many people work for companies with hundreds or thousands of employees.\n\nCompanies know groups that large wouldn't work, so they divide themselves into units small enough to work together. But to coordinate these they have to introduce something new: bosses.\n\nThese smaller groups are always arranged in a tree structure. Your boss is the point where your group attaches to the tree. But when you use this trick for dividing a large group into smaller ones, something strange happens that I've never heard anyone mention explicitly. In the group one level up from yours, your boss represents your entire group. A group of 10 managers is not merely a group of 10 people working together in the usual way. It's really a group of groups. Which means for a group of 10 managers to work together as if they were simply a group of 10 individuals, the group working for each manager would have to work as if they were a single person\u2014the workers and manager would each share only one person's worth of freedom between them.\n\nIn practice a group of people are never able to act as if they were one person. But in a large organization divided into groups in this way, the pressure is always in that direction. Each group tries its best to work as if it were the small group of individuals that humans were designed to work in. That was the point of creating it. And when you propagate that constraint, the result is that each person gets freedom of action in inverse proportion to the size of the entire tree. \\[[2](#f2n)\\]\n\nAnyone who's worked for a large organization has felt this. You can feel the difference between working for a company with 100 employees and one with 10,000, even if your group has only 10 people."}
+{"context_length":8192,"context_depth":0.0,"secret_code":"n\/a","copy":0,"context":"_(I wrote this article to help myself understand exactly what McCarthy discovered. You don't need to know this stuff to program in Lisp, but it should be helpful to anyone who wants to understand the essence of Lisp \ufffd both in the sense of its origins and its semantic core. The fact that it has such a core is one of Lisp's distinguishing features, and the reason why, unlike other languages, Lisp has dialects.)_\n\nIn 1960, [John McCarthy](http:\/\/www-formal.stanford.edu\/jmc\/index.html) published a remarkable paper in which he did for programming something like what Euclid did for geometry. He showed how, given a handful of simple operators and a notation for functions, you can build a whole programming language. He called this language Lisp, for \"List Processing,\" because one of his key ideas was to use a simple data structure called a _list_ for both code and data.\n\nIt's worth understanding what McCarthy discovered, not just as a landmark in the history of computers, but as a model for what programming is tending to become in our own time. It seems to me that there have been two really clean, consistent models of programming so far: the C model and the Lisp model. These two seem points of high ground, with swampy lowlands between them. As computers have grown more powerful, the new languages being developed have been [moving steadily](diff.html) toward the Lisp model. A popular recipe for new programming languages in the past 20 years has been to take the C model of computing and add to it, piecemeal, parts taken from the Lisp model, like runtime typing and garbage collection.\n\nIn this article I'm going to try to explain in the simplest possible terms what McCarthy discovered. The point is not just to learn about an interesting theoretical result someone figured out forty years ago, but to show where languages are heading. The unusual thing about Lisp \ufffd in fact, the defining quality of Lisp \ufffd is that it can be written in itself. To understand what McCarthy meant by this, we're going to retrace his steps, with his mathematical notation translated into running Common Lisp code.\n\n[Complete Article (Postscript)](https:\/\/sep.yimg.com\/ty\/cdn\/paulgraham\/jmc.ps?t=1595850613&)\n\n[What Made Lisp Different](diff.html)\n\n[The Code](https:\/\/sep.yimg.com\/ty\/cdn\/paulgraham\/jmc.lisp?t=1595850613&)_(This essay is derived from an invited talk at ICFP 2004.)_\n\nI had a front row seat for the Internet Bubble, because I worked at Yahoo during 1998 and 1999. One day, when the stock was trading around $200, I sat down and calculated what I thought the price should be. The answer I got was $12. I went to the next cubicle and told my friend Trevor. \"Twelve!\" he said. He tried to sound indignant, but he didn't quite manage it. He knew as well as I did that our valuation was crazy.\n\nYahoo was a special case. It was not just our price to earnings ratio that was bogus. Half our earnings were too. Not in the Enron way, of course. The finance guys seemed scrupulous about reporting earnings. What made our earnings bogus was that Yahoo was, in effect, the center of a Ponzi scheme. Investors looked at Yahoo's earnings and said to themselves, here is proof that Internet companies can make money. So they invested in new startups that promised to be the next Yahoo. And as soon as these startups got the money, what did they do with it? Buy millions of dollars worth of advertising on Yahoo to promote their brand. Result: a capital investment in a startup this quarter shows up as Yahoo earnings next quarter\u2014stimulating another round of investments in startups.\n\nAs in a Ponzi scheme, what seemed to be the returns of this system were simply the latest round of investments in it. What made it not a Ponzi scheme was that it was unintentional. At least, I think it was. The venture capital business is pretty incestuous, and there were presumably people in a position, if not to create this situation, to realize what was happening and to milk it.\n\nA year later the game was up. Starting in January 2000, Yahoo's stock price began to crash, ultimately losing 95% of its value.\n\nNotice, though, that even with all the fat trimmed off its market cap, Yahoo was still worth a lot. Even at the morning-after valuations of March and April 2001, the people at Yahoo had managed to create a company worth about $8 billion in just six years.\n\nThe fact is, despite all the nonsense we heard during the Bubble about the \"new economy,\" there was a core of truth. You need that to get a really big bubble: you need to have something solid at the center, so that even smart people are sucked in. (Isaac Newton and Jonathan Swift both lost money in the South Sea Bubble of 1720.)\n\nNow the pendulum has swung the other way. Now anything that became fashionable during the Bubble is ipso facto unfashionable. But that's a mistake\u2014an even bigger mistake than believing what everyone was saying in 1999. Over the long term, what the Bubble got right will be more important than what it got wrong.\n\n**1\\. Retail VC**\n\nAfter the excesses of the Bubble, it's now considered dubious to take companies public before they have earnings. But there is nothing intrinsically wrong with that idea. Taking a company public at an early stage is simply retail VC: instead of going to venture capital firms for the last round of funding, you go to the public markets.\n\nBy the end of the Bubble, companies going public with no earnings were being derided as \"concept stocks,\" as if it were inherently stupid to invest in them. But investing in concepts isn't stupid; it's what VCs do, and the best of them are far from stupid.\n\nThe stock of a company that doesn't yet have earnings is worth _something._ It may take a while for the market to learn how to value such companies, just as it had to learn to value common stocks in the early 20th century. But markets are good at solving that kind of problem. I wouldn't be surprised if the market ultimately did a better job than VCs do now.\n\nGoing public early will not be the right plan for every company. And it can of course be disruptive\u2014by distracting the management, or by making the early employees suddenly rich. But just as the market will learn how to value startups, startups will learn how to minimize the damage of going public.\n\n**2\\. The Internet**\n\nThe Internet genuinely is a big deal. That was one reason even smart people were fooled by the Bubble. Obviously it was going to have a huge effect. Enough of an effect to triple the value of Nasdaq companies in two years? No, as it turned out. But it was hard to say for certain at the time. \\[1\\]\n\nThe same thing happened during the Mississippi and South Sea Bubbles. What drove them was the invention of organized public finance (the South Sea Company, despite its name, was really a competitor of the Bank of England). And that did turn out to be a big deal, in the long run.\n\nRecognizing an important trend turns out to be easier than figuring out how to profit from it. The mistake investors always seem to make is to take the trend too literally. Since the Internet was the big new thing, investors supposed that the more Internettish the company, the better. Hence such parodies as Pets.Com.\n\nIn fact most of the money to be made from big trends is made indirectly. It was not the railroads themselves that made the most money during the railroad boom, but the companies on either side, like Carnegie's steelworks, which made the rails, and Standard Oil, which used railroads to get oil to the East Coast, where it could be shipped to Europe.\n\nI think the Internet will have great effects, and that what we've seen so far is nothing compared to what's coming. But most of the winners will only indirectly be Internet companies; for every Google there will be ten JetBlues.\n\n**3\\. Choices**\n\nWhy will the Internet have great effects? The general argument is that new forms of communication always do. They happen rarely (till industrial times there were just speech, writing, and printing), but when they do, they always cause a big splash.\n\nThe specific argument, or one of them, is the Internet gives us more choices."}
+{"context_length":8192,"context_depth":0.0,"secret_code":"n\/a","copy":0,"context":"_(This essay is derived from an invited talk at ICFP 2004.)_\n\nI had a front row seat for the Internet Bubble, because I worked at Yahoo during 1998 and 1999. One day, when the stock was trading around $200, I sat down and calculated what I thought the price should be. The answer I got was $12. I went to the next cubicle and told my friend Trevor. \"Twelve!\" he said. He tried to sound indignant, but he didn't quite manage it. He knew as well as I did that our valuation was crazy.\n\nYahoo was a special case. It was not just our price to earnings ratio that was bogus. Half our earnings were too. Not in the Enron way, of course. The finance guys seemed scrupulous about reporting earnings. What made our earnings bogus was that Yahoo was, in effect, the center of a Ponzi scheme. Investors looked at Yahoo's earnings and said to themselves, here is proof that Internet companies can make money. So they invested in new startups that promised to be the next Yahoo. And as soon as these startups got the money, what did they do with it? Buy millions of dollars worth of advertising on Yahoo to promote their brand. Result: a capital investment in a startup this quarter shows up as Yahoo earnings next quarter\u2014stimulating another round of investments in startups.\n\nAs in a Ponzi scheme, what seemed to be the returns of this system were simply the latest round of investments in it. What made it not a Ponzi scheme was that it was unintentional. At least, I think it was. The venture capital business is pretty incestuous, and there were presumably people in a position, if not to create this situation, to realize what was happening and to milk it.\n\nA year later the game was up. Starting in January 2000, Yahoo's stock price began to crash, ultimately losing 95% of its value.\n\nNotice, though, that even with all the fat trimmed off its market cap, Yahoo was still worth a lot. Even at the morning-after valuations of March and April 2001, the people at Yahoo had managed to create a company worth about $8 billion in just six years.\n\nThe fact is, despite all the nonsense we heard during the Bubble about the \"new economy,\" there was a core of truth. You need that to get a really big bubble: you need to have something solid at the center, so that even smart people are sucked in. (Isaac Newton and Jonathan Swift both lost money in the South Sea Bubble of 1720.)\n\nNow the pendulum has swung the other way. Now anything that became fashionable during the Bubble is ipso facto unfashionable. But that's a mistake\u2014an even bigger mistake than believing what everyone was saying in 1999. Over the long term, what the Bubble got right will be more important than what it got wrong.\n\n**1\\. Retail VC**\n\nAfter the excesses of the Bubble, it's now considered dubious to take companies public before they have earnings. But there is nothing intrinsically wrong with that idea. Taking a company public at an early stage is simply retail VC: instead of going to venture capital firms for the last round of funding, you go to the public markets.\n\nBy the end of the Bubble, companies going public with no earnings were being derided as \"concept stocks,\" as if it were inherently stupid to invest in them. But investing in concepts isn't stupid; it's what VCs do, and the best of them are far from stupid.\n\nThe stock of a company that doesn't yet have earnings is worth _something._ It may take a while for the market to learn how to value such companies, just as it had to learn to value common stocks in the early 20th century. But markets are good at solving that kind of problem. I wouldn't be surprised if the market ultimately did a better job than VCs do now.\n\nGoing public early will not be the right plan for every company. And it can of course be disruptive\u2014by distracting the management, or by making the early employees suddenly rich. But just as the market will learn how to value startups, startups will learn how to minimize the damage of going public.\n\n**2\\. The Internet**\n\nThe Internet genuinely is a big deal. That was one reason even smart people were fooled by the Bubble. Obviously it was going to have a huge effect. Enough of an effect to triple the value of Nasdaq companies in two years? No, as it turned out. But it was hard to say for certain at the time. \\[1\\]\n\nThe same thing happened during the Mississippi and South Sea Bubbles. What drove them was the invention of organized public finance (the South Sea Company, despite its name, was really a competitor of the Bank of England). And that did turn out to be a big deal, in the long run.\n\nRecognizing an important trend turns out to be easier than figuring out how to profit from it. The mistake investors always seem to make is to take the trend too literally. Since the Internet was the big new thing, investors supposed that the more Internettish the company, the better. Hence such parodies as Pets.Com.\n\nIn fact most of the money to be made from big trends is made indirectly. It was not the railroads themselves that made the most money during the railroad boom, but the companies on either side, like Carnegie's steelworks, which made the rails, and Standard Oil, which used railroads to get oil to the East Coast, where it could be shipped to Europe.\n\nI think the Internet will have great effects, and that what we've seen so far is nothing compared to what's coming. But most of the winners will only indirectly be Internet companies; for every Google there will be ten JetBlues.\n\n**3\\. Choices**\n\nWhy will the Internet have great effects? The general argument is that new forms of communication always do. They happen rarely (till industrial times there were just speech, writing, and printing), but when they do, they always cause a big splash.\n\nThe specific argument, or one of them, is the Internet gives us more choices. In the \"old\" economy, the high cost of presenting information to people meant they had only a narrow range of options to choose from. The tiny, expensive pipeline to consumers was tellingly named \"the channel.\" Control the channel and you could feed them what you wanted, on your terms. And it was not just big corporations that depended on this principle. So, in their way, did labor unions, the traditional news media, and the art and literary establishments. Winning depended not on doing good work, but on gaining control of some bottleneck.\n\nThere are signs that this is changing. Google has over 82 million unique users a month and annual revenues of about three billion dollars. \\[2\\] And yet have you ever seen a Google ad? Something is going on here.\n\nAdmittedly, Google is an extreme case. It's very easy for people to switch to a new search engine. It costs little effort and no money to try a new one, and it's easy to see if the results are better. And so Google doesn't _have_ to advertise. In a business like theirs, being the best is enough.\n\nThe exciting thing about the Internet is that it's shifting everything in that direction. The hard part, if you want to win by making the best stuff, is the beginning. Eventually everyone will learn by word of mouth that you're the best, but how do you survive to that point? And it is in this crucial stage that the Internet has the most effect. First, the Internet lets anyone find you at almost zero cost. Second, it dramatically speeds up the rate at which reputation spreads by word of mouth. Together these mean that in many fields the rule will be: Build it, and they will come. Make something great and put it online. That is a big change from the recipe for winning in the past century.\n\n**4\\. Youth**\n\nThe aspect of the Internet Bubble that the press seemed most taken with was the youth of some of the startup founders. This too is a trend that will last. There is a huge standard deviation among 26 year olds. Some are fit only for entry level jobs, but others are ready to rule the world if they can find someone to handle the paperwork for them.\n\nA 26 year old may not be very good at managing people or dealing with the SEC. Those require experience. But those are also commodities, which can be handed off to some lieutenant."}
+{"context_length":8192,"context_depth":0.0,"secret_code":"n\/a","copy":0,"context":"People who worry about the increasing gap between rich and poor generally look back on the mid twentieth century as a golden age. In those days we had a large number of high-paying union manufacturing jobs that boosted the median income. I wouldn't quite call the high-paying union job a myth, but I think people who dwell on it are reading too much into it.\n\nOddly enough, it was working with startups that made me realize where the high-paying union job came from. In a rapidly growing market, you don't worry too much about efficiency. It's more important to grow fast. If there's some mundane problem getting in your way, and there's a simple solution that's somewhat expensive, just take it and get on with more important things. EBay didn't win by paying less for servers than their competitors.\n\nDifficult though it may be to imagine now, manufacturing was a growth industry in the mid twentieth century. This was an era when small firms making everything from cars to candy were getting consolidated into a new kind of corporation with national reach and huge economies of scale. You had to grow fast or die. Workers were for these companies what servers are for an Internet startup. A reliable supply was more important than low cost.\n\nIf you looked in the head of a 1950s auto executive, the attitude must have been: sure, give 'em whatever they ask for, so long as the new model isn't delayed.\n\nIn other words, those workers were not paid what their work was worth. Circumstances being what they were, companies would have been stupid to insist on paying them so little.\n\nIf you want a less controversial example of this phenomenon, ask anyone who worked as a consultant building web sites during the Internet Bubble. In the late nineties you could get paid huge sums of money for building the most trivial things. And yet does anyone who was there have any expectation those days will ever return? I doubt it. Surely everyone realizes that was just a temporary aberration.\n\nThe era of labor unions seems to have been the same kind of aberration, just spread over a longer period, and mixed together with a lot of ideology that prevents people from viewing it with as cold an eye as they would something like consulting during the Bubble.\n\nBasically, unions were just Razorfish.\n\nPeople who think the labor movement was the creation of heroic union organizers have a problem to explain: why are unions shrinking now? The best they can do is fall back on the default explanation of people living in fallen civilizations. Our ancestors were giants. The workers of the early twentieth century must have had a moral courage that's lacking today.\n\nIn fact there's a simpler explanation. The early twentieth century was just a fast-growing startup overpaying for infrastructure. And we in the present are not a fallen people, who have abandoned whatever mysterious high-minded principles produced the high-paying union job. We simply live in a time when the fast-growing companies overspend on different things.If you have a US startup called X and you don't have x.com, you should probably change your name.\n\nThe reason is not just that people can't find you. For companies with mobile apps, especially, having the right domain name is not as critical as it used to be for getting users. The problem with not having the .com of your name is that it signals weakness. Unless you're so big that your reputation precedes you, a marginal domain suggests you're a marginal company. Whereas (as Stripe shows) having x.com signals strength even if it has no relation to what you do.\n\nEven good founders can be in denial about this. Their denial derives from two very powerful forces: identity, and lack of imagination.\n\nX is what we _are_, founders think. There's no other name as good. Both of which are false.\n\nYou can fix the first by stepping back from the problem. Imagine you'd called your company something else. If you had, surely you'd be just as attached to that name as you are to your current one. The idea of switching to your current name would seem repellent. \\[[1](#f1n)\\]\n\nThere's nothing intrinsically great about your current name. Nearly all your attachment to it comes from it being attached to you. \\[[2](#f1n)\\]\n\nThe way to neutralize the second source of denial, your inability to think of other potential names, is to acknowledge that you're bad at naming. Naming is a completely separate skill from those you need to be a good founder. You can be a great startup founder but hopeless at thinking of names for your company.\n\nOnce you acknowledge that, you stop believing there is nothing else you could be called. There are lots of other potential names that are as good or better; you just can't think of them.\n\nHow do you find them? One answer is the default way to solve problems you're bad at: find someone else who can think of names. But with company names there is another possible approach. It turns out almost any word or word pair that is not an obviously bad name is a sufficiently good one, and the number of such domains is so large that you can find plenty that are cheap or even untaken. So make a list and try to buy some. That's what [Stripe](http:\/\/www.quora.com\/How-did-Stripe-come-up-with-its-name?share=1) did. (Their search also turned up parse.com, which their friends at Parse took.)\n\nThe reason I know that naming companies is a distinct skill orthogonal to the others you need in a startup is that I happen to have it. Back when I was running YC and did more office hours with startups, I would often help them find new names. 80% of the time we could find at least one good name in a 20 minute office hour slot.\n\nNow when I do office hours I have to focus on more important questions, like what the company is doing. I tell them when they need to change their name. But I know the power of the forces that have them in their grip, so I know most won't listen. \\[[3](#f1n)\\]\n\nThere are of course examples of startups that have succeeded without having the .com of their name. There are startups that have succeeded despite any number of different mistakes. But this mistake is less excusable than most. It's something that can be fixed in a couple days if you have sufficient discipline to acknowledge the problem.\n\n100% of the top 20 YC companies by valuation have the .com of their name. 94% of the top 50 do. But only 66% of companies in the current batch have the .com of their name. Which suggests there are lessons ahead for most of the rest, one way or another.\n\n**Notes**\n\n\\[1\\] Incidentally, this thought experiment works for [nationality and religion](identity.html) too.\n\n\\[2\\] The liking you have for a name that has become part of your identity manifests itself not directly, which would be easy to discount, but as a collection of specious beliefs about its intrinsic qualities. (This too is true of nationality and religion as well.)\n\n\\[3\\] Sometimes founders know it's a problem that they don't have the .com of their name, but delusion strikes a step later in the belief that they'll be able to buy it despite having no evidence it's for sale. Don't believe a domain is for sale unless the owner has already told you an asking price.\n\n**Thanks** to Sam Altman, Jessica Livingston, and Geoff Ralston for reading drafts of this.A couple months ago I got an email from a recruiter asking if I was interested in being a \"technologist in residence\" at a new venture capital fund. I think the idea was to play Karl Rove to the VCs' George Bush.\n\nI considered it for about four seconds. Work for a VC fund? Ick.\n\nOne of my most vivid memories from our startup is going to visit Greylock, the famous Boston VCs. They were the most arrogant people I've met in my life. And I've met a lot of arrogant people. \\[1\\]\n\nI'm not alone in feeling this way, of course. Even a VC friend of mine dislikes VCs. \"Assholes,\" he says.\n\nBut lately I've been learning more about how the VC world works, and a few days ago it hit me that there's a reason VCs are the way they are. It's not so much that the business attracts jerks, or even that the power they wield corrupts them."}
+{"context_length":8192,"context_depth":0.0,"secret_code":"n\/a","copy":0,"context":"The biggest component in most investors' opinion of you is the opinion of other investors. Which is of course a recipe for exponential growth. When one investor wants to invest in you, that makes other investors want to, which makes others want to, and so on.\n\nSometimes inexperienced founders mistakenly conclude that manipulating these forces is the essence of fundraising. They hear stories about stampedes to invest in successful startups, and think it's therefore the mark of a successful startup to have this happen. But actually the two are not that highly correlated. Lots of startups that cause stampedes end up flaming out (in extreme cases, partly as a result of the stampede), and lots of very successful startups were only moderately popular with investors the first time they raised money.\n\nSo the point of this essay is not to explain how to create a stampede, but merely to explain the forces that generate them. These forces are always at work to some degree in fundraising, and they can cause surprising situations. If you understand them, you can at least avoid being surprised.\n\nOne reason investors like you more when other investors like you is that you actually become a better investment. Raising money decreases the risk of failure. Indeed, although investors hate it, you are for this reason justified in raising your valuation for later investors. The investors who invested when you had no money were taking more risk, and are entitled to higher returns. Plus a company that has raised money is literally more valuable. After you raise the first million dollars, the company is at least a million dollars more valuable, because it's the same company as before, plus it has a million dollars in the bank. \\[[1](#f1n)\\]\n\nBeware, though, because later investors so hate to have the price raised on them that they resist even this self-evident reasoning. Only raise the price on an investor you're comfortable with losing, because some will angrily refuse. \\[[2](#f2n)\\]\n\nThe second reason investors like you more when you've had some success at fundraising is that it makes you more confident, and an investors' opinion of [you](convince.html) is the foundation of their opinion of your company. Founders are often surprised how quickly investors seem to know when they start to succeed at raising money. And while there are in fact lots of ways for such information to spread among investors, the main vector is probably the founders themselves. Though they're often clueless about technology, most investors are pretty good at reading people. When fundraising is going well, investors are quick to sense it in your increased confidence. (This is one case where the average founder's inability to remain poker-faced works to your advantage.)\n\nBut frankly the most important reason investors like you more when you've started to raise money is that they're bad at judging startups. Judging startups is hard even for the best investors. The mediocre ones might as well be flipping coins. So when mediocre investors see that lots of other people want to invest in you, they assume there must be a reason. This leads to the phenomenon known in the Valley as the \"hot deal,\" where you have more interest from investors than you can handle.\n\nThe best investors aren't influenced much by the opinion of other investors. It would only dilute their own judgment to average it together with other people's. But they are indirectly influenced in the practical sense that interest from other investors imposes a deadline. This is the fourth way in which offers beget offers. If you start to get far along the track toward an offer with one firm, it will sometimes provoke other firms, even good ones, to make up their minds, lest they lose the deal.\n\nUnless you're a wizard at negotiation (and if you're not sure, you're not) be very careful about exaggerating this to push a good investor to decide. Founders try this sort of thing all the time, and investors are very sensitive to it. If anything oversensitive. But you're safe so long as you're telling the truth. If you're getting far along with investor B, but you'd rather raise money from investor A, you can tell investor A that this is happening. There's no manipulation in that. You're genuinely in a bind, because you really would rather raise money from A, but you can't safely reject an offer from B when it's still uncertain what A will decide.\n\nDo not, however, tell A who B is. VCs will sometimes ask which other VCs you're talking to, but you should never tell them. Angels you can sometimes tell about other angels, because angels cooperate more with one another. But if VCs ask, just point out that they wouldn't want you telling other firms about your conversations, and you feel obliged to do the same for any firm you talk to. If they push you, point out that you're inexperienced at fundraising \u2014 which is always a safe card to play \u2014 and you feel you have to be extra cautious. \\[[3](#f3n)\\]\n\nWhile few startups will experience a stampede of interest, almost all will at least initially experience the other side of this phenomenon, where the herd remains clumped together at a distance. The fact that investors are so much influenced by other investors' opinions means you always start out in something of a hole. So don't be demoralized by how hard it is to get the first commitment, because much of the difficulty comes from this external force. The second will be easier.\n\n**Notes**\n\n\\[1\\] An accountant might say that a company that has raised a million dollars is no richer if it's convertible debt, but in practice money raised as convertible debt is little different from money raised in an equity round.\n\n\\[2\\] Founders are often surprised by this, but investors can get very emotional. Or rather indignant; that's the main emotion I've observed; but it is very common, to the point where it sometimes causes investors to act against their own interests. I know of one investor who invested in a startup at a $15 million valuation cap. Earlier he'd had an opportunity to invest at a $5 million cap, but he refused because a friend who invested earlier had been able to invest at a $3 million cap.\n\n\\[3\\] If an investor pushes you hard to tell them about your conversations with other investors, is this someone you want as an investor?\n\n**Thanks** to Paul Buchheit, Jessica Livingston, Geoff Ralston, and Garry Tan for reading drafts of this.In high school I decided I was going to study philosophy in college. I had several motives, some more honorable than others. One of the less honorable was to shock people. College was regarded as job training where I grew up, so studying philosophy seemed an impressively impractical thing to do. Sort of like slashing holes in your clothes or putting a safety pin through your ear, which were other forms of impressive impracticality then just coming into fashion.\n\nBut I had some more honest motives as well. I thought studying philosophy would be a shortcut straight to wisdom. All the people majoring in other things would just end up with a bunch of domain knowledge. I would be learning what was really what.\n\nI'd tried to read a few philosophy books. Not recent ones; you wouldn't find those in our high school library. But I tried to read Plato and Aristotle. I doubt I believed I understood them, but they sounded like they were talking about something important. I assumed I'd learn what in college.\n\nThe summer before senior year I took some college classes. I learned a lot in the calculus class, but I didn't learn much in Philosophy 101. And yet my plan to study philosophy remained intact. It was my fault I hadn't learned anything. I hadn't read the books we were assigned carefully enough. I'd give Berkeley's _Principles of Human Knowledge_ another shot in college. Anything so admired and so difficult to read must have something in it, if one could only figure out what.\n\nTwenty-six years later, I still don't understand Berkeley. I have a nice edition of his collected works. Will I ever read it? Seems unlikely."}
+{"context_length":8192,"context_depth":0.0,"secret_code":"n\/a","copy":0,"context":"I don't think Apple realizes how badly the App Store approval process is broken. Or rather, I don't think they realize how much it matters that it's broken.\n\nThe way Apple runs the App Store has harmed their reputation with programmers more than anything else they've ever done. Their reputation with programmers used to be great. It used to be the most common complaint you heard about Apple was that their fans admired them too uncritically. The App Store has changed that. Now a lot of programmers have started to see Apple as evil.\n\nHow much of the goodwill Apple once had with programmers have they lost over the App Store? A third? Half? And that's just so far. The App Store is an ongoing karma leak.\n\n\\* \\* \\*\n\nHow did Apple get into this mess? Their fundamental problem is that they don't understand software.\n\nThey treat iPhone apps the way they treat the music they sell through iTunes. Apple is the channel; they own the user; if you want to reach users, you do it on their terms. The record labels agreed, reluctantly. But this model doesn't work for software. It doesn't work for an intermediary to own the user. The software business learned that in the early 1980s, when companies like VisiCorp showed that although the words \"software\" and \"publisher\" fit together, the underlying concepts don't. Software isn't like music or books. It's too complicated for a third party to act as an intermediary between developer and user. And yet that's what Apple is trying to be with the App Store: a software publisher. And a particularly overreaching one at that, with fussy tastes and a rigidly enforced house style.\n\nIf software publishing didn't work in 1980, it works even less now that software development has evolved from a small number of big releases to a constant stream of small ones. But Apple doesn't understand that either. Their model of product development derives from hardware. They work on something till they think it's finished, then they release it. You have to do that with hardware, but because software is so easy to change, its design can benefit from evolution. The standard way to develop applications now is to launch fast and iterate. Which means it's a disaster to have long, random delays each time you release a new version.\n\nApparently Apple's attitude is that developers should be more careful when they submit a new version to the App Store. They would say that. But powerful as they are, they're not powerful enough to turn back the evolution of technology. Programmers don't use launch-fast-and-iterate out of laziness. They use it because it yields the best results. By obstructing that process, Apple is making them do bad work, and programmers hate that as much as Apple would.\n\nHow would Apple like it if when they discovered a serious bug in OS\u00a0X, instead of releasing a software update immediately, they had to submit their code to an intermediary who sat on it for a month and then rejected it because it contained an icon they didn't like?\n\nBy breaking software development, Apple gets the opposite of what they intended: the version of an app currently available in the App Store tends to be an old and buggy one. One developer told me:\n\n> As a result of their process, the App Store is full of half-baked applications. I make a new version almost every day that I release to beta users. The version on the App Store feels old and crappy. I'm sure that a lot of developers feel this way: One emotion is \"I'm not really proud about what's in the App Store\", and it's combined with the emotion \"Really, it's Apple's fault.\"\n\nAnother wrote:\n\n> I believe that they think their approval process helps users by ensuring quality. In reality, bugs like ours get through all the time and then it can take 4-8 weeks to get that bug fix approved, leaving users to think that iPhone apps sometimes just don't work. Worse for Apple, these apps work just fine on other platforms that have immediate approval processes.\n\nActually I suppose Apple has a third misconception: that all the complaints about App Store approvals are not a serious problem. They must hear developers complaining. But partners and suppliers are always complaining. It would be a bad sign if they weren't; it would mean you were being too easy on them. Meanwhile the iPhone is selling better than ever. So why do they need to fix anything?\n\nThey get away with maltreating developers, in the short term, because they make such great hardware. I just bought a new 27\" iMac a couple days ago. It's fabulous. The screen's too shiny, and the disk is surprisingly loud, but it's so beautiful that you can't make yourself care.\n\nSo I bought it, but I bought it, for the first time, with misgivings. I felt the way I'd feel buying something made in a country with a bad human rights record. That was new. In the past when I bought things from Apple it was an unalloyed pleasure. Oh boy! They make such great stuff. This time it felt like a Faustian bargain. They make such great stuff, but they're such assholes. Do I really want to support this company?\n\n\\* \\* \\*\n\nShould Apple care what people like me think? What difference does it make if they alienate a small minority of their users?\n\nThere are a couple reasons they should care. One is that these users are the people they want as employees. If your company seems evil, the best programmers won't work for you. That hurt Microsoft a lot starting in the 90s. Programmers started to feel sheepish about working there. It seemed like selling out. When people from Microsoft were talking to other programmers and they mentioned where they worked, there were a lot of self-deprecating jokes about having gone over to the dark side. But the real problem for Microsoft wasn't the embarrassment of the people they hired. It was the people they never got. And you know who got them? Google and Apple. If Microsoft was the Empire, they were the Rebel Alliance. And it's largely because they got more of the best people that Google and Apple are doing so much better than Microsoft today.\n\nWhy are programmers so fussy about their employers' morals? Partly because they can afford to be. The best programmers can work wherever they want. They don't have to work for a company they have qualms about.\n\nBut the other reason programmers are fussy, I think, is that evil begets stupidity. An organization that wins by exercising power starts to lose the ability to win by doing better work. And it's not fun for a smart person to work in a place where the best ideas aren't the ones that win. I think the reason Google embraced \"Don't be evil\" so eagerly was not so much to impress the outside world as to inoculate themselves against arrogance. \\[[1](#f1n)\\]\n\nThat has worked for Google so far. They've become more bureaucratic, but otherwise they seem to have held true to their original principles. With Apple that seems less the case. When you look at the famous [1984 ad](http:\/\/www.uriahcarpenter.info\/1984.html) now, it's easier to imagine Apple as the dictator on the screen than the woman with the hammer. \\[[2](#f2n)\\] In fact, if you read the dictator's speech it sounds uncannily like a prophecy of the App Store.\n\n> We have triumphed over the unprincipled dissemination of facts.  \n>   \n> We have created, for the first time in all history, a garden of pure ideology, where each worker may bloom secure from the pests of contradictory and confusing truths.\n\nThe other reason Apple should care what programmers think of them is that when you sell a platform, developers make or break you. If anyone should know this, Apple should. VisiCalc made the Apple II.\n\nAnd programmers build applications for the platforms they use. Most applications\u2014most startups, probably\u2014grow out of personal projects. Apple itself did. Apple made microcomputers because that's what Steve Wozniak wanted for himself. He couldn't have afforded a minicomputer. \\[[3](#f3n)\\] Microsoft likewise started out making interpreters for little microcomputers because Bill Gates and Paul Allen were interested in using them. It's a rare startup that doesn't build something the founders use."}
+{"context_length":8192,"context_depth":0.0,"secret_code":"n\/a","copy":0,"context":"_(This article describes the spam-filtering techniques used in the spamproof web-based mail reader we built to exercise [Arc](arc.html). An improved algorithm is described in [Better Bayesian Filtering](better.html).)_\n\nI think it's possible to stop spam, and that content-based filters are the way to do it. The Achilles heel of the spammers is their message. They can circumvent any other barrier you set up. They have so far, at least. But they have to deliver their message, whatever it is. If we can write software that recognizes their messages, there is no way they can get around that.\n\n\\_ \\_ \\_\n\nTo the recipient, spam is easily recognizable. If you hired someone to read your mail and discard the spam, they would have little trouble doing it. How much do we have to do, short of AI, to automate this process?\n\nI think we will be able to solve the problem with fairly simple algorithms. In fact, I've found that you can filter present-day spam acceptably well using nothing more than a Bayesian combination of the spam probabilities of individual words. Using a slightly tweaked (as described below) Bayesian filter, we now miss less than 5 per 1000 spams, with 0 false positives.\n\nThe statistical approach is not usually the first one people try when they write spam filters. Most hackers' first instinct is to try to write software that recognizes individual properties of spam. You look at spams and you think, the gall of these guys to try sending me mail that begins \"Dear Friend\" or has a subject line that's all uppercase and ends in eight exclamation points. I can filter out that stuff with about one line of code.\n\nAnd so you do, and in the beginning it works. A few simple rules will take a big bite out of your incoming spam. Merely looking for the word \"click\" will catch 79.7% of the emails in my spam corpus, with only 1.2% false positives.\n\nI spent about six months writing software that looked for individual spam features before I tried the statistical approach. What I found was that recognizing that last few percent of spams got very hard, and that as I made the filters stricter I got more false positives.\n\nFalse positives are innocent emails that get mistakenly identified as spams. For most users, missing legitimate email is an order of magnitude worse than receiving spam, so a filter that yields false positives is like an acne cure that carries a risk of death to the patient.\n\nThe more spam a user gets, the less likely he'll be to notice one innocent mail sitting in his spam folder. And strangely enough, the better your spam filters get, the more dangerous false positives become, because when the filters are really good, users will be more likely to ignore everything they catch.\n\nI don't know why I avoided trying the statistical approach for so long. I think it was because I got addicted to trying to identify spam features myself, as if I were playing some kind of competitive game with the spammers. (Nonhackers don't often realize this, but most hackers are very competitive.) When I did try statistical analysis, I found immediately that it was much cleverer than I had been. It discovered, of course, that terms like \"virtumundo\" and \"teens\" were good indicators of spam. But it also discovered that \"per\" and \"FL\" and \"ff0000\" are good indicators of spam. In fact, \"ff0000\" (html for bright red) turns out to be as good an indicator of spam as any pornographic term.\n\n\\_ \\_ \\_\n\nHere's a sketch of how I do statistical filtering. I start with one corpus of spam and one of nonspam mail. At the moment each one has about 4000 messages in it. I scan the entire text, including headers and embedded html and javascript, of each message in each corpus. I currently consider alphanumeric characters, dashes, apostrophes, and dollar signs to be part of tokens, and everything else to be a token separator. (There is probably room for improvement here.) I ignore tokens that are all digits, and I also ignore html comments, not even considering them as token separators.\n\nI count the number of times each token (ignoring case, currently) occurs in each corpus. At this stage I end up with two large hash tables, one for each corpus, mapping tokens to number of occurrences.\n\nNext I create a third hash table, this time mapping each token to the probability that an email containing it is a spam, which I calculate as follows \\[1\\]: (let ((g (\\* 2 (or (gethash word good) 0))) (b (or (gethash word bad) 0))) (unless (< (+ g b) 5) (max .01 (min .99 (float (\/ (min 1 (\/ b nbad)) (+ (min 1 (\/ g ngood)) (min 1 (\/ b nbad))))))))) where word is the token whose probability we're calculating, good and bad are the hash tables I created in the first step, and ngood and nbad are the number of nonspam and spam messages respectively.\n\nI explained this as code to show a couple of important details. I want to bias the probabilities slightly to avoid false positives, and by trial and error I've found that a good way to do it is to double all the numbers in good. This helps to distinguish between words that occasionally do occur in legitimate email and words that almost never do. I only consider words that occur more than five times in total (actually, because of the doubling, occurring three times in nonspam mail would be enough). And then there is the question of what probability to assign to words that occur in one corpus but not the other. Again by trial and error I chose .01 and .99. There may be room for tuning here, but as the corpus grows such tuning will happen automatically anyway.\n\nThe especially observant will notice that while I consider each corpus to be a single long stream of text for purposes of counting occurrences, I use the number of emails in each, rather than their combined length, as the divisor in calculating spam probabilities. This adds another slight bias to protect against false positives.\n\nWhen new mail arrives, it is scanned into tokens, and the most interesting fifteen tokens, where interesting is measured by how far their spam probability is from a neutral .5, are used to calculate the probability that the mail is spam. If probs is a list of the fifteen individual probabilities, you calculate the [combined](naivebayes.html) probability thus: (let ((prod (apply #'\\* probs))) (\/ prod (+ prod (apply #'\\* (mapcar #'(lambda (x) (- 1 x)) probs))))) One question that arises in practice is what probability to assign to a word you've never seen, i.e. one that doesn't occur in the hash table of word probabilities. I've found, again by trial and error, that .4 is a good number to use. If you've never seen a word before, it is probably fairly innocent; spam words tend to be all too familiar.\n\nThere are examples of this algorithm being applied to actual emails in an appendix at the end.\n\nI treat mail as spam if the algorithm above gives it a probability of more than .9 of being spam. But in practice it would not matter much where I put this threshold, because few probabilities end up in the middle of the range.\n\n\\_ \\_ \\_\n\nOne great advantage of the statistical approach is that you don't have to read so many spams. Over the past six months, I've read literally thousands of spams, and it is really kind of demoralizing. Norbert Wiener said if you compete with slaves you become a slave, and there is something similarly degrading about competing with spammers. To recognize individual spam features you have to try to get into the mind of the spammer, and frankly I want to spend as little time inside the minds of spammers as possible.\n\nBut the real advantage of the Bayesian approach, of course, is that you know what you're measuring. Feature-recognizing filters like SpamAssassin assign a spam \"score\" to email. The Bayesian approach assigns an actual probability. The problem with a \"score\" is that no one knows what it means. The user doesn't know what it means, but worse still, neither does the developer of the filter."}
+{"context_length":8192,"context_depth":0.0,"secret_code":"n\/a","copy":0,"context":"_(This essay is derived from a talk at Google.)_\n\nA few weeks ago I found to my surprise that I'd been granted four [patents](http:\/\/paulgraham.infogami.com\/blog\/morepatents). This was all the more surprising because I'd only applied for three. The patents aren't mine, of course. They were assigned to Viaweb, and became Yahoo's when they bought us. But the news set me thinking about the question of software patents generally.\n\nPatents are a hard problem. I've had to advise most of the startups we've funded about them, and despite years of experience I'm still not always sure I'm giving the right advice.\n\nOne thing I do feel pretty certain of is that if you're against software patents, you're against patents in general. Gradually our machines consist more and more of software. Things that used to be done with levers and cams and gears are now done with loops and trees and closures. There's nothing special about physical embodiments of control systems that should make them patentable, and the software equivalent not.\n\nUnfortunately, patent law is inconsistent on this point. Patent law in most countries says that algorithms aren't patentable. This rule is left over from a time when \"algorithm\" meant something like the Sieve of Eratosthenes. In 1800, people could not see as readily as we can that a great many patents on mechanical objects were really patents on the algorithms they embodied.\n\nPatent lawyers still have to pretend that's what they're doing when they patent algorithms. You must not use the word \"algorithm\" in the title of a patent application, just as you must not use the word \"essays\" in the title of a book. If you want to patent an algorithm, you have to frame it as a computer system executing that algorithm. Then it's mechanical; phew. The default euphemism for algorithm is \"system and method.\" Try a patent search for that phrase and see how many results you get.\n\nSince software patents are no different from hardware patents, people who say \"software patents are evil\" are saying simply \"patents are evil.\" So why do so many people complain about software patents specifically?\n\nI think the problem is more with the patent office than the concept of software patents. Whenever software meets government, bad things happen, because software changes fast and government changes slow. The patent office has been overwhelmed by both the volume and the novelty of applications for software patents, and as a result they've made a lot of mistakes.\n\nThe most common is to grant patents that shouldn't be granted. To be patentable, an invention has to be more than new. It also has to be non-obvious. And this, especially, is where the USPTO has been dropping the ball. Slashdot has an icon that expresses the problem vividly: a knife and fork with the words \"patent pending\" superimposed.\n\nThe scary thing is, this is the _only_ icon they have for patent stories. Slashdot readers now take it for granted that a story about a patent will be about a bogus patent. That's how bad the problem has become.\n\nThe problem with Amazon's notorious one-click patent, for example, is not that it's a software patent, but that it's obvious. Any online store that kept people's shipping addresses would have implemented this. The reason Amazon did it first was not that they were especially smart, but because they were one of the earliest sites with enough clout to force customers to log in before they could buy something. \\[[1](#f1n)\\]\n\nWe, as hackers, know the USPTO is letting people patent the knives and forks of our world. The problem is, the USPTO are not hackers. They're probably good at judging new inventions for casting steel or grinding lenses, but they don't understand software yet.\n\nAt this point an optimist would be tempted to add \"but they will eventually.\" Unfortunately that might not be true. The problem with software patents is an instance of a more general one: the patent office takes a while to understand new technology. If so, this problem will only get worse, because the rate of technological change seems to be increasing. In thirty years, the patent office may understand the sort of things we now patent as software, but there will be other new types of inventions they understand even less.\n\nApplying for a patent is a negotiation. You generally apply for a broader patent than you think you'll be granted, and the examiners reply by throwing out some of your claims and granting others. So I don't really blame Amazon for applying for the one-click patent. The big mistake was the patent office's, for not insisting on something narrower, with real technical content. By granting such an over-broad patent, the USPTO in effect slept with Amazon on the first date. Was Amazon supposed to say no?\n\nWhere Amazon went over to the dark side was not in applying for the patent, but in enforcing it. A lot of companies (Microsoft, for example) have been granted large numbers of preposterously over-broad patents, but they keep them mainly for defensive purposes. Like nuclear weapons, the main role of big companies' patent portfolios is to threaten anyone who attacks them with a counter-suit. Amazon's suit against Barnes & Noble was thus the equivalent of a nuclear first strike.\n\nThat suit probably hurt Amazon more than it helped them. Barnes & Noble was a lame site; Amazon would have crushed them anyway. To attack a rival they could have ignored, Amazon put a lasting black mark on their own reputation. Even now I think if you asked hackers to free-associate about Amazon, the one-click patent would turn up in the first ten topics.\n\nGoogle clearly doesn't feel that merely holding patents is evil. They've applied for a lot of them. Are they hypocrites? Are patents evil?\n\nThere are really two variants of that question, and people answering it often aren't clear in their own minds which they're answering. There's a narrow variant: is it bad, given the current legal system, to apply for patents? and also a broader one: is it bad that the current legal system allows patents?\n\nThese are separate questions. For example, in preindustrial societies like medieval Europe, when someone attacked you, you didn't call the police. There were no police. When attacked, you were supposed to fight back, and there were conventions about how to do it. Was this wrong? That's two questions: was it wrong to take justice into your own hands, and was it wrong that you had to? We tend to say yes to the second, but no to the first. If no one else will defend you, you have to defend yourself. \\[[2](#f2n)\\]\n\nThe situation with patents is similar. Business is a kind of ritualized warfare. Indeed, it evolved from actual warfare: most early traders switched on the fly from merchants to pirates depending on how strong you seemed. In business there are certain rules describing how companies may and may not compete with one another, and someone deciding that they're going to play by their own rules is missing the point. Saying \"I'm not going to apply for patents just because everyone else does\" is not like saying \"I'm not going to lie just because everyone else does.\" It's more like saying \"I'm not going to use TCP\/IP just because everyone else does.\" Oh yes you are.\n\nA closer comparison might be someone seeing a hockey game for the first time, realizing with shock that the players were _deliberately_ bumping into one another, and deciding that one would on no account be so rude when playing hockey oneself.\n\nHockey allows checking. It's part of the game. If your team refuses to do it, you simply lose. So it is in business. Under the present rules, patents are part of the game.\n\nWhat does that mean in practice? We tell the startups we fund not to worry about infringing patents, because startups rarely get sued for patent infringement. There are only two reasons someone might sue you: for money, or to prevent you from competing with them. Startups are too poor to be worth suing for money. And in practice they don't seem to get sued much by competitors, either."}
+{"context_length":8192,"context_depth":0.0,"secret_code":"n\/a","copy":0,"context":"The Segway hasn't delivered on its initial promise, to put it mildly. There are several reasons why, but one is that people don't want to be seen riding them. Someone riding a Segway looks like a dork.\n\nMy friend Trevor Blackwell built [his own Segway](http:\/\/tlb.org\/#scooter), which we called the Segwell. He also built a one-wheeled version, [the Eunicycle](http:\/\/tlb.org\/#eunicycle), which looks exactly like a regular unicycle till you realize the rider isn't pedaling. He has ridden them both to downtown Mountain View to get coffee. When he rides the Eunicycle, people smile at him. But when he rides the Segwell, they shout abuse from their cars: \"Too lazy to walk, ya fuckin homo?\"\n\nWhy do Segways provoke this reaction? The reason you look like a dork riding a Segway is that you look _smug_. You don't seem to be working hard enough.\n\nSomeone riding a motorcycle isn't working any harder. But because he's sitting astride it, he seems to be making an effort. When you're riding a Segway you're just standing there. And someone who's being whisked along while seeming to do no work \u2014 someone in a sedan chair, for example \u2014 can't help but look smug.\n\nTry this thought experiment and it becomes clear: imagine something that worked like the Segway, but that you rode with one foot in front of the other, like a skateboard. That wouldn't seem nearly as uncool.\n\nSo there may be a way to capture more of the market Segway hoped to reach: make a version that doesn't look so easy for the rider. It would also be helpful if the styling was in the tradition of skateboards or bicycles rather than medical devices.\n\nCuriously enough, what got Segway into this problem was that the company was itself a kind of Segway. It was too easy for them; they were too successful raising money. If they'd had to grow the company gradually, by iterating through several versions they sold to real users, they'd have learned pretty quickly that people looked stupid riding them. Instead they had enough to work in secret. They had focus groups aplenty, I'm sure, but they didn't have the people yelling insults out of cars. So they never realized they were zooming confidently down a blind alley.One of the biggest things holding people back from doing great work is the fear of making something lame. And this fear is not an irrational one. Many great projects go through a stage early on where they don't seem very impressive, even to their creators. You have to push through this stage to reach the great work that lies beyond. But many people don't. Most people don't even reach the stage of making something they're embarrassed by, let alone continue past it. They're too frightened even to start.\n\nImagine if we could turn off the fear of making something lame. Imagine how much more we'd do.\n\nIs there any hope of turning it off? I think so. I think the habits at work here are not very deeply rooted.\n\nMaking new things is itself a new thing for us as a species. It has always happened, but till the last few centuries it happened so slowly as to be invisible to individual humans. And since we didn't need customs for dealing with new ideas, we didn't develop any.\n\nWe just don't have enough experience with early versions of ambitious projects to know how to respond to them. We judge them as we would judge more finished work, or less ambitious projects. We don't realize they're a special case.\n\nOr at least, most of us don't. One reason I'm confident we can do better is that it's already starting to happen. There are already a few places that are living in the future in this respect. Silicon Valley is one of them: an unknown person working on a strange-sounding idea won't automatically be dismissed the way they would back home. In Silicon Valley, people have learned how dangerous that is.\n\nThe right way to deal with new ideas is to treat them as a challenge to your imagination \ufffd not just to have lower standards, but to [switch polarity](altair.html) entirely, from listing the reasons an idea won't work to trying to think of ways it could. That's what I do when I meet people with new ideas. I've become quite good at it, but I've had a lot of practice. Being a partner at Y Combinator means being practically immersed in strange-sounding ideas proposed by unknown people. Every six months you get thousands of new ones thrown at you and have to sort through them, knowing that in a world with a power-law distribution of outcomes, it will be painfully obvious if you miss the needle in this haystack. Optimism becomes urgent.\n\nBut I'm hopeful that, with time, this kind of optimism can become widespread enough that it becomes a social custom, not just a trick used by a few specialists. It is after all an extremely lucrative trick, and those tend to spread quickly.\n\nOf course, inexperience is not the only reason people are too harsh on early versions of ambitious projects. They also do it to seem clever. And in a field where the new ideas are risky, like startups, those who dismiss them are in fact more likely to be right. Just not when their predictions are [weighted by outcome](swan.html).\n\nBut there is another more sinister reason people dismiss new ideas. If you try something ambitious, many of those around you will hope, consciously or unconsciously, that you'll fail. They worry that if you try something ambitious and succeed, it will put you above them. In some countries this is not just an individual failing but part of the national culture.\n\nI wouldn't claim that people in Silicon Valley overcome these impulses because they're morally better. \\[[1](#f1n)\\] The reason many hope you'll succeed is that they hope to rise with you. For investors this incentive is particularly explicit. They want you to succeed because they hope you'll make them rich in the process. But many other people you meet can hope to benefit in some way from your success. At the very least they'll be able to say, when you're famous, that they've known you since way back.\n\nBut even if Silicon Valley's encouraging attitude is rooted in self-interest, it has over time actually grown into a sort of benevolence. Encouraging startups has been practiced for so long that it has become a custom. Now it just seems that that's what one does with startups.\n\nMaybe Silicon Valley is too optimistic. Maybe it's too easily fooled by impostors. Many less optimistic journalists want to believe that. But the lists of impostors they cite are suspiciously short, and plagued with asterisks. \\[[2](#f2n)\\] If you use revenue as the test, Silicon Valley's optimism seems better tuned than the rest of the world's. And because it works, it will spread.\n\nThere's a lot more to new ideas than new startup ideas, of course. The fear of making something lame holds people back in every field. But Silicon Valley shows how quickly customs can evolve to support new ideas. And that in turn proves that dismissing new ideas is not so deeply rooted in human nature that it can't be unlearnt.\n\n\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\n\nUnfortunately, if you want to do new things, you'll face a force more powerful than other people's skepticism: your own skepticism. You too will judge your early work too harshly. How do you avoid that?\n\nThis is a difficult problem, because you don't want to completely eliminate your horror of making something lame. That's what steers you toward doing good work. You just want to turn it off temporarily, the way a painkiller temporarily turns off pain.\n\nPeople have already discovered several techniques that work. Hardy mentions two in _A Mathematician's Apology_:\n\n> Good work is not done by \"humble\" men. It is one of the first duties of a professor, for example, in any subject, to exaggerate a little both the importance of his subject and his importance in it.\n\nIf you overestimate the importance of what you're working on, that will compensate for your mistakenly harsh judgment of your initial results."}
+{"context_length":8192,"context_depth":0.0,"secret_code":"n\/a","copy":0,"context":"I recently told applicants to Y Combinator that the best advice I could give for getting in, per word, was\n\n> Explain what you've learned from users.\n\nThat tests a lot of things: whether you're paying attention to users, how well you understand them, and even how much they need what you're making.\n\nAfterward I asked myself the same question. What have I learned from YC's users, the startups we've funded?\n\nThe first thing that came to mind was that most startups have the same problems. No two have exactly the same problems, but it's surprising how much the problems remain the same, regardless of what they're making. Once you've advised 100 startups all doing different things, you rarely encounter problems you haven't seen before.\n\nThis fact is one of the things that makes YC work. But I didn't know it when we started YC. I only had a few data points: our own startup, and those started by friends. It was a surprise to me how often the same problems recur in different forms. Many later stage investors might never realize this, because later stage investors might not advise 100 startups in their whole career, but a YC partner will get this much experience in the first year or two.\n\nThat's one advantage of funding large numbers of early stage companies rather than smaller numbers of later-stage ones. You get a lot of data. Not just because you're looking at more companies, but also because more goes wrong.\n\nBut knowing (nearly) all the problems startups can encounter doesn't mean that advising them can be automated, or reduced to a formula. There's no substitute for individual office hours with a YC partner. Each startup is unique, which means they have to be advised by specific partners who know them well. \\[[1](#f1n)\\]\n\nWe learned that the hard way, in the notorious \"batch that broke YC\" in the summer of 2012. Up till that point we treated the partners as a pool. When a startup requested office hours, they got the next available slot posted by any partner. That meant every partner had to know every startup. This worked fine up to 60 startups, but when the batch grew to 80, everything broke. The founders probably didn't realize anything was wrong, but the partners were confused and unhappy because halfway through the batch they still didn't know all the companies yet. \\[[2](#f2n)\\]\n\nAt first I was puzzled. How could things be fine at 60 startups and broken at 80? It was only a third more. Then I realized what had happened. We were using an _O(n2)_ algorithm. So of course it blew up.\n\nThe solution we adopted was the classic one in these situations. We sharded the batch into smaller groups of startups, each overseen by a dedicated group of partners. That fixed the problem, and has worked fine ever since. But the batch that broke YC was a powerful demonstration of how individualized the process of advising startups has to be.\n\nAnother related surprise is how bad founders can be at realizing what their problems are. Founders will sometimes come in to talk about some problem, and we'll discover another much bigger one in the course of the conversation. For example (and this case is all too common), founders will come in to talk about the difficulties they're having raising money, and after digging into their situation, it turns out the reason is that the company is doing badly, and investors can tell. Or founders will come in worried that they still haven't cracked the problem of user acquisition, and the reason turns out to be that their product isn't good enough. There have been times when I've asked \"Would you use this yourself, if you hadn't built it?\" and the founders, on thinking about it, said \"No.\" Well, there's the reason you're having trouble getting users.\n\nOften founders know what their problems are, but not their relative importance. \\[[3](#f3n)\\] They'll come in to talk about three problems they're worrying about. One is of moderate importance, one doesn't matter at all, and one will kill the company if it isn't addressed immediately. It's like watching one of those horror movies where the heroine is deeply upset that her boyfriend cheated on her, and only mildly curious about the door that's mysteriously ajar. You want to say: never mind about your boyfriend, think about that door! Fortunately in office hours you can. So while startups still die with some regularity, it's rarely because they wandered into a room containing a murderer. The YC partners can warn them where the murderers are.\n\nNot that founders listen. That was another big surprise: how often founders don't listen to us. A couple weeks ago I talked to a partner who had been working for YC for a couple batches and was starting to see the pattern. \"They come back a year later,\" she said, \"and say 'We wish we'd listened to you.'\"\n\nIt took me a long time to figure out why founders don't listen. At first I thought it was mere stubbornness. That's part of the reason, but another and probably more important reason is that so much about startups is [counterintuitive](before.html). And when you tell someone something counterintuitive, what it sounds to them is wrong. So the reason founders don't listen to us is that they don't _believe_ us. At least not till experience teaches them otherwise. \\[[4](#f4n)\\]\n\nThe reason startups are so counterintuitive is that they're so different from most people's other experiences. No one knows what it's like except those who've done it. Which is why YC partners should usually have been founders themselves. But strangely enough, the counterintuitiveness of startups turns out to be another of the things that make YC work. If it weren't counterintuitive, founders wouldn't need our advice about how to do it.\n\nFocus is doubly important for early stage startups, because not only do they have a hundred different problems, they don't have anyone to work on them except the founders. If the founders focus on things that don't matter, there's no one focusing on the things that do. So the essence of what happens at YC is to figure out which problems matter most, then cook up ideas for solving them \u2014 ideally at a resolution of a week or less \u2014 and then try those ideas and measure how well they worked. The focus is on action, with measurable, near-term results.\n\nThis doesn't imply that founders should rush forward regardless of the consequences. If you correct course at a high enough frequency, you can be simultaneously decisive at a micro scale and tentative at a macro scale. The result is a somewhat winding path, but executed very rapidly, like the path a running back takes downfield. And in practice there's less backtracking than you might expect. Founders usually guess right about which direction to run in, especially if they have someone experienced like a YC partner to bounce their hypotheses off. And when they guess wrong, they notice fast, because they'll talk about the results at office hours the next week. \\[[5](#f5n)\\]\n\nA small improvement in navigational ability can make you a lot faster, because it has a double effect: the path is shorter, and you can travel faster along it when you're more certain it's the right one. That's where a lot of YC's value lies, in helping founders get an extra increment of focus that lets them move faster. And since moving fast is the essence of a startup, YC in effect makes startups more startup-like.\n\nSpeed defines startups. Focus enables speed. YC improves focus.\n\nWhy are founders uncertain about what to do? Partly because startups almost by definition are doing something new, which means no one knows how to do it yet, or in most cases even what \"it\" is. Partly because startups are so counterintuitive generally. And partly because many founders, especially young and ambitious ones, have been trained to win the wrong way. That took me years to figure out. The educational system in most countries trains you to win by [hacking the test](lesson.html) instead of actually doing whatever it's supposed to measure. But that stops working when you start a startup. So part of what YC does is to retrain founders to stop trying to hack the test."}
+{"context_length":8192,"context_depth":0.0,"secret_code":"n\/a","copy":0,"context":"Here's a simple trick for getting more people to read what you write: write in spoken language.\n\nSomething comes over most people when they start writing. They write in a different language than they'd use if they were talking to a friend. The sentence structure and even the words are different. No one uses \"pen\" as a verb in spoken English. You'd feel like an idiot using \"pen\" instead of \"write\" in a conversation with a friend.\n\nThe last straw for me was a sentence I read a couple days ago:\n\n> The mercurial Spaniard himself declared: \"After Altamira, all is decadence.\"\n\nIt's from Neil Oliver's _A History of Ancient Britain_. I feel bad making an example of this book, because it's no worse than lots of others. But just imagine calling Picasso \"the mercurial Spaniard\" when talking to a friend. Even one sentence of this would raise eyebrows in conversation. And yet people write whole books of it.\n\nOk, so written and spoken language are different. Does that make written language worse?\n\nIf you want people to read and understand what you write, yes. Written language is more complex, which makes it more work to read. It's also more formal and distant, which gives the reader's attention permission to drift. But perhaps worst of all, the complex sentences and fancy words give you, the writer, the false impression that you're saying more than you actually are.\n\nYou don't need complex sentences to express complex ideas. When specialists in some abstruse topic talk to one another about ideas in their field, they don't use sentences any more complex than they do when talking about what to have for lunch. They use different words, certainly. But even those they use no more than necessary. And in my experience, the harder the subject, the more informally experts speak. Partly, I think, because they have less to prove, and partly because the harder the ideas you're talking about, the less you can afford to let language get in the way.\n\nInformal language is the athletic clothing of ideas.\n\nI'm not saying spoken language always works best. Poetry is as much music as text, so you can say things you wouldn't say in conversation. And there are a handful of writers who can get away with using fancy language in prose. And then of course there are cases where writers don't want to make it easy to understand what they're saying\u2014in corporate announcements of bad news, for example, or at the more [bogus](https:\/\/scholar.google.com\/scholar?hl=en&as_sdt=1,5&q=transgression+narrative+postmodern+gender) end of the humanities. But for nearly everyone else, spoken language is better.\n\nIt seems to be hard for most people to write in spoken language. So perhaps the best solution is to write your first draft the way you usually would, then afterward look at each sentence and ask \"Is this the way I'd say this if I were talking to a friend?\" If it isn't, imagine what you would say, and use that instead. After a while this filter will start to operate as you write. When you write something you wouldn't say, you'll hear the clank as it hits the page.\n\nBefore I publish a new essay, I read it out loud and fix everything that doesn't sound like conversation. I even fix bits that are phonetically awkward; I don't know if that's necessary, but it doesn't cost much.\n\nThis trick may not always be enough. I've seen writing so far removed from spoken language that it couldn't be fixed sentence by sentence. For cases like that there's a more drastic solution. After writing the first draft, try explaining to a friend what you just wrote. Then replace the draft with what you said to your friend.\n\nPeople often tell me how much my essays sound like me talking. The fact that this seems worthy of comment shows how rarely people manage to write in spoken language. Otherwise everyone's writing would sound like them talking.\n\nIf you simply manage to write in spoken language, you'll be ahead of 95% of writers. And it's so easy to do: just don't let a sentence through unless it's the way you'd say it to a friend.\n\n**Thanks** to Patrick Collison and Jessica Livingston for reading drafts of this.Technology tends to separate normal from natural. Our bodies weren't designed to eat the foods that people in rich countries eat, or to get so little exercise. There may be a similar problem with the way we work: a normal job may be as bad for us intellectually as white flour or sugar is for us physically.\n\nI began to suspect this after spending several years working with startup founders. I've now worked with over 200 of them, and I've noticed a definite difference between programmers working on their own startups and those working for large organizations. I wouldn't say founders seem happier, necessarily; starting a startup can be very stressful. Maybe the best way to put it is to say that they're happier in the sense that your body is happier during a long run than sitting on a sofa eating doughnuts.\n\nThough they're statistically abnormal, startup founders seem to be working in a way that's more natural for humans.\n\nI was in Africa last year and saw a lot of animals in the wild that I'd only seen in zoos before. It was remarkable how different they seemed. Particularly lions. Lions in the wild seem about ten times more alive. They're like different animals. I suspect that working for oneself feels better to humans in much the same way that living in the wild must feel better to a wide-ranging predator like a lion. Life in a zoo is easier, but it isn't the life they were designed for.\n\n**Trees**\n\nWhat's so unnatural about working for a big company? The root of the problem is that humans weren't meant to work in such large groups.\n\nAnother thing you notice when you see animals in the wild is that each species thrives in groups of a certain size. A herd of impalas might have 100 adults; baboons maybe 20; lions rarely 10. Humans also seem designed to work in groups, and what I've read about hunter-gatherers accords with research on organizations and my own experience to suggest roughly what the ideal size is: groups of 8 work well; by 20 they're getting hard to manage; and a group of 50 is really unwieldy. \\[[1](#f1n)\\]\n\nWhatever the upper limit is, we are clearly not meant to work in groups of several hundred. And yet\u2014for reasons having more to do with technology than human nature\u2014a great many people work for companies with hundreds or thousands of employees.\n\nCompanies know groups that large wouldn't work, so they divide themselves into units small enough to work together. But to coordinate these they have to introduce something new: bosses.\n\nThese smaller groups are always arranged in a tree structure. Your boss is the point where your group attaches to the tree. But when you use this trick for dividing a large group into smaller ones, something strange happens that I've never heard anyone mention explicitly. In the group one level up from yours, your boss represents your entire group. A group of 10 managers is not merely a group of 10 people working together in the usual way. It's really a group of groups. Which means for a group of 10 managers to work together as if they were simply a group of 10 individuals, the group working for each manager would have to work as if they were a single person\u2014the workers and manager would each share only one person's worth of freedom between them.\n\nIn practice a group of people are never able to act as if they were one person. But in a large organization divided into groups in this way, the pressure is always in that direction. Each group tries its best to work as if it were the small group of individuals that humans were designed to work in. That was the point of creating it. And when you propagate that constraint, the result is that each person gets freedom of action in inverse proportion to the size of the entire tree. \\[[2](#f2n)\\]\n\nAnyone who's worked for a large organization has felt this. You can feel the difference between working for a company with 100 employees and one with 10,000, even if your group has only 10 people."}