Ask HN: How do you create a successful community from zero?

Depending on the type of community you need to jumpstart, intentionally leave off commenting as a feature in the very early beginning if you can. For example if you're starting a Reddit-like site of link posting and voting. While the comments will drive the community later, the comment system reveals the sockpuppets have no souls. Alternatively of course just try to post as many stray comments as you can, the problem is that's extremely time consuming at any scale.

Also shorten the amount of content shown on the front page. So for example once your community is very successful, using Reddit as the example again, you might expand how many posts you show on the frontpage to 30. In the beginning, start it at 10, 12 or 15. That way if you need to post via sockpuppets, doing 20 or 30 links per day will properly fill the frontpage and then some.

The approach I last used to sockpuppet start a community, was to create ~300-400 or so fake user accounts, then have a posting system that takes content I want to post, and randomly stack load the content to be posted into the future (add in some forced shake to the random so it's not too obvious what the posting time thresholds are). Run a job against that, when the next piece of content is ready to be posted assign a stray sockpuppet account to 'post' it. The reason for the stack loading, is so you can do all the content posting for a day, or multiple days, in one rush to optimize the time you're spending; so you can load up 30 or 90 posts in one go and the system will jigger the timing between them for when they get publicly posted, so it seems natural.

In my case I also had voting on the content and used a decaying approach to the scores (so content scores melt, cycling the content off the frontpage, comparable to a Reddit or HN approach). I put together an approach where each sockpuppet piece of content had a preordained max score that would be reached over the course of three days at a declining rate of voting (eg day one the content gets 75% of its max votes; day two it gets 20%; day three it gets 5%). Then I added a job that throws a random stray vote against any piece of content in the system every N amount of time to add some more projected fake activity into the system. As necessary it's pretty trivial to simulate each sockpuppet doing the fake voting as well, if you need to show sockpuppet activity on a profile page or such. Some sockpuppet content would randomly receive very low preordained vote scores (1 in 10 posts or similar), eg 0 to 5 votes as the max outcome, so that not all content would seem popular.

I made a possessor sockpuppet account, so that if I logged into that super account, on each page load it would put me into the shoes of a different sockpuppet account, so I could browse around the site leaving comments or doing any manual behavior and each action would go to a different sockpuppet without having to log in / out constantly. That further sped up manual intervention actions I had to take to boost the appearance of human activity.

Finally there was an ability to follow other users and get a feed from that (of their posts). So I simulated sockpuppets following and unfollowing eachother. That was annoying to get right but not terribly difficult and it works well as a good simulation that there is human activity on the site.

Naturally you probably - depending on the community - don't have to fool highly observant tech engineers, you have to fool average users that have no idea about the technical aspect of simulating something like this.

The point of it of course is that, if such a proto community doesn't exist, its going to be very difficult for a community to grow.

The hugely successful SWSX campaign moved Twitter out of the tech-only sphere into other areas. That got non-tech people using it, but by the time of the time of the Aston Kutcher vs CNN race to a million followers (less than 2 years later) it already had its own momentum.

I remember that was how plentyoffish started.