The Flexibility to Explore: Zuckerberg on Facebook’s Early History
The mob scene outside Stanford’s Memorial Auditorium this Saturday resembled the prelude to a big rock concert, and in some ways, it was. For today’s generation of young tech entrepreneurs, Mark Zuckerberg—the headline speaker, for the third year running, at Y Combinator’s Startup School—is like Mick Jagger, Madonna, and Justin Bieber rolled into one.
In his on-stage interview with the Facebook CEO, Y Combinator founding partner Paul Graham focused on Zuckerberg’s brief time at Harvard and the forces that brought Facebook (NASDAQ: FB) into being in early 2004. Conscious that he was speaking to an audience of aspiring entrepreneurs, Zuckerberg hit on a few key lessons, which I’ll summarize here before diving into the full interview transcript.
—It might have been possible to start Facebook before 2004, but users would have shared much less information with each other. Zuckerberg said he perceives an exponential curve of social sharing—he called it a “social networking version of Moore’s Law”—and predicted that within 10 years, Internet users will be sharing 1,000 times as much as they share today.
—When it began to expand beyond Harvard, Facebook deliberately targeted campuses that already had their own social networking systems (Stanford, Columbia, and Yale) because Zuckerberg wanted to prove that Facebook was “the best service out there.”
—Facebook grew slowly at first (only adding a new server when there was enough ad revenue to support it), which gave the team time to work out the kinks. “It took a year for us to get a million users,” Zuckerberg said. “And we thought that was incredibly fast. It wasn’t as quick as a lot of things are today. Having that time to bake it was really valuable.”
—Young entrepreneurs shouldn’t focus on starting companies; they should focus on finding fundamental problems where they can have a big impact. “I didn’t start Facebook to start a company,” Zuckerberg said. “I started it because I really wanted this personally, and I believed it should exist and I thought it should be global and I wanted to play a part in doing that … I never really understood the psychology of knowing that you want to start a company before you decide what you want to do.”
—If you start a company too soon in your career, you might get locked into something that isn’t important or rewarding. “One of the amazing things about being in college [is] that you can work on all these hobbies and try out a lot of stuff—it’s this amazing flexibility that most people take for granted. Once you decide that you want to start a company, and you’re going to do it with somebody else, you now immediately have to check with someone else if you want to change your mind on something.”
—Zuckerberg said that if he hadn’t started Facebook, he probably would have finished college and wound up with a job at Microsoft. On the other hand, a succession of people, including his mother and investor Peter Thiel, predicted (correctly) that he’d drop out somewhere along the way.
Zuckerberg also talked about university e-mail systems as Facebook’s first layer of identity management; his love of classics and psychology; his propensity at Harvard for hacking instead of doing homework; whether the Dunbar Number really puts an upper limit on human sociability; Facebook’s competition with MySpace; and the company’s (non-)decision to move to Palo Alto.
What follows is a full writeup of the talk, based on my recording and typed notes. Note that there was a jocular tone to many of the interactions between Graham and Zuckerberg that doesn’t really come across in the written transcript. Also, the photo above is from Zuckerberg’s 2010 appearance at Startup School—YC didn’t allow cameras in the room this time around, perhaps in deference to Facebook’s new status as a public company.
Paul Graham: I’m going to ask a lot about the very early days of Facebook. How long before 2004 could something like Facebook have succeeded? What were the last things that neeeded to fall into place? Could someone have done it in 1995 or 2000?
Mark Zuckerberg: Interesting question. There were certain elements we bootstrapped off of and used to hack early identity. One of the things that people don’t think much about today is that early on we wanted to establish this culture of real identity. There weren’t really any other online services or communities where people were their real selves before that. One of the ways we determined that someone was who they said they were and that their credentials were real were everyone had school e-mail addresses. I don’t know how much before 2004—it was probably around 2000 that all schools began issuing e-mail addresses but that was really this critical thing. This counterintuitive thing.
PG: So school e-mail addresses were the original source of identity.
MZ: That’s how we knew what school you were in. It also kept people from setting up fake accounts, because people only had one school account. Being able to bootstrap off that was this nice early thing that helped establish this culture of identity. Before we got to a million or 10 million people, we were able to bootstrap off that culture and we’ve been able to keep it, even though most people in the world now don’t have e-mails that were issued by some institution that vouches for their identity.
PG: Now people get their identify from Facebook. You’re the source of identity!
MZ: It comes around. But to your question about when it would have been possible. One of the trends we see is that the amount any individual share in a given year seems to be growing at an increasingly exponential rate.
PG: Zuckerberg’s Law?
MZ: I don’t call it that. But it is a kind of social networking version of Moore’s Law. And it’s interesting. What that suggests to me is, if you expect that this rate is going to double every year for 10 years, two to the 10th is 1,024 … So I think the question is not `would Facebook have been possible before 2004?’ It would have been in some capacity, but people would have shared less. And if you fast forward five years, there are going to be versions of social services that people use to share way more. Anyone here should be thinking about, 10 years from now people are going to be sharing 1,000 times as much stuff as today if this trend continues. What things are going to have to exist in the world in order for that to be possible?
PG: Instagram for toilets.
MZ: Instagram is doing really well. So that is a good frontier.
PG: When you first launched in the very beginning the features were a profile, name, who you are, and also what house or dorm you lived in, and what courses you were taking. Do you think it would it have worked without that? Wouldn’t it have been enough to just have profiles?
MZ: It’s a really interesting question. We evolved and wanted to be a more general service. But I remember there was this early debate that Dustin and I had. We had to do some manual work for every new school that we added to Facebook, and in order to do that we went through and parsed the course catalogs for each school to make sure the data was clean. And I remember Dustin saying we could just expand so much faster, it would just be easier to launch new schools if we didn’t have to have the course catalogs. We just had this really long debate about what quality meant for us and the culture we wanted to have. In retrospect maybe it wouldn’t have made a huge difference. But it definitely set the tone that there was a lot of data on Facebook that you could rely on. Tools like a college-specific thing, which was valuable beyond the person-to-person thing.
A lot of you guys, in the projects you work on, are going to have a lot of similar questions. There is the famous 80/20 rule, where you can get 80 percent of the benefit by doing 20 percent of the work. But you can’t just 80/20 everything. There have to be certain things that you just are the best at. Where you go way further than anyone else, and you establish this kind of quality so that your product is the best thing out there. We really had to that one, we had to do enough of those things or else we wouldn’t have been the best service out there.
PG: Do you recall when you stopped adding course catalogs?
MZ: It was when we expanded beyond colleges.
PG: Really? So you kept doing courses?
MZ: I think we eventually figured out a way to crowdsource this so it was a bit easier. We had enough data that we could extrapolate from the colleges that were in place. But we did it probably way longer than was rational.
PG: Do you remember how much your first server cost?
MZ: $85 [per month].
MZ: Yeah, I remember that, because that was the gating factor for us launching new schools. We had this philosophy from the very beginning where we didn’t want to be this project—it wasn’t even a company at the very beginning—but we didn’t want to be burning a lot of money. We weren’t planning on raising a lot of money. We didn’t want to be one of these things where you raise a bunch of money and lose a bunch of money and somehow pull it through at the end. So $85 for the first server. We had this thing in the beginning where we were running some kind of ad network.
PG: Do you remember what your first ad was?
MZ: Um … No, I don’t. This was Eduardo [Saverin]’s job. Basically he was responsible for making sure that we had enough money to keep things running, in the beginning. So a server was $85. Dustin and I worked to make sure that you could fit more schools on each server. And whenever we had more money we would bring in another $85-a-month server.
PG: You didn’t spend money if you didn’t have money.
MZ: Not really.
PG: That was a real strength then.
MZ: Yeah, it was actually good because sometimes it’s really nice to have the time to get your product to be awesome. One of the things that was interesting was, you know at the time, I don’t know how many of you will remember but at the time Friendster was having massive scaling problems. They grew quickly and they couldn’t scale. The fact that we could go college by college and offer new features and make it more efficient and make sure they worked, I think that was really key. You are talking about people who had never built a company. We had never built any large-scale software for anything. So having that period where we could just bake it. People like to talk these days about how these services grow so quickly. And Facebook did grow quickly. But I think it took a year for us to get a million users. And we thought that was incredibly fast. It wasn’t as quick as a lot of things are today. Having that time to bake it was really valuable.
PG: Didn’t all of Harvard get on it? Like 6,000 students?
MZ: I think there are like 5,000 undergraduates at Harvard. And we got something like two-thirds or three-quarters of Harvard in the first two weeks. The vast majority.
PG: Was there ever a school where you opened it, and it didn’t stick?
MZ: Some schools took longer than others. What we basically did was, we launched at Harvard first because I wanted it. I built it for myself. I really wanted to use this service. This is one of the ironies—I started building Facebook because I wanted to use it in college and then I immediately left college.
PG: But it’s expanded to everyone outside college, so it worked out.
MZ: But yeah, so then after Harvard all of these schools, a lot of students from other schools started writing to us and asking us to expand. And we weren’t looking to start a company right then. I thought something like this would eventually exist at large scale. But one of the interesting juxtapositions that was going on at the time was, I remember distinctly I had this one friend who I went for pizza with almost every night and we did all our problem sets together in computer science. And at the time I remember talking to him about how I was working on this Facebook thing and I thought it would be cool for Harvard and I really was excited about it because I wanted to use it, but at the same time I felt that over time, someone would definitely go build a version of this for the world. But it wasn’t going to be us, it was going to be, you know, Microsoft. Or someone who builds software for hundreds of millions of people. We were college students! We were not qualified in any way to build this. So back to your question—I went off on a tangent—the first set of schools that we launched at after Harvard were schools that had school-specific social networks. Stanford had something, Columbia had something, and I think Yale had something.
PG: Why would you choose schools that already had their own social networks?
MZ: Well, I wanted to go to the schools that I thought would be the hardest for us to succeed at. I knew that if we had a product that was better than everything else that the students were making at other colleges, then it would be worth investing in and putting time in. But I didn’t want to just get into a project where there would be this huge legacy of maintaining it and ultimately there would be [all these competitors]. So we thought this was going to be good and we launched it at Yale, Stanford, and Columbia.
PG: So you probably could have gone into random schools and succeeded, and you chose those because they had competition.
MZ: I think what we saw at those schools that people wanted to use something like this. So we wanted to make sure that what we had was way better than anything else that was out there.
PG: I read the Crimson article when you launched. It said “Hundreds of People Sign Up for New Facebook Website.” They said in this article that the Harvard computer services people were working on a university-wide face book. The problem was that they couldn’t figure out how to restrict the information. If you had come to Harvard and this had already existed, do you think you would have started Facebook?
MZ: I don’t know. There is this trend I was talking about earlier where each year people are sharing more and more. You can map out, at any given point you can look at the Internet and say okay there is enough sharing to support certain products. Wikipedia came before Facebook because there was a small amount of sharing of information about all of these public entities. But in order to have enough sharing to support some basic information so you could look up anyone and find some interesting stuff about them, we had to travel along this curve. A couple of years earlier someone could have done something more basic, but a couple of years later you would obviously have been able to build something that was more encompassing and allows people to learn more about the people around them. Our continual mission and job is to keep on building that next thing and that’s what we live for at Facebook and that’s what excites us.
PG: So if the university had built this face book, you would have built Facebook anyway.
MZ: It’s obviously hard to know how this would have played out. One of the interesting things about Facebook was that it wasn’t just a picture and some basic info. It pretty quickly gave people the ability to share more stuff. One of the early stories that is pretty instructive for anybody trying to build a startup is, we really listened to what our users wanted. That means both qualitatively listening to what they say and quantitatively looking at the behaviors they take. And in the beginning, we had one photo on your profile. And what we observed was this behavior where every day a lot of people would upload a new profile picture. And our takeaway from this was there was this very strong demand to have a service where people could share more photos. And actually it wasn’t until later that we had the server capacity and the engineering team to build a whole photo sharing service, and that now has become obviously one of the key parts of Facebook. I think we are over 300 or 400 million photos shared a day now. But obviously nothing that a university built would have supported that. If a university had built it, the little mug shot on your [I.D.] card would have been your picture.
PG: They would have chosen your picture for you.
PG: Do you remember when you first went to college what you planned to do afterward? Did you want to go to graduate school? Were you going to get a job?
MZ: When I first went to college, I was actually planning on being a classics major. I loved classics in high school, Latin and Greek. I find them fascinating. And my sister actually did go on and do that. When I was in college I actually wasn’t a computer science major, I was a psychology major. I didn’t actually get around to taking that many classes, because I left pretty quickly. So I took more computer science classes than psychology classes.
PG: So you had no plan. You were going to be a barista.
MZ: No … [laughter]
PG: At some point you got sidetracked into programming.
MZ: Growing up, I always had a lot of respect for Microsoft and what they had built. A lot of people from Harvard went to Microsoft. So maybe I would have done that. It’s obviously hard to say. Later, I made this bet with my sister, the classics PhD. I remember when I was starting college, she bet me that she would finish college before me. I’m like all right, I’ll take that bet. And then after I dropped out I was talking to my mom, and she was like, “Yeah, I always knew you would drop out of college.” And I was like, “Thanks Mom!”
PG: Did she mean that you would zoom out the top, or fall out of the bottom?
MZ: I never asked her.
PG: Do you think your parents knew that you would always want to run your own show?
MZ: They would probably say yes.
PG: Did you know you wanted to start a startup?
MZ: Now that is the interesting part of being in a place like this [Startup School] where a lot of you guys are thinking about starting companies. For me, so much of the lesson that I feel I’ve learned is, I think it’s really hard to decide to start a company. I didn’t start Facebook to start a company. I started it because I really wanted this personally and I believed it should exist and I thought it should be global and I wanted to play a part in doing that. It was just wanting to build it, and have it be this hobby, and eventually it got the momentum to become a company. But I never really understood the psychology of knowing that you want to start a company before you decide what you want to do. I know that’s different from your philosophy on this.
PG: No, believe me, I know too many people where their company started them rather than vice versa.
MZ: Going back to your question of why did we open first at colleges that already had competitors. I had this big fear, I think, of getting locked into doing things that aren’t actually the most impactful things. This is the trait, I think, that entrepreneurs have. They just have this laser-like ability to find where they will have the most impact. If you take on a project, especially if you hire people, you are going to do that project. There are obviously ways you can exit and all that. But I think having the flexibility to explore a lot of different things—which you can do when you’re in college, which is one of the amazing things about being in college, that you can work on all these hobbies and try out a lot of stuff—it’s this amazing flexibility that most people take for granted. Once you decide that you want to start a company, and you’re going to do it with somebody else, you now immediately have to check with someone else if you want to change your mind on something. I think people really undervalue the option value of flexibility.
PG: So stay in college!
MZ: I think, explore what you want to do before committing is really the key thing. And keep yourself flexible. You can definitely do that within the framework of a company. But I would be wary about starting a company too rigidly. Because you are going to change what you do. People talk about pivots all the time, as if your thing didn’t work so you pivot. Facebook pivoted many times. We were college, and then we were not college. We were a website, and then we were a platform. You are going to change what you do.
PG: There is another word for what you are talking about: Expansion.
PG: I’m curious. When you first started: There is a difference between making something where people sign up, and making something where people keep coming back. What was the feature that kept people coming back to Facebook over and over again? Unless they were updating their profile.
MZ: I think it really just gets down to what makes humans human. This goes back to my study of psychology. The human brain is uniquely wired to process things about people. When I look out [at the audience] I see faces, I don’t see chairs or the room around people. We are hard wired to think about people. There are whole parts of the visual cortex that just process the slightest micro movements in faces. This is what people are. This is how we process the world.
I heard a story recently which is interesting, most humans if you take an MRI when they’re dreaming, they dream about social interactions. Humans are the only animal that does this. So, OK, when I thought about the Internet before Facebook—Google and search engines were amazing, you could get access to information about anything you wanted, but you couldn’t learn about the people around you. Most of this information was out there, it just hadn’t been indexed by a search engine. There had to be a service that let people share the things they wanted and let them control it. I think Facebook did that. One definition of technology is that it extends natural abilities. Glasses extend your ability to see. Steve Jobs once famously compared a computer to being “a bicycle for your mind.” And basically extending your ability to think. And the word “computer” comes from the Latin for “think together.” A social network, I think, extends people’s very real social capacity. You hear all these approximations. There is this famous Dunbar’s Number: That humans have the capacity to maintain empathetic relationships with about 150 people. I think Facebook extends that.
PG: On Facebook do you see certain groups stop at 150?
MZ: When people sign up the average amount of friends they get is 150, but after that it expands and you can keep in touch with a lot more people. So, given that, I actually think one of the lessons from that is, do something fundamental. I think a lot of people in a lot of companies that I see are operating on small problems. And it’s cool if you are an entrepreneur and you have a partner and you’re building a company to solve some tangible problem, but I think the most interesting things operate on these phenomena in the world which are really just fundamental things about how the world operates.
PG: So what you did was something that was fundamental for a small market. Then you just expanded the market, from Harvard students to everyone.
MZ: Well, it was fundamental for me. I felt this need really acutely. I really wanted this. It turned out that this wasn’t just for college students. Almost everybody in the world who has friends and family wants to stay in touch with those people.
PG: In retrospect—this is a bit of a controversial question perhaps—but do you think MySpace had a chance, once you started the center of gravity shifted. From the point you started expanding out of Harvard, it seems like in retrospect they were doomed.
MZ: I don’t see it that way.
PG: You think they could have won?
MZ: No, it’s not about winning and losing, it’s about doing something valuable. Almost every product and category was going to get transformed. There are things that MySpace did that Facebook has never done. You know, MySpace was a much better service early on for meeting new people. Facebook was never primarily about meeting new people. It was about staying active with the people you knew. And kind of mapping out the real relationships that existed. I think part of the issue is that they saw us growing and they felt threatened by that and they tried to copy what we were doing. You are never going to win that way. Of all these interesting social services and apps that are being built today—think about all the new apps that you guys install on your phones. There are so many interesting things. Eight of the top 10 iOS apps plug into Facebook. 50 percent of the top 400 iOS apps plug into Facebook. They are all socially integrated, but the companies that are getting started now that are just trying to copy the stuff, that has already been done, aren’t going be successful.
PG: Do you think MySpace could have survived if they’d gone into some marginal territory?
MZ: It’s a question of bringing real value. I think people have a fundamental need to stay connected with the people they know. I think people have many fundamental needs, to meet people and expand their horizons as well. That has never been the primary problem Facebook is trying to solve. And, I think it’s something we can do, or that someone else could do using our platform, or something that someone else could build independently. I also never bought this music thing for MySpace. They always said they were a music service. I’m not sure.
PG: Before we go I want to ask you about how you ended up out here. How did you end up in a house in Palo Alto.
MZ: I remember bits of the story. I wrote the first version of Facebook in January of 2004. The reason I did it in January was at the time Harvard had this weird intersession thing.
PG: They don’t have reading period anymore?
MZ: I think they changed it. Now they just kick you out if you start anything interesting there. [Laughter] I think they are actually changing that. Strike that! But they had this thing before where in January you basically had this dead month where you could study for finals. Hypothetically.
PG: I remember reading that you started Facebook during reading period. It was because you had this time where you weren’t too busy with stuff.
MZ: Yeah, although I actually probably should have been studying. I was taking this course on the Rome of Augustus. And on the final there are these pieces of art that you studied from class and on the final they show you some pieces of art and you have to write about the historical significance of that. I hadn’t really done much of the studying for the class. I spent most of my time programming and reading things that I enjoyed. I could have used reading period to study but instead I was building Facebook. So instead what I did was I hacked together this website where I went and downloaded from the course website the 200 or so images that were going to be potentially on the final. And I built this very simple site where it showed one of the images and then you could contribute what you thought was significant about that image and you could see what other people thought was significant about it. And then you’d go Next and it would pop up a random one. And I e-mailed this to the class list, I said “Hey guys, I’ve built this study tool,” and they just populated it. And it worked for me and it was wonderful. After that they mentioned that the grades on the final had never been higher. [Applause.]
PG: You crowdsourced your study tool.
MZ: There are a lot of interesting social dynamics that you can applied to almost any issue in homework. But yeah, I built the first version in January. I spent some of that time at Harvard and I also went and visited a couple of friends, one at Stanford and one at Caltech. And at the time, I had never even been out to California before.
PG: And you went in January. What did you think?
MZ: I remember coming into SFO and driving down 101 and I saw these buildings for all these technology companies came from, and it was amazing. And then the weather of course was awesome. And I remember, I had been at Harvard through freshman year, and then I stayed there for the summer, and then sophomore year, so by the time that sophomore summer came around my friends and I were, “Let’s go somewhere else, let’s go someplace in California.” So we decided to get a place in Palo Alto. At the time we actually were not thinking about moving to California and dropping out. What was in our mind was, it will be neat to be around all of these great companies, and maybe we’ll find something to build a company out of. But surely this isn’t it.
So we went out to California. And I remember this conversation where one day Dustin pulled me aside, and were getting a ton of users, and we have an increasing number of servers, and we have no ops guys, and this was before EC2 so you had to manage your own servers, and he was like, “This is really hard. I don’t think we can do this and take a full course load.” So Harvard has this policy where you can take as much time as you want off from school. So why don’t we just take one term off and then just try to get it under control and build the tool so we can go back next semester and it will run more autonomously? So we did that, and of course we raised money from Peter Thiel, and we told him we might go back.
PG: You told him you might go back to school?
MZ: Yeah, I think he didn’t believe us. [Laugher] There is this long history of other people thinking I was going to drop out well before I did. But, so, then you know spring term came along and we hadn’t quite built the tool and the automation, so let’s take another term off, and finally at some point we just figured that we were out there. But by then we had millions of users.
PG: So you didn’t decide not to go back until you had millions of users. Wow.
MZ: Sometimes I joke about it. Harvard has this policy where you can go back whenever you want. I might still.
PG: I’m sure they wouldn’t mind in your case. Are we out of time? Mark has a wedding to go to.
MZ: I actually do. It’s the guy who I said I used to go out for pizza with him and do our CS problem sets. He joined Facebook and we’re really good friends and he’s getting married. So I have to go to that. Thank you guys.