Paul Buchheit (Google)


  • 2023.03.10 - The Social Radars - Paul Buchheit
    • Paul: (...) the very first version of Gmail that I built was built based off of just the Usenet code that I had been working on previously. So I mentioned earlier, the first time I tried to build an email project, I ran into this problem where I was writing a bunch of code for a while and then I got bored and distracted. And so one of the things I had learned over the years is that just my own psychology, I have to always be keeping myself engaged in the product. And so, for me, what I really like is to always be launching things. And so with Gmail, I was able to build the first version in a day and then launch it. And so what I did was I just took this Usenet indexing code that I had from the previous project, and then I took my email, my mailbox, and I just wrote a script to convert the email to look like Usenet data. And then I just shoved my email into this Usenet search and then I just launched it. I just sent it out to engineering to be like, "Hey, guys, I built this email search. Let me know what you think." And so that's the first version of day zero of Gmail was I launched this product.

      Jessica: What did people say?

      Paul: People say, "Well, it's useful, but it would be better if it had my email instead of yours." And so that's a feature request.

      Jessica: Yeah.

      Paul: And so it was all this iterative. So I built the first version in one day and then I got feature request, which is people wanted to search their own email, not just my email. So then I'm like, "Okay, now I know what to do." And so I went to work and I made this thing that would crawl through the home directories and suck up people's email and index it and then partition it so that it would just get one person at a time. And then I had ... That was version two, so I launched it again. Okay, now I have email search with your email. And so then people would be like, "Okay, this is cool, and now I want to reply to an email." I'm like, "Okay." So then I went in and added a reply feature and just iteratively would launch a new version every day or two. And then just constantly people would complain about, now they want to send email or whatever, they want an address book and just kept adding feature by feature or it's too slow or whatever the problem is and try to keep making the product better and better. And we continued through this process and there were a lot of complicated ups and downs in there as well. But, ultimately, we came up with this goal that it would be ready to launch to the world once we had a hundred happy users. And this is something we still use sometimes in advising startups because it's actually a pretty good way of working on a product because it's easy to get people to tolerate a product, but you some of them to actually like it, but you don't need all of them to like it. And so I would do ... I actually embedded just a little one-question survey inside of Gmail, which is a little box would pop up and say, "Are you happy with Gmail? Yes or no?" And it's just like, yes or no. And I would get a list of names then and I could go through the list of nos. I would just go talk to them.

      Jessica: Because they're all part of Google's engineering team.

      Paul: Yeah, it was just inside, so it'd be like down the hall. So I would just go down the hall and be like, "Okay, since you're not happy with Gmail, what's it going to take to make you happy?" And so just ... It's basically door-to-door sales. It's like, "What's it going to take to get you to be a happy user?" And for some people, they'd be like, "Well, basically it needs to be exactly like Outlook." And I'm like, "Well, that's never going to happen," so cross this person off the list. But other people, it's something much more attainable. They just needed one feature. There's like something they didn't like or whatever. So basically, there's some people who are easy to make happy and some people who are difficult. And so you just focus on the ones that are easy. And then eventually, we made it to a hundred. And then sometimes, eventually, I would get the Outlook people, too, because back then Outlook had this bug where once your mailbox on disc would reach, I think, two gigabytes, it would corrupt the data and lose everything.
      (...)
      Jessica: (...) Can I just, as an aside, say listening to you tell this story and saying that you were the kind of person who would keep going, you always wanted to be launching and you would talk to users, you're like the dream founder, I have to say. If I were fund ... I wish I could have funded you.

      Paul: That feedback loop is what makes it fun. So if I work on a project just by myself without that loop, I lose motivation. And so it's also part of the problem that have ... Fast forward, by the time I left Google in 2006, the release cycles on Gmail were like, you could check something in, but it's going to be months by the time it makes it all the way out to the world or whatever. And I just like fast iterations. So when we had FriendFeed later on, I would write the code, I would push it live, and then if it's good, I would check it in. I just like immediate feedback loops where you put something out there and then people love it or they hate it or whatever, and you learn and you go forward. This ponderously slow movement where you do something and then you don't see the effects for many months, to me, is just demotivational.