A Conversation with Paul Graham - Moderated by Geoff Ralston

Well, thank you for coming this morning. We are trying something a little bit different this startup school year. We are not just having our weekly two lectures, but we are having some conversations with notable people, and I couldn't be happier to have one of the most notable startup people here. He'll get embarrassed when I say anything complimentary, so I'll try to avoid it. Paul Graham, my friend, and the founder of Y Combinator back in 2005, and we're just going to talk, and hopefully it won't be boring.

Paul: Yes.

Jeff: So you and I got to know each other strangely at a— it was kind of a startup at the time, Yahoo, back in the late '90s.

Paul: '98?

Jeff: Yeah. Guys know he wrote Yahoo Mail? Some of it?

Paul: Yeah, Y Combinator has the people who wrote Yahoo Mail and Gmail. That is weird, right? Paul Buchheit, who actually has more of a claim on actually writing Gmail than I do on writing Yahoo. I wrote some important parts of Yahoo Mail, but Paul was kind of by himself doing Gmail almost, and we were a little more intentional with Yahoo Mail and had a team creating something called RocketMail.

Jeff: But it is a little weird that both of us ended up at YC years later. But Paul was at Yahoo; I was at Yahoo because we created something called RocketMail, and we were purchased in '97. Paul’s company was bought in '98. How can you? You had this weird idea for a product for something you thought people would want in the mid-'90s, which had to do with like SaaS before SaaS was a thing. How did you come up with this idea for Viaweb, the company that Yahoo bought?

Paul: Okay, so what we made was something that is now called a web app. I believe at the time I was the first one, and that's why the company was called Viaweb, because it worked via the web. The way I remember very vividly—we thought back then everybody thought that writing software was identical with writing software that ran on the client. And we’re still, for those of you who remember the mid-'90s, the client meant Windows, right? And like, we knew how to write software for UNIX, and we liked UNIX; we did not know how to write software for Windows, but we knew enough to know that we did not want to learn. So we were really, really highly motivated to figure out how to write software without having to write software to run on Windows.

Paul: And so we thought to ourselves, well, maybe—well, you know, the way I first thought of it actually was that you would have this— you would have it was an online store builder, right? And I thought that you could send it updates by email. That was—I thought you could send it updates by email, right? And it would change your website. And then I remember thinking, well, if you could cost us for shopping carts or whatever, if you can use SMTP, why not HTTP to update your website? Whoa! You could cut the software could run on the server, and you could control it by clicking on links in the browser. Would that even work? Right? But I remember I was like sleeping on a mattress on the floor in Robert’s spare bedroom. And I remember I sat up in bed when I thought of this, like the letter L, right? It was actually like one of those scenes in a movie. I'm like, oh my God, we could avoid having to write software on Windows at all. We never at that point were thinking about the benefits of web apps, which are great as it turned out; we were just thinking we could like actually do this without ever having to learn pointing Windows.

Jeff: You did, yeah.

Paul: At AOL, that like you could deploy then without having to ship software. Did none of that avoiding Windows. We could actually write—thank you, Gates—never have to learn Windows, because like the Netscape people had already gone through the pain of making Netscape work on Windows. Hey, just like that was like a hole onto Windows, you know, and people could get through the hole and click on links in the browser and control the software. And so we immediately—we weren't even sure this would work because it was so weird, but we immediately sat down and tried to write this really, really clunky website builder that worked by clicking on links. It was terrible, but it did actually work; it could make a website, and I'm like, holy [__], this actually works!

Jeff: That's weird that like an innovative idea like that just sort of comes out of nowhere. I mean, it doesn’t seem innovative now because all websites work that way, but the first one—like that’s going from 0 to 1. How do you like—it's very common, I mean.

Paul: Right, very common for ideas in retrospect to seem obvious, and you and Matt, you sort of historians who weren't there, sort of straighten out all the kinks in the development, and it just seems like, oh, they had this, you know—it's like this idea. I always hate it when people represent startup ideas as light bulbs, you know. Because as well as being the most clichéd metaphor imaginable, it's also false. You don't just like have this idea and then realize it, right? No, it's more like you have this sort of inkling that something—you could do something that somebody else hasn't done before, and it's probably a bad idea, but you're too lazy to learn Windows, and so you want to do it that way, right?

Paul: Or like who would—or like Zuck starting Facebook, you know? Just like, let’s see what happens, right? The way implausible if this would ever be a startup, which is good for you all. Hopefully, some of your ideas seem really implausible, right? I love it when ideas seem implausible in the right way. Over the years, I mean, I already had just from being a hacker, sort of a good nose for this sort of thing, but it has become refined over all these years of dealing with startups.

Paul: And when people have—I was talking to some people yesterday, and they were talking about two possible things they could do, and one of them seemed so—it seemed if not like it was just sort of taking advantage of something that seems like it shouldn't even be possible, right? Like making this software run on the server and just like never making software that seemed to the users to run on their computer but actually was just talking to them through the browser, that kind of naughtiness. And then I said like, pick that one, because it seems— it seems so outrageous; it seems so hilarious. It seems like there are all these ways we have talking about what's a good idea, and when you try to match up ideas with them, it's always sort of a complicated—you know, it's like, is this idea naughty or is it—is it—is it, like, not quite intuitive or not so javea? And it's really always hard to take any particular idea and say, is that the one with potential? Well, you can't tell. In fact, that's another thing—it's not just that the outcomes of startups are hard to predict. I think to some degree they're actually indeterminate; there’s a huge amount of luck involved.

Paul: You know, and so I think even if Y Combinator were as good at picking startups as you could possibly get—I mean, in fact, you have like 150 startups per batch, of which maybe five will become giant or something like that. Hopefully high—that would be good.

Jeff: Yeah, well, it depends on your definition of giant, right?

Paul: It's a power-law distribution, where like, if for some definitions of giant, one of it—one is taking any batches, we'll think if five are any definition of giant, yeah, we'll be happy.

Jeff: So talking about—I like—oh yeah, the indeterminacy of—yes, even if you were able to pick perfectly that doesn't mean you could get the batches down to only five startups. You probably have to pick twenty, I think, even if you were perfect.

Paul: So it's just—it's hard. It’s hard to know. There’s this strange mix of extraordinary hard work, an idea that actually has something to it, but is not obvious. It's hard for founders to know.

Jeff: Yeah, you don't know yourselves.

Paul: You don't know yourselves. I mean, I mean, you don't know, comma, yourselves—not that you don't know yourself. So that it's also true. So, and that's why it's good to be a hacker, because if you're some business person, you know, and you think, I am going to work on business opportunities, right? Well, I don't know if this is a business opportunity, so I'm not gonna work on it, whereas someone who's a hacker who's just doing things for the fun of it or because they don't like Windows, yeah, laziness to this is actually in some cases really helpful—at least some kinds of—some kinds of later.

Paul: I'm a little scared about this meme escaping into the wild, right? Can't you just see—well, yeah, you just see—it's truly—they said laziness is good, he said from the sofa, as he ate another doughnut, you know? It would be better if people thought laziness wasn't good because the glaze enos—the kind of—the people to you—that’s good. They do, and they would do anyway, even if they thought it's weird.

Jeff: Yeah. Well, using laziness not in the normal sense of laziness, right? Because if you're lazy, there's no such thing as a lazy founder that builds a big company. There's—but a certain sort of—it’s sort of a reticence to do something for some reason. I'm calling it lazy waiting; it’s not really, it seems like sort of slaps—it makes you, you know, if you have a natural hatred for gratuitous slaps, like, then what that produces is elegant solutions.

Paul: So, in 1995-ish, I was trying every single idea I could possibly try to start some internet business, and I was doing it all by myself, which was awful. And what I didn't know how to do was to find people to do it with. And I kind of ran into a bunch of people, and I had no clue whether they would be good people to start companies with. And I was thinking about your experience. You said you were like sleeping on Robert’s—well, you started, you started Viaweb with two fascinating people: Robert Morris and Trevor Blackwell.

Paul: Robert Morris and Trevor Blackwell. That's a—that was a weird mix of names, anyway. How did you pick them as co-founders? How did that happen? And when did you decide they would be good co-founders? Were they good co-founders?

Paul: Alright, so, yeah, right. They're extremely—worked—they were extremely truculent co-founders who were very good at programming. But I remember Robert was like not into the whole startup thing the entire way; like, whenever he was on the board, and whenever we got—and whenever you get an acquisition offer, the board has to at least consider it, right? And so we would get these crap lowball acquisition offers—we'll give you, like, two million dollars in our stock, right? Robert would say, like, yes, no, no. Robert would say, like, we would vote on it, and Robert would always say, well, you know, I have to be honest here; like, this would at least mean we could stop working on this thing. And so, yes, I would take this deal, right? I made a deal with Robert that if he ever made a million dollars out of Viaweb, he would get an earring.

Paul: Robert's not the kind of guy who would enjoy having an earring, let me tell you. And so, as soon as the deal closed, me and Trevor frog-marched him to the place in Harvard Square where people get their things pierced. Never heard this before! He got him an earring. There are photographs on the internet of Robert Morris with an earring that he didn't keep. Not for long; we should have specified that. I never thought—I never thought to specify, you know, I thought you'd actually get the million dollars.

Jeff: No, no, I really actually did. But I thought he would wear it for a bit longer. When his fiancée saw it, she fell on the floor.

Jeff: Why did you choose—why did I try killing Trayvon first, okay? So I chose Robert because he was my co-conspirator in everything. And I don't mean—I was always the lead, and he was just the co-conspirator when he did things. I would be his co-conspirator too. So, like, the internet worm, if—19—you guys know Robert Morris's is pretty famous all on his own.

Jeff: Yeah, he invented buffer overflow, and I remember when he told me about it, and I said, wow, what a cool idea! We should totally do that. He was the first famous hacker who really got in trouble, I think.

Paul: Yeah, really in trouble. He was, the way it’s true, he was the first person to be prosecuted under the Computer Fraud and Abuse Act of 1986. So if you're looking for a co-founder, look at people who have been prosecuted because it works.

Jeff: User, he is actually a convicted felon. Okay, this is another meme we don't necessarily want to escape out there—that convicted felons are what you want to look for as co-founders. They're convicted for—but maybe it was funny because the FBI, like, law enforcement kind of does things by the book, right? Like, imagination is a big thing in Sherlock Holmes stories, but not in actual everyday law enforcement, and so they have these—the FBI agent told me, like, they have these motives. They look for sex, drugs, money, revenge, right? And so, like, Robert did it out of curiosity, and that's just—it’s just not on the list.

Paul: Hey, they really had a hard time figuring out what the hell was going on. I think even within the government, there were lots of people who didn’t understand what was going on. So back to Robert and Trevor—woo-hoo! Oh, so why did I pick Robert and Trevor? Yeah. Robert, Robert, Robert. I picked because I would do everything with him, right? We had all these schemes. And so you knew you could work with him.

Paul: I mean, I could work with him when I knew he was really good. He was a really good programmer; he could program as fast as he could type. Of course, this was in C, which is a lot of very verbose language, but he could program as fast as he could type. He was amazing. He would, like, if he didn’t like what the operating system would do, he would edit the source and recompile it, and then it would do what he wanted. So truculence aside, he was exceptional.

Paul: Like, when he showed up at Harvard, undergrads were only allowed to have accounts on like—the official undergrad computing system off in the Science Center. And there were all these real computers in the CS department and the LaTeichin Lab, and so they wouldn't give him an account there. And so he just walked up to a machine, you know, switched it to single user with him as the super user, gave himself an account, and then switched it back.

Jeff: Hey, he was curious!

Paul: Yeah, well, he also wanted accounts on the right machines. He got kicked—this whole thing will be about Robert! Robert's such an interesting character; I shouldn't talk about him. Maybe more interesting than anything else we can talk about. He got kicked out of Harvard as an undergrad for reconnecting Harvard to the internet.

Jeff: Um, Harvard had been one of the early internet nodes? That’s so funny to like—like get in trouble for disconnecting a school from the internet?

Paul: Connecting it!

Jeff: Connecting it?

Paul: Can it happen for connecting a school? Where were we? Well, we’re still trying to figure out why the heck—oh! I should tell you, though, about how we—tell us about the disconnecting really fast!

Paul: No, no, it's okay. So we have all day? We don’t do it; no, we don’t! Okay, how long have we been—? We have enough time; we have the same watch, more or less.

Robert: Okay, it was one of the very early internet nodes, and by the time Robert arrived in as an undergrad, the connection had died from bit decay. That's like—no one noticed! That’s how unimportant the internet was back then, or rather ARPANET, as it was then called. Very early internet, right?

Paul: So RTM spent one whole semester working on getting Harvard reconnected.

Jeff: RTM is Robert’s?

Paul: Yes, yes, Robert, and he, like, didn’t do any work for any of his classes, and he got such bad grades he got kicked out for a year. So we actually used that later as a recruiting technique because we knew there might be other people who got kicked out for spending all their time on some project, and so we actually went around Harvard and posted up this poster saying, did you get kicked out working on some project of your own? We'd like to hire you.

Jeff: Yeah!

Paul: I think there's probably something made tons of money off of options ’cause we got acquired—you enjoy life out of school so much, I don't think he'd go back until he was 25.

Jeff: So, alright, why did you—why Trevor?

Paul: Well, so you picked Robert. Robert was the guy you did everything with, and he was great, okay. So after we'd been doing—before the wave, when you picked Robert, did you know he was gonna be a pain in the ass in so many ways as you like, or was that a surprise to you?

Paul: Well, I knew he was a pain in the ass. I didn't know the company was gonna be so hard, and it would bring out so much pain in him. So you didn’t realize how hard—

Jeff: Tell me the way we got Trevor was very indicative, yeah. So, a month in, we'd been working on this company for a month, and Robert rebels! He says, we've been working on this company for a whole month, and it’s still not done! And I thought, hmm, maybe we need more programmers. So, I said, alright, Robert, who's the smartest person you know in grad school? And he said, Trevor, right? And I said, Trevor, right? And Trevor's one of those people who seems a lot smarter than you. Seems so, but Trevor is actually super smart.

Paul: So we went and recruited Trevor, and he said, oh OK, that's—that's a really good Trevor. But almost not a Canadian— Trevor has two modes.

Jeff: Oh, okay! And oh, I'm sorry, which is what he does when he does something that breaks everything, which was very common.

Paul: Very common. And so we got Trevor, I recruited Trevor to work on the thing; he does nothing for two weeks, at least as far as I can tell! He just disappears, and then, two weeks later, he says, come into my office! And he's rewritten all our goddamn software in Smalltalk—typical!

Jeff: Yes!

Paul: So I said, alright, Trevor, we’re not using that, but do this instead, right? And so Trevor was like this super productive hacking monster, but you didn't know that in advance; you just trusted Robert when Robert told you he was really smart? Well, if Robert said he was the smartest, then he's definitely the smartest!

Paul: But there’s kind of a deep lesson in that, and how you pick, whether it's employees or co-founders—it's you find someone you trust and then find someone they trust.

Jeff: Yeah, although here's an interesting point—that works better for some qualities than others. Like intelligence, intelligent people can judge other intelligent people, but trustworthy people cannot judge other trustworthy people. In fact, trustworthy people are often fooled by untrustworthy people.

Paul: So like Kate Quarto is like our architect who did—this is like super trustworthy, but she's always fooled by easily sneaky schemers, you know, who take advantage of her. So I guess that means you take recommendations from people, but then you have to add your own—no, no, you can totally trust, but if Robert says someone's smart and I disagree, I think I must be wrong, right?

Jeff: Yeah, but you need to—this works for some qualities better than others; let’s hold, right? You want them to be trustworthy when they join you as a co-founder, so you have to add that for the trustworthiness.

Jeff: Yes, this whole thing of like networking doesn’t work; it doesn’t—you have to have someone you can judge whether people are trustworthy, which I'm—I am not even good at; I am not good at the person who does that for Y Combinator is Jessica, the social radar, the secret power behind everything.

Paul: So you did hire people at Viaweb too? Yes. What did you learn about hiring in your like your first startup? Like, so you learned that if they were getting lousy grades, your start working on a project that seemed good, which is kind of counterintuitive.

Paul: A lot of things we seem to do at YC end up being counterintuitive. Startups in general are very, very counterintuitive, and so that's actually why YC exists. If it was obvious what to do in starting a startup, we wouldn't—YC couldn't really add much by teaching people how to do it, right? And so I often say—I often, in fact, practically every day if I do office hours, I'm reminded of this—that like what Y Combinator does is tell founders things that they ignore, right?

Paul: So we tell people don’t hire too fast, and then they go off and hire too fast, and then they come back later and say, oh, I wish we’d listened. But all the things we tell people—we tell people the counterintuitive stuff, not the obvious stuff. And then what counterintuitive means is it sounds wrong, and so they go with their gut and do the wrong thing instead, and then hopefully catch the mistake in time.

Jeff: Why is it that startups are so counterintuitive?

Paul: You know, that would be—if that would be very much writing an essay about that would be worth writing an essay about why startups are so counterintuitive. I don't know—I don't know.

Jeff: I don't know. I could come up with some theories, but I have a feeling the answer is so interestingly complicated that it would be unlikely to come—

Paul: And it's really just—we were just talking about one of the things that—you did write an essay about, which is somewhat counterintuitive. When you start a startup and you're like, okay now I'm gonna grow, and then you write a whole bunch of stuff, but you've already—software that helps you grow—and you're at a whole—YC says don’t do that!

Paul: Do things that don't scale in the beginning.

Jeff: Yeah, do things that don't scale. Because growth is—maybe you can start by saying what it means to do things that don't scale.

Paul: Because I'm afraid the best thing you did not already—I'm sure you've all read all, as well as I say, but this is actually very important because most of you all have early startups. And when you do an early startup, it's become our mantra at YC, and I’ve seen it again and again—you should not try to do too much in the beginning. You should do things that don’t scale.

Paul: What doing things that don’t scale means specifically is doing things in a sort of handmade, artisanal, painstaking way that you feel like—yeah, it would be great if you could do things that way forever—but you—in the back of your mind, you think to yourself, well, there’s no way we can keep doing this and become giant!

Jeff: Right, and what do things that don’t scale means is do those things early on anyway because you know if you don’t do them, you’ll never be big, and you’ve got nothing to lose. And also, you learn a lot from it!

Paul: You learn a lot. And so I discovered a lot of this things that we teach startups at Y Combinator, or you teach startups at Y Combinator—I’m retired, I don’t know if you know that—but a lot of things you guys teach startups are things that I hit myself and didn’t realize that they were actually common startup lessons.

Paul: So one of them is doing things very manually for your early customers. Like, it's so important to get early customers that if you have to do a ton of manual stuff, that's okay; you'll learn a lot from it.

Jeff: Did you guys do manual stuff?

Paul: Oh, yeah, totally! And we thought, oh, we’re doing it wrong, this is so lame. And in fact, it was exactly the right thing. We made an online store builder; you could build a store on the internet and sell stuff, and we would go to customers and say, would you like to use our easy online store builder? And they would say no, and we would say, but you want an online store, right? And they would say, yeah, and they would say, well, what if we used our software for you to make an online store, and then you could have it? Would that be good? And they’d say, alright!

Jeff: And so you were like, we can’t get anyone to use our software, but at least they’re willing to let us use it for them!

Paul: Exactly! And it seemed—it seemed so lame; I learned so much [__] about like direct marketing. As it turns out, it’s called—we were members of the DMA, the Direct Marketing Association. That’s what, you know, in every business card—carrying member of the DMA, we were in every business.

Jeff: They have a name for the business; that’s not the name.

Paul: Outsiders call it, like, fast food.

Jeff: No, no, no, they don't call it that; they call it fast casual within the industry, right? And in the catalog business, that used to be called the direct marketing industry, right? So that's what the DMA meant back in those days—the catalog business. We signed up for every catalog; you know, his catalogs you get in the mail? We would just like write a wide right and ask for more!

Paul: Well, you had this like bookshelf full of every catalog ever! Yeah, it seems almost like if you were building a new search engine. And what you ought to do is have your box out there, and then just take all the searches manually and get results and throw them out there and see what happens!

Jeff: I’m trying to think what we did for mail that was manual in the beginning. We—we sort of had a manual ad server; like, we didn’t really have ad serving technology; we just throw—like, I wrote it, so it wasn’t really good, and we would just throw them in there.

Jeff: There's a deep lesson there for people, though: like, what happens if you think you know the solution, and you build lots of stuff for it? You’re gonna be wrong, because you just don’t know, and having to use our software myself made it much better because I would be using our software to make someone’s website, right? And I was also the guy who wrote the site builder.

Paul: And so I would be using it and think to myself, this is inconvenient. It’s like going in the middle of building their site; I would change the software. I mean ship, as in like MV, and like that's like—that's the great thing about server-based software: Ship is the UNIX MV command!

Jeff: So well, know CP; I would want to keep a copy! But—but I would like change the software in the middle of using it and then go back to working on their website, and boy, did that make that— that made the software much better that I had to use it!

Paul: Yeah, so, um, so picking co-founders, you want to pick the best people who you can somehow believe are trustworthy. It's probably helpful for a lot of people; I think a lot of people taking the class are sole founders now.

Jeff: Yeah, why is it that being a sole founder is so difficult?

Paul: Wow, there are so many different reasons. I think the hardest part is more L—there’s no one—there's like if you have multiple people, they keep one another going when things are going badly, and there’s no one to cheer you up.

Jeff: That's what's hard for me. I remember '95—I’d be like, I know this internet thing is gonna be real, but damn!

Paul: Oh, so you were a sole founder?

Jeff: Yeah, yeah. That’s why his hair is greater than mine!

Paul: Yeah! Like being a sole founder’s hard; it's hard because there's a million reasons why what you're doing isn’t going to work, and mostly, you’re right. So it’s easy to convince yourself. You used to say this, and I always remember—you used to say, look, you have two choices as a founder!

Jeff: Do you remember what you would say after that?

Paul: Well, there's so many different cases—there are versions of this. What I would say, you could either quit or get rich!

Jeff: Mmm!

Paul: Yeah, die or get rich! Like the company's a lot of versions—well, that's what happens with startups. Mostly, the company either dies, or it's something—you succeed; you get rich, and you know getting rich can mean you know you have a nice little business where you're the boss and you have 20 or 50 employees.

Jeff: Or it could mean your Airbnb, right?

Paul: They're both good outcomes— not as good for your investors necessarily, but for you, it’s really good. And doing things that don’t scale, that’s hot, you know? And internalizing what that really means, I was thinking about, you know, one of the—it's the sort of things that don’t scale. We talked about at YC all the time; we say you’re supposed to— for success—you talk to users and write code. You—you talked—the flower.

Paul: And if you think about it, talking to users, especially as sort of the founder, doesn’t really scale very well, but that’s what it forces you to do. If you have to do everything that doesn’t scale, it forces you to learn, right? In the beginning, you only have ten customers; you want to grow to 20%. You want to go 10% next week, 10% weeks—an ambitious goal—you only got to get one more customer; you can go out and do that very manually, right? And then next week you have eleven customers, you have to get 1.1 customers, right? So this is basically one, right? You just keep going out there and doing things manually, and as long as your growth rate is good, that doesn’t matter how small the number is because a grow— a constant growth rate means exponential growth, and that means the base number will soon take care of itself.

Jeff: It’s kind of thinking about what makes companies good, what makes companies fail, and I was thinking about maybe how we could talk a little bit about some of those aspects of how people can kind of figure out how they're doing.

Jeff: Here we have thousands of companies out in startup school and trying to figure out what to do next and whether they’re doing the right thing and whether they’re failing and whether they’re on the right path. What makes companies fail most of the time is poor execution by the founders, right?

Paul: We often talk to startups who are worried about competitors, and one good event—one advantage of Y Combinator having funded so many companies is you have a really large data set; and like, how many companies have been killed by competitors? One more than one out of nineteen hundred, right?

Paul: So I tell startups that basically you have the same protection against competitors that light aircraft have against crashing into other light aircraft when flying through clouds. You know what the protection is? Space is large!

Jeff: Hey, these little Cessnas don’t have any radar in them! You know, you can’t see in clouds! This falls under the rubric of counterintuitive advice because one of the things that happens all the time is people come up and say, oh my god, there’s another new company! What should I do? You know, that’s in exactly my space. What should I do, right?

Paul: And it’s sort of like if you’re running the hundred meters, right, and suddenly there appears another lane with another runner in it, what should you do? You run as fast as you can just like you presumably were, or hey, if they're better than you, they'll win, and if not, then you'll win!

Jeff: And that's a good reminder! Could everyone please silence your phones if you haven't already?

Paul: Yeah, what else? Welcome—like one of the things we do a lot is try to judge Wright companies and founders Nick who’s gonna succeed, and you guys are doing the same thing except you’re trying to pick yourselves. Hopefully, what is it that then makes startup succeed?

Paul: Well, you have to start with a good founding team, ideally—not just one person, that’s my quote. You can do it with one person.

Jeff: Well, I couldn't.

Paul: I ended up having to join other people. Like, I actually think—oh, that's majority; that was not rock, you know? I was—no way, like, I couldn't do it; I'm not built that way. I think most people aren't built that way.

Jeff: That’s my questions about Robert and Trevor. You know, it's just so difficult. I would not try and start a startup just with just myself as the founder.

Paul: Mmm, I don't think I could do it, yeah. It's lonely enough anyway.

Jeff: Think you can, but I couldn’t. So, you need a good solid founding team who know one another well and who can work together just like stuff. Pause—there are like knowing one another well— why does that even matter?

Paul: Because what happens in a startup will shake your relationship. If there’s any sort of flakiness or uncertainty or disloyalty or whatever lingering in there, it will emerge under the stresses of a startup.

Jeff: Yeah, it’s like a marriage, only harder; you spend more time with that person.

Paul: Other differences, there's potentially—yeah, okay, so you get a good founding team; you get someone who you can stand stress with.

Paul: So there's two categories of stuff. There are two categories of answers we could give here, like there’s the answer you give to investors about how to pick startups, and then there's, and then there’s only a subset of that is useful to founders, right? So like, if like, if you— you know, you could tell investors you want to pick smart people, right? But founders are as smart as they are.

Jeff: If they party, what can we tell them—get smarter?

Paul: Right! And actually, it turns out it's not that important to be smart. It’s much, much more important to be determined.

Paul: If you imagine this hypothetical person who's like a hundred out of hundred for smart and a hundred out of a hundred for determination, right? And then you start taking away determination, you know, it doesn’t take very long before you have this sort of ineffectual but brilliant person. Whereas if you take someone who’s like super, super determined and you start taking away smartness bit by bit, I mean, eventually you get to some guy who owns a lot of taxi medallions but he’s still rich, right? Or like a trash hauling business or something like that—but like, you can take away a lot of smart.

Paul: Yeah, take a lot of smart and take away a bunch of ethics at the same time, and then you get that guy, right? He could be President! Yeah! I’ve been thinking about what, like, Zuck thought would—Trump would actually teach an important lesson.

Jeff: Yeah!

Paul: Yeah, for startups, it’s all about determination. I was thinking about these axes we think about, and those are two, like the intelligence axis and the determination axis, and obviously, if you were really intelligent, really determined, that’s awesome.

Jeff: By the way, only one person has to be super determined, right?

Paul: Right! So, like, you know, you can have one person who really wants to start the one—the company and look in person— you and Robin exactly, who after a month is like, what, we’re still working? But that depends on this tight—bond where you can drag that person along with you; otherwise, they'll go away.

Paul: But I was thinking there’s actually another axis, which is you have to have the ability somewhere on your team to—to project—uh—to think of, because your idea is gonna suck, and then it’ll suck less and less, and so you have to have the ability to think, where should we like to navigate?

Paul: Like, you have to have some sort of creativity that allows you to—to like make the right choices.

Jeff: And you know, actually, you might not have to be that creative if you care enough about users; you can just follow what what it—what will make users happy, the way a scientist follows the truth. And eventually, without much thinking on your—you—the need to grow will give you this product idea that’s actually the product—that's the result of evolution. So maybe you’re either sort of Steve Jobs and can intuit what customers need, or you just have to have the skill of actually talking to customers and understanding what they’re saying, which isn’t always obvious.

Paul: You know, Steve Jobs’s trick was he satisfied users to accept. He was the user! It’s like I don’t want to have any phone jack anymore, so no more phone jacks for everybody!

Jeff: Right?

Paul: Maybe he was just lucky that he—that his desires were generic enough that it was closer to everybody.

Jeff: Yeah, yeah, yeah. He was living in the future!

Paul: Yeah, like he didn’t really like the internet except for reading the New York Times, so he never really got that right.

Jeff: Hmm, interesting! You work at Apple at one point after you got bought?

Paul: Did for this— It was months!

Jeff: What was it like? What was Steve like? Could you tell from where you were in Apple what Steve was like?

Paul: Yeah! I got to talk to him a few times; he’s different.

Jeff: Tell us—tell us how are their useful lessons you can learn from him or Zia one off?

Paul: I wouldn’t presume; I only met with him from a handful of times, and I wouldn’t presume to know too much about them.

Jeff: Yeah, you had a lot of indirect evidence and stories and stuff like that, yeah! Like, do you ever see founders at YC you think to yourself, oh, this is like Steve Jobs?

Paul: I don't think we'd— I don't think we’d fund—Steve Jobs at YC! Well, that's not good!

Jeff: Yeah!

Paul: And the reason is especially when Steve Jobs was starting off, he was an [__]!

Jeff: Oh, I think actually, you would find him because Steve Jobs would make you find him!

Paul: Maybe! Right? Maybe we try—like we try to avoid people who who seem like they’re [__].

Paul: And I think—like, you know you wrote a whole essay on being nice—yeah, the Ronco principle, right! There are a couple couples, yeah, be nice!

Jeff: Yeah!

Paul: And called be good! Be good, right? And I don't think he would have passed the test, especially when he was younger, which is—I worry about that because it's true that it's usually good to be nice and friendly and benevolent.

Jeff: He wasn't!

Paul: Maybe if YC had caught him early, we could have helped!

Jeff: Maybe! Maybe that's part of what made him successful in the beginning? I don't know; I don't know.

Paul: You work—didn't you work at Apple at one point after you got bought? Did for this—It was months!

Jeff: What was it like? What was Steve like? Could you tell from where you were in Apple what Steve was like?

Paul: Yeah! I got to talk to him a few times; he’s different.

Jeff: Tell us—tell us how are their useful lessons you can learn from him or Zia one off?

Paul: I wouldn’t presume; I only met with him from a handful of times, and I wouldn’t presume to know too much about them.

Jeff: Yeah, you had a lot of indirect evidence and stories and stuff like that, yeah! Like, do you ever see founders at YC you think to yourself, oh, this is like Steve Jobs?

Paul: I don't think we'd— I don't think we’d fund—Steve Jobs at YC! Well, that's not good!

Jeff: Yeah!

Paul: And the reason is especially when Steve Jobs was starting off, he was an [__]!

Jeff: I think actually, you would find him because Steve Jobs would make you find him, maybe, right? Maybe we try—like we try to avoid people who seem like they’re [__].

Paul: And I think—you know, you wrote a whole essay on being nice—yeah, the Ronco principle, right. There are a couple couples, yeah, be nice.

Jeff: Yeah!

Paul: And call it be good; be good! Right? And I don't think he would have passed the test, especially when he was younger, which is— I worry about that because it's true that it's usually good to be nice and friendly and benevolent.

Jeff: He wasn’t!

Paul: Maybe if YC had caught him early, we could have helped!

Paul: Maybe! Maybe that’s part of what made him successful in the beginning? I don’t know; I don’t know.

Jeff: It worked for Donald Trump.

Paul: Donald—oh, did it?

Jeff: That’s a different kind of example, right?

Paul: Yeah, we should know what like— in business determination matters more, but like business is not identical with startups.

Paul: I want to leave these guys some time to ask questions, but I want—like, what are they gonna get wrong? What mistakes will these people make that maybe hearing them might help? It probably won't because you won't listen, but what you will get wrong is that you will not pay enough attention to users.

Paul: You will have—you will make up some idea in your own head that you will call your vision, and then you will spend a lot of time thinking about your vision in a café by yourself and build some elaborate thing without going and talking to users because that's doing sales, which is a pain in the ass, and they might say no!

Jeff: And—and you'd be way better off finding someone, anyone, who has a problem that they will pay you to fix, and fixing it, and then seeing if you can find more people like that.

Paul: Best case is if you yourself have the problem, right? But like you will not ship fast enough because you don't want to face the—you’re embarrassed to ship something unfinished, and you don't want to face the likely feedback.

Paul: It's humbling and humiliating when they tell you how shitty what you built actually is, right? But that’s the only way you get it better.

Jeff: Yeah. So you will shrink from contact with the real world or contact with your users; that's the mistake you will make.

Paul: Yeah, there's a Paul Graham-ism in there, which is all around launching. Launching before you’re ready. Tell people the story of its Reid Hoffman, actually, who harped on that one early—not me.

Jeff: Yeah.

Paul: He—there’s I think he said if you’re not embarrassed with what you launched, rather than you launched too late.

Jeff: I remember when I was—I was joining Y Combinator, and at the same time, this is not necessarily a good idea of launching—I imagine K12, this EdTech version of YC, and I came in to talk to Paul, and he’s like, so how’s it going with Imagine K12? Because he helped out a lot, and I said, well, it was probably called something else.

Paul: Well?

Jeff: No, no, no, still; it was by then. He actually gave us the name of Magic K12, and so it was called Imagine K12, but it’s like, when you get a launch—and I was like, well, we're going on our PR plan and we're finishing the software, so this was like—you know, we’re gonna have a summer batch, and this was probably in February or something.

Jeff: And I said, so we have like, you know probably in like six weeks. He said—he just launched! When he just launched! Now! And I was like, damn, he just did it to me! And we did—we launched; we launched like within a week!

Paul: Yeah.

Jeff: We launched.

Paul: Good, yeah.

Jeff: Like within a week, we launched, and so that is a mistake you’ll make—you’ll wait too long; you’ll—you’ll be embarrassed; you’ll—we launched YC probably in less than a week of having the idea.

Paul: Yeah.

Jeff: Of course, launching it was merely saying—was merely posting a website, and we didn’t have—we didn’t have any software in the beginning.

Paul: Here!

Jeff: People—we just had this ASCII form that people would fill out and reply to us by email. The first couple batches of applications were by email. We would print them out and pass them around amongst the partners, and then we would write grades on the top like we were grading papers, and then we would sort this heap of papers, and that was the application process.

Jeff: You’re talking about doing things that don't scale—try to do that with thousands of applications now; it would be harder! So we have lots of software that does that because eventually you scale.

Paul: You know, eventually, you build software when you learn what software you ought to be building. It’d be fun to spend a lot more time talking about how YC got started, but I think rather than that, let's—how do you have left? We can probably go to around 12, then and all videos that get too long, or like no one wants to watch anyway.

Paul: So we’ve already been going for about 40 minutes or so, so let’s open it up to any questions you guys might have.

Audience: Yeah, are you saying that if you recruit your co-founder, your friends as co-founders, you might lose them as friends?

Paul: I don't think that's a big danger. I don't think you have—

Jeff: So the question was, should I have my friends be co-founders and then maybe risk losing them as friends, or should I just take someone off the street where I'm—I’m not—I don’t have the risk of losing them as friends, but they might break my company?

Paul: Is that was that your question? Do I get that right?

Jeff: I already answered it, dude, come on!

Paul: I just wanted to make sure they heard the question too, just because I repeated the question.

Audience: Uh, how do you deal with different commitment levels of co-founders?

Audience: And if you do have to fire a co-founder, what should you do?

Paul: Okay, so the way you deal with different commitment levels is you ask yourself, would I rather have 30% of this person or a hundred percent of some other person? And in my case, I would rather have like 10% of Robert Morris's brain than a hundred percent of almost anybody else's, so is an easy decision.

Paul: How do you— could you fire co-founders in Y Combinator? The answer is you— you ask Carolyn Levy. Y Combinator has so much experience with dealing with crap like firing co-founders, and like you just go and talk to them, and they say here, you do this and this and this, right? And probably it’s online too, so you could Google it, really— is it? YC published how to—

Jeff: We— we have published! We’ve actually— if you look at some of the talks Carolyn has given and Christie, I’m pretty sure they've covered that particular issue.

Paul: One way you deal with it, by the way, is making sure that someone has a little bit more stock, perhaps like—it's usually better to have somewhat equal shares—not completely equal.

Audience: Yeah.

Paul: And if one person has 51%, the other person has 49%, then they can— you know— deadlock.

Audience: Yes!

Paul: So when you talk about launch— is launching to like private beta—a couple of people good enough, what do you have just launch? So what does launch really mean?

Paul: It means you have actually—okay, so the question is, is launching in a private beta to a few people good enough for you to have to launch to the general public?

Paul: I will take launching in just a private beta to a couple people because that’s closer to the— that’s closer to full launch than like just sitting, thinking of your idea in a café, next!

Audience: Wouldn't you come out of the currently raising in the project maybe a little bit late?

Paul: So basically, you’re asking about ICOs?

Audience: Yes, what do you think about ICOs, Paul?

Paul: I have—I haven’t—I know nothing about this world.

Paul: I’ve heard there’s a huge amount of crypto money floating around, so like if there’s money floating around, it’s often a good idea for startups to try and get it. Beyond that, I don’t know, but I will say like one of the things we've seen a lot at Y Combinator is, it’s not always right to raise too much money and even if you don’t—like the IPO thing by definition of— it’s always wrong, yeah?

Paul: The ways—a lot of money, and the thing about there’s a couple reasons for that: one, you dilute more than you necessarily need to; but the other reason is that money just sitting there has sort of a gravitational—it makes you suspend it, and it makes you do stupid things!

Paul: So it’s like these companies that have raised a billion dollars on ICOs and, I don't know, if they'll ever have a real company where you get— like, you get a bunch of LPS and raise a billion dollars, and then you just hire so many more people than you actually need, right?

Audience: I know that’s the difference!

Jeff: When I see? Oh, I get it! But just having a billion dollars when you only barely have an idea is not necessarily—you know, Magic Leap raised 2.3 billion dollars, and they seem to have almost nothing!

Paul: By the way, that is the trick for getting your question answered! I learned that a long time ago. You raise your hand as your fin—as people are finishing the preceding question, and I see that guy's hand over there; he knows the trick!

Audience: Go ahead!

Audience: What is it—my best tips for a great user interview?

Paul: Well, you want to figure out not just what they think is wrong, but what's actually wrong? Like, what's missing in their life? Right?

Paul: So just start talking to them, like, say, what would you like to do? What would you like to be able to do that you can’t? And they might just, you know, they might, they'll tend to give you an answer that's like a subset, right?

Paul: Like, you know, if you were asking someone about email, they'd like to say, I like you know, I—it’s really important to me to be able to mark emails as unread, right?

Paul: And like, what does it mean unread? That whole “unread” thing is a sign that the whole inbox is being misused; it's what’s really going on is it’s a to-do list, you know? And that’s why you have to mark items as unread, so they stay on your to-do list.

Paul: Um, and so you could— so you start asking them what's wrong, and then you try and figure out what they really mean, what they're really getting at, which they themselves might not know, you know. And then you start asking them, so what if you could do such-and-such? What if you could do such-and-such? Right?

Jeff: Okay, we should ask someone on this side. You’ve taken over, so you just keep calling them; you’re doing great, but you—you pick!

Audience: Right there in the way corner.

Audience: How do you justify the financial risk of launching a shitty product?

Paul: Okay, so the risk of launching early is not as great as the risk of launching late; it's not like there's only risk on one side, right? There's a risk of launching late too!

Paul: So you gotta have a rule of thumb about when to launch, and my rule of thumb, the way I like—I have this phrase for it that didn’t become as sticky as Minimum Viable Product because it describes what the Minimum Viable Product has, but it’s a quantum of utility!

Paul: Right? Um, launch as soon as you have a quantum of utility, which means as soon as there's one person in the world who is glad that you launched because now they can do something that they couldn’t do when there's—when you—if you don’t—if you have something that if you launched it, no one would be happy, then you're not ready to launch, right?

Paul: As soon as there's at least somebody out there who would say, oh, here’s this thing I can use because I can do X, and before that, I couldn't do it, then you should launch as soon as you have a quantum of utility.

Audience: It sounds like there’s some wolves in the utility function. Like, if it makes too many people happy and, like, one person happy and, like, a hundred really unhappy, maybe it’s not ready, but if you, you know, Pablo Kaido, he says if you have ten people who really love your product, that’s a good place to be!

Paul: Yeah, all you need is if you—that’s why if there's ten people who are super excited, totally, totally launch!

Paul: Unlike—and nobody else cares, that is perfectly fine! That’s great!

Audience: Okay, well, the danger of building—

Paul: Okay, so the question is how do you decide between building what people need and building what people want, right?

Paul: Well, the problem is what you really want to be doing is build something people will choose to use or buy more—becoming your user or something like that.

Paul: And so here's a counterexample to the whole need versus want thing: What people need is healthy food; what people buy is unhealthy food!

Jeff: So if you're a business, which one do you want to do?

Paul: Okay, I mean, yeah! Ideally, you want to make healthy food that also tastes good, right? But really, um, you can't—you can't push this too far.

Paul: You can't push this too far. I mean, I myself would not start some company making junk food or something like that, but if you are making food, don’t—like, or something where the need and want are very different, be real careful about picking need, you know, or you’ll go out, and you know, your goal will be to cleanse the world of sin or something like that.

Jeff: It’s normal as a founder to be, to see multiple different solutions.

Paul: Yeah, sure, absolutely! You just have to pick one, and pick the one that you think could get could get usage quickest.

Paul: What’s the founder DNA?

Founder DNA bothers you; you mean what makes a—what makes someone a founder?

Audience: Yes!

Paul: It’s you choose to be one; what makes someone a good founder? A combination of positive and negative things.

I think positive things are things like determination and you know, willingness to try new things. But then there may also be negative things, like if you are—I think if you have worked for a large company for twenty years, you might not be a founder unless you were forced to for visa reasons; because if you were the kind of person that would—that would make a good founder, you wouldn't be able to stand working for a large company for twenty years, right?

Paul: And so actually we’ve noticed certain companies, where the alumni of these companies tend to make bad founders, and it's because no one who would make a good founder could have stood working for these companies for very long.

Jeff: Okay, maybe we should pick one on this side!

Audience: Okay, what do you—what do you price during launch?

Paul: Well, you can probably guess if you know your business, you know roughly what you should charge, so you should probably just go with whatever you'd guess.

Paul: If you want to be more scientific about it, you could try talking to some potential users. Most of the time, you have some sort of tame users who are actual users but are also friends of yours or something like that, or family, and you can ask them, tell me honestly, like, what would you actually pay for this, right?

Paul: You can always change your prices later, though; if you want to lower your prices, no one’s gonna complain and if you want to raise your prices, you just grandfather your existing users, which if you have exponential growth will always be a tiny subset of your total users, and then no one will complain about that either! So don’t sweat about it too much one way.

Paul: I think about that, it just seems like in the beginning what you need more than anything is customers, and you want them to pay. So you will need to pick a price that gets you customers, so you’ll learn because you learn so much from those customers.

Jeff: And you're right, don’t lose—don’t leave your customer price!

Paul: Yeah, 'cause in the beginning, you just want growth because the customers teach you!

Paul: Alright, two more questions. You choose; you pick to—you.

Audience: Okay, right up front here—angel investing.

Audience: So, my name is Karan, and I’m with Chat Goose, which is a chat manager, so three tips for dealing with and finding angel investors if you don’t have, you know, the original goals, etc.

Paul: Well, we never talked about fundraising at all!

Audience: Well, maybe I'll have to do this whole thing again—it's good funds.

Paul: Fundraising as a—as a Nakia is the thing to occupy your thoughts as a founder is overrated! It’s better not to be spending all your time thinking about fundraising.

Paul: We come to it much later in the class; we don’t need to!

Audience: Alright!

Paul: Okay, so the question was how to find angel investors.

Paul: Well, that’s the part—that’s the thing I’m particularly ill-suited to answering that question because Y Combinator is like this machine for funneling angel investors into all the startups that it accepts.

Paul: So like, how do you find angel investors if you don’t—it’s like asking a pilot how to, you know, walk along and track on the ground. I know the least about that!

Paul: Here’s the thing: angel investors are looking for you, so get out there!

Paul: Go to, I don’t know, well, there’s—there’s lots of startup events; the angels are ancient, like they’re not necessarily good angels, but angels are out there looking.

Paul: And the—and the best way to actually meet angels is to get one to invest in you and then to introduce you to people!

Paul: Actually, I know the answer to this! I know the answer! Find—find people who work at startups and ask them to introduce you to their investors. That’s the way to do it! That's good—that’s a good answer!

Audience: Okay, last question! Paul, you get to—you get to choose!

Audience: Maybe from this side!

Audience: Why is YC thinking of going after high school students to fund?

Paul: Well, I don't know because I don’t think on behalf of YC anymore, but you better not be!

Paul: Because that would be an evil thing to do! There are—at YC, high school students could, there are plenty of high school students who could start successful startups.

Paul: But they shouldn't! Just because you could start a successful startup doesn’t mean you should, because if you start a successful startup, like that means like the—you know, the footloose and fancy free days of your life are over!

Jeff: Hey, like you're going for that company!

Paul: But we have funded kids in high school, but only because they’re already going, like—like it’s that we’re not encouraging that!

Jeff: I agree! You think to go to high school and encourage kids to, like, do this incredibly hard thing instead of how fun—and like being kids; that’s awful!

Paul: I actually think that it’s MoMA churn optimization to write—what you should be doing when you’re in high school and even college is you should be figuring out what the options are, not picking one option and running with it, right?

Jeff: And you know, it just turns out that usually people who are in school and then stop to start companies, it's like, you know, there are dropouts from Harvard; we know the classic examples.

Jeff: But if you have like—if you drop out and just go do it, and it's you’re on a path, that's one thing! But most people like they’re dabbling and they have a fallback, and that’s not a good way to be a startup founder!

Paul: You have to be a hundred percent committed! And so we'll fund people in college, maybe even high school, if they’re already clearly a hundred percent committed; otherwise, no!

Jeff: Whatever trend, I think that trend you're talking about is that—you graduate from high school and why bother getting a college degree because you’re already a fantastic hacker, so I'm going to hire you!

Paul: And—and if that person wants to enter the professional workforce, well, that’s great!

Jeff: I don’t know; I was—I don’t think it would have been good for me!

Paul: It wouldn’t have been good for me either, but you know, like, go off and start just some startup.

Paul: I just think between 18 and 22, you grow a lot as a human being who’s human and humane, and it's not clear what society will turn into if we—if all those folks enter the workforce.

Jeff: It’s good to like mess around and do a whole bunch of different things in your early 20s if you could have worked, yeah?

Paul: Whether it’s—whether this messing around doing a whole bunch of things takes the form of college or something else, just like—you’re careful because like starting a startup is like catching a dragon by the tail.

Paul: If it works, it works—yeah. Be careful! At what point in your life you do that?

Paul: Thank you very much for coming in!

Jeff: Thank you, Paul!