Steve Huffman talks about Reddit's early days, including how they solved the cold-start problem by seeding content themselves and the product decisions like adding comments and self-posts. Paul Graham and Y Combinator pushed them to launch before they felt ready.
Chapters
I know in the beginning you guys were curating the content yourselves, and that had to be a pretty arduous process.
Yeah, so somebody asked me before I came up here, like, how do you get over the cold start problem? And I think that's actually one of the hardest parts about starting any business, is how do you get your first few customers? For us at Reddit, it was important that, you know, we're building a community. So nobody wants to be a part of a community that's empty. Nobody wants to be in a ghost town. So what we did is Alexis and I every day submitted all of the content. I had a script that I'd run that would scrape a bunch of websites like Slashdot and New York Times and Boing Boing and this and that. And it would list all of the URLs that I had never seen before. And so on the first day, it was just a bunch of garbage. Like all these navigational links and stuff like that. But on the second day, it was just all content. And so from then on, I would just run that every morning. I'd get a big list of maybe 100 new URLs. And I'd read them all. And I'd submit the good ones to Reddit. And we had a special submit page. And Reddit's submit page at the time just had fields for title and URL. And mine had a third field for username. So I could I could put in the URL, I could put in the title, I could make up a username, and then when I hit submit, it would automatically register that user and then submit the content as that user. And so we just did that for the first few months to make it look like there was a community there.
At what moment did you realize there was a community there?
Yeah, that was in August. So we started in June. Sometime in August, I took a day off, and I don't remember why. I think I was just tired.
Understandably.
And at the time, the way the front page algorithm worked is only links that were under 24 hours old would be shown. And our front page was 25 links long. So if there were fewer than 25 links submitted that day, the front page would actually just start to dry up. It would just be white because we didn't have 25 links to show. And that was a problem sometimes. Like on days where we were really busy and we didn't and submit a bunch of links, you might go to Reddit and there'd be like 5 links there. And so I took the day off. And at the end of the day, I had this panicked realization. I was like, I haven't been on Reddit all day. It's probably blank. And so I open up Reddit and I look, and it was full of content. And that was a really special day because that's when we knew that we had real users submitting real content.
How'd that feel?
It felt great. The hair still stands up on my arm when I tell that story. It was a really special moment. 'Cause at that point we were— up until that point we limped into Y Combinator. Everybody around us was much smarter than we were.
I know that also at one point you were pushed out of the nest to even put Reddit live, right?
Yes, yes. We had a special relationship with Paul because of how we got in and so he was always He was just badgering me. Largely his advice was directionally correct. And it's still kind of the cornerstone of Y Combinator's advice is launch soon. Launch as quickly as possible. 'Cause that's when you start learning. And so he would just every day, every day. And I'm like, Paul, it's been like 2 weeks. And so he sent me this email. At one point that was like, I don't know why you haven't launched. This is on June 22nd. We started on June 1st. He was like, on June 22nd, he's like, I don't know why you haven't launched yet. It's either because you're not capable of doing it or you're just waiting until it's perfect. And I don't know which is worse. And I was just like, kind of mad. And also, this is my idol chewing me out a little bit. I sent him a link the next day. I was like, here it is. And it's like I sent him this link that was running off my laptop of the first version of Reddit. And then 2 days later, he wrote a blog post about startups or something and he linked to it. And so I had this screen next to my main monitor that just had the error logs for when I was developing. And all of a sudden it just starts scrolling really, really fast. And it's because Paul launched us without telling me by linking to Reddit from his essay.
That's awesome. That's awesome. So now that you have the community and it's starting to grow, what sort of things are you— you're looking at your users, what sort of information are you getting from to iterate on the platform and what are you learning?
You know, there's lots of, I guess, kind of theories about product development and this and that. We weren't really collecting data the way you would now. We barely knew how much traffic we had. But we were users, and so a lot of what we were doing in those early days was just following our own product intuition. Like, what would we like to see? And sometimes users would be asking for features and they'd have a good idea and we'd do that. But often users don't ask for the feature, the correct feature, but behind their request is a problem. And if you can kind of figure out what that problem is, then that's a good starting place to think about what should I be building. And so there's a million small things. One of the examples, an example of one of the biggest ones was, it was a little bit later, I think it was that fall. We had, what users would do, it was after we added comments. So the first version of Reddit didn't even have comments. So we added comments, which of course the first comment was about like, this is the end of Reddit. And what users would do is they would submit a link to the comment page for itself. Because the ID of each link was in the URL. It was just a counter. And it still works that way, right? You go to reddit.com/r/subreddit/ and there's a number there. They would guess what that number would be for their post. They'd take the newest post and basically add 1 to that, and they'd submit a link to the comment page for itself. And so it's like a self-referencing post. So instead of— because at the time everything on Reddit was a link to some external site. And so this would create a link to a comment page, which was pretty cool. Now occasionally, like, users would get it wrong or 2 links would be submitted at the same time, and so they'd submit a link to somebody else's comment page. But this is where, like, AskReddit started. Users would say, AskReddit, and they would just link to an empty comment page and people would answer. And so we saw users doing this, and it was kind of technical and confusing. So we made it a little bit easier. So instead of having to guess at the URL, you could just type in self in the URL. And that would make automatically the self-referencing link. Self-referencing link. And now these days, those self posts are 60% of our content. And that idea— I don't think— we would have never built that on our own. We built that because the users were already doing it for us.
