Paul Buchheit was Google's 23rd employee. He was the creator and lead developer of Gmail, Google's web-based email system, which anticipated most aspects of what is now called Web 2.0. As part of his work on Gmail, Buchheit developed the first prototype of AdSense, Google's program for running ads on other websites. He also suggested the company's now-famous motto, "Don't be evil," at a 2000 meeting on company values. Although not a founder, Buchheit probably contributed more to Google than many founders do their startups. Gmail was in effect a startup within Google--a dramatically novel project on the margins of the company, initiated by a small group and brought to fruition against a good deal of resistance. Livingston: Take me back to how things got started. Was Gmail a side project or commissioned by Google? Buchheit: A little bit of both, actually. I started working on email software a long time ago. I think it was maybe 1996, but it was just a little project. I had all these ideas that never really went anywhere. Oddly enough, I think I was calling it Gmail at the time, for some other reason. It was just a random project-- not necessarily the predecessor to Gmail--but it was something that I'd been thinking about because I'd been sort of unhappy with email for a long time. It was before Hotmail and I was in college at the time. If you wanted to check your email, you'd have to go back to your dorm room. I thought, "That's so stupid. I should be able to just check it anywhere." So I wanted to make some kind of web-based email. But I really didn't know what I was doing, so it didn't go anywhere. I wrote something, but it was never useful and never got off the ground. So fast-forward to much later: I was here at Google and I had worked on Google Groups, which is not exactly the same, but it's related. After the first 161 Paul Buchheit Creator, Gmail 12 CHAPTER generation of Google Groups had mostly wrapped up, they asked me if I wanted to build some type of email or personalization product. It was a pretty non-specific project charter. They just said, "We think this is an interesting area." Of course, I was excited to work on that. Livingston: So they didn't ask for an email product? Buchheit: They were very general--just kind of saying, "Yeah, we think there's something interesting to do here," but it wasn't like they gave me a list of features. People really weren't sure what it was. And this was when Google was still pretty much thought of as exclusively search, so even the idea of doing something like email was strange. A lot of people were kind of unsure. At this point, it wouldn't seem like a big deal, but at the time it was a little bit controversial. For quite a while I was just working on it by myself. I actually started out with some of the Groups code, just because I was familiar with it. I built the first version of Gmail in 1 day, just using the Groups code, but it only searched my email. I released that to some Googlers and people said it was useful, so it progressed from there. Livingston: When you built this first version, was your vision to create a better email program or was it to build something that would allow you to search through your emails? Buchheit: Both. Search is obviously very important. It was central to what we were doing at the time and it's really useful for managing your email. I had ambitions of doing more than that, but search seemed like the natural first step--it was one of the things that was most obviously a problem. Everyone here had lots of email. This company is a little bit email crazy. I get 500 emails a day. So there was a very big need for search. That was the most obvious thing that I could do, and it was also one of the easiest. So I built this first version and it only searched my email, but even that was useful for other people, because we had a lot of the same email. So then they said, "It would be even better if I could search my own email." Livingston: You could search for keywords, senders, etc.? Buchheit: Yes, it was free text, just like Google is, but for email. Livingston: Was it supposed to be your full-time gig or was it part of your 20-percent-of-your-time projects? Buchheit: Nothing's totally full-time, but it was mostly full-time. I still had some other projects that I would have to spend some time on, and inevitably I end up with side projects just because something catches my eye and I go off and work on it for a little bit. I think I may have something to do with 20 percent projects as well because I've created a few things on the side. AdSense, the content-targeted ads, was actually something that, if I recall, I did on a Friday. It was an idea that we had talked about for a long time, but there was this belief that somehow it wouldn't work. But it seemed like an interesting problem, so one evening I implemented this content-targeting system, just as sort of a side project, not because I was supposed to. And it turned out to work. Livingston: This is Google's AdSense now? 162 Founders at Work Buchheit: It's the same concept. What I wrote was just a throwaway prototype, but it got people thinking because it proved that it was possible, and that it wasn't too hard because I was able to do it in less than a day. After that, other people took over and did all the hard work of making it into a real product. Livingston: You have done two groundbreaking things at Google. Buchheit: Probably. I've done a lot of random things. Mostly what I do doesn't turn into anything, because I like to just try out ideas and a lot of them don't go anywhere. Livingston: So you work on Gmail for a day, you can see you're on to something--then what happened? Buchheit: For quite a while, it was just myself; and then another person, Sanjeev Singh, started working on it. But switching projects here, especially back then, wasn't easy. It wasn't like one day, you're suddenly on a new project. So he still ended up spending a lot of his time on enterprise search, which he was working on at the time. It was quite a while before Sanjeev could really spend most of his time on Gmail. So it was pretty slow for a long time. It was mostly just me; then me and Sanjeev; then later on another person, Jing Lim, started. It was a very slow kind of progression. And people were still a little bit uncertain about the whole idea of doing something as different as email. Livingston: When was the moment when you said, "This is big and we're going to launch this"? Buchheit: Several days after launch! It was a big project. Sometimes it seemed as though we weren't ever going to make it out. Livingston: Tell me about some of the most challenging parts. Buchheit: There's a lot that was challenging about it, just because it's very big, for one thing. We gave everyone a gigabyte of storage to start with. At the time, the standard was around 2 or 4 megabytes. A lot of people actually didn't think that was real. They thought it was a joke--partially because we launched on April 1. They also thought it wasn't possible. It can be a little bit tricky, because it's a lot of data if you actually do the math: you have millions of users and they all have a lot of data, and then, to make the system really reliable, you need to keep several copies of the data, backups and everything like that. It requires a lot of research. It's a lot of machines and a lot of systems to make that all work without requiring an army of people to maintain the system and keep it running. There's a very complicated system problem there. We were also doing a lot of things that were new to Google. And I guess this is one difference between a regular startup and starting within Google--I think it's a little bit different now, but at that time there was still this vision that, "We only do web search." Now we do lots of neat products that go beyond that, but at the time, a lot of people inside the company were sort of unsure. The idea of doing this product that was receiving all the email--and we had to store the email, which is a different systems problem, really, from web search, because Paul Buchheit 163 in web search you go out and you crawl the web and index that data and the latencies are different. We go fetch a page and it gets searchable a little bit later. But in email, everything has to be instant, and of course you can't lose any of the data either. It turns out to make a big difference in how you build things. A lot of the strategies that you might use for web search can be problematic when you apply them to email at a systems level, simply because you need to make everything so fast. It has to happen right away. You can't say, "Well, we receive email and then in half an hour it will appear." Which is actually how it worked in one of my early versions--the email would come in and I had this little script that would incorporate it into the index, but it generated this long lag, and so that wasn't really great. All of those little details add up to creating a lot of challenges, just to get it all right. The JavaScript was a big deal as well, because at the time that we first started doing the interface in JavaScript, most people thought of JavaScript as a tool for pop-up advertising and other obnoxious things like that. This was before the whole Ajax thing, so a lot of people were pretty skeptical that JavaScript could work reliably. Not without justification--it is a little bit tricky because if you do things wrong, you'll crash the browser. So making all of that work and work really well took some learning and figuring out the right techniques and where to draw the line about which features are a good idea and which aren't. Livingston: Which was your favorite feature? Buchheit: That's hard to pin down. Actually one of the things that we added very early on, which at this point seems pretty obvious, but it turned out to be really nice, is the autocomplete when you type in the email addresses. Once you have it, it just seems so obvious. "Why wouldn't you have autocomplete?" Livingston: This was a first? Buchheit: None of the other web mail providers had autocomplete. Now you don't really even think about it, but it makes a big difference. You can send email so fast and you don't have to remember the addresses. To my knowledge, we were the first web mail provider to do it. Desktop products would have things like that sometimes, but no web mail was doing that at the time. Livingston: Was it always your plan to archive everything and not delete emails and have the massive storage needs? Buchheit: You can delete email. The idea was that there's valuable information in email and we thought, "Why would you perform these actions?" For deleting, we found three or four reasons why you might delete things. One is that you're running out of space--which was the most common reason for deleting things, because you only had a 2-megabyte quota. We said, "If we give people enough storage, then they won't run into that problem." The second reason was that people would delete things just because email quickly became unmanageable if they didn't. So we said, "We've got search, we'll try to make that efficient." I can handle--I don't know how many millions 164 Founders at Work of messages are in my email now--but it's not a problem. They don't get in the way. They're just there, and if ever I want to find that message from four years ago where someone made some funny comment about Gmail that is ironic at this point, then I can go back and find it. I guess the third reason was that there's something in the email that the person's really nervous about and they just want to get rid of it. But that's pretty uncommon. So we said, "You want to provide the ability to delete things, but ordinarily it isn't really necessary, because most of the reasons are actually just consequences of limitations elsewhere." Livingston: What else were brand new features that the world hadn't seen? Buchheit: Conversation view was new--when you click on a conversation and you get all of the messages as cards instead of separate emails. Livingston: Was that your idea? Buchheit: This was a consequence of a few things. One is that I'd worked on Groups, where we had done some of the same threading. Second was the fact that we have so much email internally. We'd have these conversations where someone sends out an email and then four different people reply to the same thing, and some of them would be like five hours later and you'd think, "This has been covered five times already and you keep responding." It turned out part of the reason people were organizing their mail so aggressively is because they were trying to put the conversations back together. They'd put them all in the same folder--or they would forget and put them in the wrong folder and then the conversation would get split and they could never find the reply to this message. There were all these little tools and tricks that people had for reassembling the conversations. Why not just put them all together to start with? At some point, we said, "Let's hide the quoted text too." Because that way you can just read it much faster without having to read the same content over and over. We were also looking forward to integrating chat/IM. We didn't have time to include chat in the original launch, but it was in the early prototypes because we very much wanted to integrate chat and email--they belong together. So one thing we did was to think about email from a chat perspective, as though we were adding email to chat instead of the other way around. Of course chat is very much conversation-oriented--nobody thinks about individual chat messages. So the conversation view also came out of that--for a while we even formatted the email to look more like a chat conversation. Livingston: It sounds like you really took the user's perspective when you designed Gmail. Buchheit: Absolutely, that's very much how it developed. Every time we would get irritated by some little problem, or one of the users would say, "I have this problem, it isn't working for me," we'd just spend time thinking about it, looking at what the underlying problems are and how we can come up with solutions to make it better for them. Paul Buchheit 165 Livingston: How big was your group by the time it launched? Only three of you? Buchheit: There were a lot more people at that point. It depends which people you count, but it was about a dozen. Livingston: Was there a time then when you said, "We need more programmers to get this going"? Buchheit: I was always asking for more people. We still ask for more people. There's so much more we could do. The product is nice, but every day there are things that I find that I want to change. But when you're operating a big service, it also takes a lot of work just to deal with growth and improvements. A lot of the improvements are invisible. For example, I think we added 43 new languages. You don't necessarily notice that as an English user, but for most of the world, it's a big deal. There's just so much work as the product becomes big and needs to support millions of users. Livingston: When you launched, had you already had users? Buchheit: Literally from day one, we had users internally. One nice thing about Google is that we can just release things internally and have this great population of testers, essentially. So people inside have been using Gmail for a long time. The code name was Caribou. Initially, I called it Gmail, and then we realized that was not really very subtle, so we changed it to Caribou. Livingston: Did you choose Caribou? Buchheit: Yeah. There's a Dilbert cartoon where he's talking about "Project Caribou," and I thought it was a funny name, so I used it. Livingston: Tell me about one of the darkest days of the project, when you felt that you couldn't do this. And tell me about one of the most euphoric days. Buchheit: There are a variety of dimensions to the darkest days. Like I said, a lot of times it was sort of controversial, especially in the very early days, because people weren't sure if we should even be doing this. So the general attitude would swing, and when it would swing against us, that was very hard to deal with. Later on, not as much. We would have some system problems internally. In a previous generation, it wasn't as redundant as what we finally released, and the hard disk in one of our machines that had everyone's email stopped working. I came in and everyone I walked past would ask me, "When is Caribou going to be back up?" I was walking into the machine room with screwdrivers, and people saw me and were like, "Oh no!" I managed to take apart the hard drive and transplant the electronics from another drive, so nothing was lost. Through the whole thing, we've never lost any data, which is kind of unbelievable considering everything that happened. A lot of the machines that Google is built on--commodity is the polite word for them--they're regular PCs and so they're not always the most reliable. The most fun was, of course, launching. Nothing is more exciting than finally getting it out there for the world and seeing that people like it. 166 Founders at Work Livingston: Were there any disasters on launch day? Buchheit: Nothing major. It went surprisingly smoothly. There are always little problems but nothing so bad that I remember it. But then again, I'd been awake for 70 hours at that point. I was awake for about 3 days, because I was furiously assembling the last bits--sort of stitching together some systems to actually make it public, like the login system. And just testing everything. Livingston: Did you sleep well that night? Buchheit: Strangely enough, when I went home, I had a hard time going to sleep. Livingston: Since Google was totally focused on search at the time, was there ever a point where you worried that your project would get canned? Buchheit: All the time. Again, it was sort of a much earlier time than now where it fits in nicely. It was really kind of the first thing that diverged from the simple idea of web search. Even Groups is still basically search--it's just search over public usenet posts. So it took a while for people to get used to the idea of something different. You have to remember that the situation between Google and Yahoo was different at that time. It was sort of a different company with different concerns. Livingston: Is Gmail still invitation-only? Buchheit: No, you can sign up with a cell phone. Livingston: And on Blogger, right? Buchheit: We've extended it in a bunch of different directions. All university students can sign up, because we wanted to make it available to students. Livingston: What was the idea behind the invitation-only signup? Buchheit: There were a few different factors. Again, I mention that this is a really big thing in terms of the amount of data and everything else. A big concern has always been that I don't want to lose any of that data, because of course nobody wants to lose their email. If something goes wrong with web search, you can go back and crawl the Web again, but with email, if it's gone, it's gone. I was very concerned about keeping the systems operational. So part of it was just controlling the rate of adoption so that you don't exceed any of those limits. You always want to make sure that the current users are getting a good service. Also, it controls some of the abuse, by making it harder for, let's say, a spammer to get 10 million accounts, which also would be bad. Livingston: Who did you learn things from at Google? Did you have mentors? Buchheit: I didn't know anything about building these large systems before working at Google. So I'd look at how different parts of Google work and sort of say, "Does that apply to us? Can we reuse that technique?"--since there was already a successful model of how to do these things. That was part of the challenge, just figuring out when to copy other parts of Google and when to say, "Our problem is too different from theirs. We have to do something new." Paul Buchheit 167 That took us a while to figure out. You don't want to ignore all of those lessons, because that would be a big mistake, but at the same time, sometimes you really are just solving a different problem. For example, the update issue: we needed to be able to update instantly. Something like search, you can have a little latency. If a document doesn't get added for a few minutes, it's not a big deal. So at a system design level, that actually makes a huge difference, even though it's a seemingly small difference when you describe it. Livingston: It seems like one advantage of having a startup-like project within a big company is that you have access to all its resources. Tell me about some other valuable things. Buchheit: I think the people are the biggest resource. There are really smart people around, so you could just go talk to them and say, "How are we going to do this?" and brainstorm solutions. You can just go talk to people, whether it's the engineers... and Larry and Sergey are actually really smart. Yesterday, I heard someone making a comment like, "These guys get lucky and now they think they're smart." But in fact, they really are smart and have good ideas. Sometimes people think that these guys just got lucky, and luck is always a factor in everything, but it isn't sufficient. It takes more than luck to build something that successful. So there are lots of good resources, in the people and also systems. We get machines--we don't have to build the machines ourselves--so it's nice to have that infrastructure. Storage turns out to be a surprisingly difficult problem. It's not solved. There are network attached storage (NAS) appliances, but they tend to be expensive and they have some other problems. Then you have what we do with PCs, and that's technically pretty challenging--to take this big network of machines that are unreliable and build a big, reliable storage system out of it. We're getting a lot closer, but it probably isn't something that some startup could pull off the shelf, at least not without paying for it. Livingston: Was there anyone else at Google commissioned to work on an email program at the same time? Buchheit: No. It's possible someone else was doing something on the side, but I don't know of any. Livingston: Did you get a Google's Founders Award? Buchheit: No, most of what we did predated the Founders Awards. But things mostly worked out for us anyway. Livingston: What surprised you most looking back on the whole process? Was it about 2 years? Buchheit: It depends where you draw the line, but it was a couple of years. I think some of the systems problems were a little bit harder than we realized to start with. I keep mentioning this idea of updating data quickly. It really soaks in at a lot of levels when you have to make your latencies be very low. If you have 168 Founders at Work a machine that's down, what do you do? You have to be able to respond to everything that goes wrong very quickly, so that's challenging. I was actually surprised to some extent at how positively some of the things we did were received. We were pretty nervous about some of our features. The idea of doing the whole thing in JavaScript--internally a lot of people were very unsure about that, but I think that our users loved it. It actually worked better than we expected it to. We were pretty nervous about it, because there are so many browsers out there and they all have plug-ins and some of these plug-ins will cause problems for you. It's really worked out better than we thought it would. Livingston: Earlier, you said "it worked out" for you. Most founders take the risk of starting a startup for the potential reward of a liquidity event. Did you get a bonus or something similar? Buchheit: There are lots of bonuses inside of Google and I don't know what the average is, but the bonuses in general can be very significant--much more so than at other companies. For me personally, I've been here long enough that there's only one bonus that matters, right? Which is part of why, for newer employees, things like the Founders Grants are much more important, because they're not going to get stock at a nickel a share or whatever. So something like a Founders Award isn't necessarily that important to me, but would be for newer employees. Livingston: What number employee were you? Buchheit: 23. Livingston: How did you join Google? Buchheit: I was working at Intel in the area and was kind of bored. I was looking around for something more interesting and I emailed Google my resume. Interestingly enough, the first time I emailed my resume, it bounced because their mail server was down. But I emailed it again the next day and it got through and they called me up. I came in and took a job. It worked out well, but it wasn't like I saw this company and said, "Oh wow, this is going to succeed!" I just thought it would be fun. It looked like there were some smart people and it was kind of interesting work--that it would be more fun than my old job. Livingston: Did you get any compensation for doing this project that was such a big success within the company? Buchheit: It's hard for me to know even, because, even after the initial stock grants, throughout the history of the company they've given follow-on grants. So I don't know what mine would have been if I wasn't working on Gmail. Livingston: I heard you came up with the famous "Don't be evil" principle. Can you give me the background? Buchheit: I believe that it was sometime in early 2000, and there was a meeting to decide on the company's values. They invited a collection of people who had Paul Buchheit 169 been there for a while. I had just come from Intel, so the whole thing with corporate values seemed a little bit funny to me. I was sitting there trying to think of something that would be really different and not one of these usual "strive for excellence" type of statements. I also wanted something that, once you put it in there, would be hard to take out. It just sort of occurred to me that "Don't be evil" is kind of funny. It's also a bit of a jab at a lot of the other companies, especially our competitors, who at the time, in our opinion, were kind of exploiting the users to some extent. They were tricking them selling search results--which we considered a questionable thing to do because people didn't realize that they were ads. Livingston: The users didn't know? Buchheit: Companies would just mix the ads in with the regular search results so people would think it was a search result. It's kind of like fake news or something. In a newspaper, they're usually pretty good about separating out which things are advertisements and which aren't. But the search engines at the time were all selling search results and mixing them in with the real ones, so it was a little bit of a differentiator that we always said that we would never do that-- and haven't. So it was all those inspirations, and I just thought it was a catchy little phrase. But the real fun of it was that people get a little uncomfortable with anything different, so throughout the meeting, the person running it kept trying to push "Don't be evil" to the bottom of the list. But this other guy, Amit Patel, and I kept kind of forcing them to put it up there. And because we wouldn't let it fall off the list, it made it onto the final set and took on a life of its own from there. Amit started writing it down all over the building, on whiteboards everywhere. It's the only value that anyone is aware of, right? It's not the typical meaningless corporate statement or platitude. Livingston: You mentioned that Gmail was "controversial" internally. Can you expand? Buchheit: I think, in general, people are uncomfortable with things that are different. Even now when I talk about adding new features to Gmail, if it isn't just a small variation or rearranging what's already there, people don't like it. People have a narrow concept of what's possible, and we're limited more by our own ideas about what's possible than what really is possible. So they just get uncomfortable, and they kind of tend to attack it for whatever reason. But for me, I am more interested in things that are new, and so I'm always excited just to see what will happen. That was actually one of the biggest reasons I joined Google in the first place. It wasn't so much that I was convinced that it was a good business; I just thought it was interesting and I was excited to see what would happen. Likewise, with Gmail, part of the excitement was just seeing how the world would respond. I kind of like uncertainty to some extent, because it's a little bit of suspense and excitement and adventure, almost, right? And you can learn a 170 Founders at Work lot even if things don't work out. But not everyone likes adventure. A lot of people seem to be against uncertainty, actually. In all areas of life. I'm suddenly reminded that, for a while, I asked people, if they were playing Russian roulette with a gun with a billion barrels (or some huge number, so in other words, some low probability that they would actually be killed), how much would they have to be paid to play one round? A lot of people were almost offended by the question and they'd say, "I wouldn't do it at any price." But, of course, we do that every day. They drive to work in cars to earn money and they are taking risks all the time, but they don't like to acknowledge that they are taking risks. They want to pretend that everything is risk-free. Livingston: Wasn't it controversial when you tried to test out the AdSense idea? Buchheit: Yes, absolutely. Everyone hated it. Many people were kind of mad at me because they didn't really go for the whole concept. It was something that had been talked about, and people agreed that it was not workable, it was not a good idea. So, to some extent, they were agitated that I wasted my time. Livingston: But you did it in one day? Buchheit: Yeah, pretty much. Livingston: And they were still annoyed? Buchheit: Different people to different degrees. There were only a few people who were sort of upset about my distracting from the main task. Other people just didn't like the concept, because it's obviously something that's very controversial and it isn't immediately obvious when you just hear about the idea and you haven't really used it. At first, it kind of seems a little bit wrong, right? Just because it's very unfamiliar. So it takes some getting used to. But people got used to it and then they were OK with it. Livingston: Most startup founders have investors, but you had Larry and Sergey to answer to. What's it like having them as your investors, in a way? Buchheit: I think it's probably reasonable. I've never had other investors so I don't have a lot of perspective, but they are very open to crazy ideas--more so than almost anyone I've ever met. I used to tell people my ideas, and then they'd explain to me that I just didn't understand how the world worked and why I was wrong about whatever. One of the exceptional things for me, coming to Google, was that it was the first time that I would tell people my crazy ideas and they'd say, "Oh, yeah, that's a good idea. I was thinking the same thing." So it was an environment with many people who are open to these kind of unusual ideas, and this is especially true with Larry and Sergey. Livingston: So they aren't "risk-averse" like so many investors. Buchheit: Obviously they consider risk and so forth, but they are definitely more open to the idea of something unexpected or different. Which I believe is very much their own thinking. Paul Buchheit 171 Livingston: What advice would you give someone who was working at a big technology company (that wasn't like Google in terms of encouraging new ideas) if they had a great idea that they thought could help the company? Buchheit: It depends on your situation. It depends how risk-averse you are. You should consider going to work at Google, start a startup, or go to another place where you are going to have that opportunity. For someone who's pretty far down in a company, if they are going to try to change the whole culture of the company, I'm skeptical. When I was leaving Intel, one of my managers there was trying to convince me, "You don't have to leave to do the startup thing. There are startup opportunities inside of Intel." Livingston: When you were working on it, were you working startup hours? Did it feel like a startup? Buchheit: Oh yeah. We had a pretty tight little team. We have really smart people and they are fun to work with. I'm not a morning person, so I'm always here at night. My normal hours were something like noon until 3:00 a.m. It's hard to go home at night, because you get working and you say, "I'm just going to make this one last improvement." Then, the next thing you know, it's 3:00 a.m. Livingston: Did it affect your relationship with your wife? Buchheit: No, it was nothing new. I've always been like this, so she was used to it. It's actually a much bigger change now, because I see her every day. But, as I say, for these people, it depends on their situation if they can take that risk of joining a startup or moving to a new city if they don't live in the right place. For me, I was actually single at the time, I didn't have a mortgage, so the idea of joining a little startup that may well be destroyed was just like, "That will be fun." Because I kind of thought, "Even if Google doesn't make it, it will be educational and I'll learn something." Honestly, I was pretty sure AltaVista was going to destroy Google.