I can speak to this with some authority. I led a 20-person tigerteam several years ago in order to investigate 3rd party Reddit search. I've been invited by four different admins to discuss taxonomy. And when Karmawhores debuted, I was #3 comment karma, #8 well-rounded.

And I'm here to say - dude don't know what the fuck he talkin' bout. Neither do you, neither does Team Hubski.

HERE IS THE PROBLEM

Any aggregator - I don't care who they are, or how much money they have - becomes a piece of Borges' Library. There is bottomless information there, expanding out from singularity at the speed of light. As time goes on, parsing it grows less and less possible; what you want is there, but you'll never find it.

Reddit has no taxonomy whatsoever. Want a subreddit? Make a subreddit. Want a better subreddit? Make a new subreddit and hope the people you like come from the old subreddit you hate. It's a hashtag problem - you follow #yolo because you're a douche, then all these wannabe douches start tagging everything with #yolo, so now you have to tag things with #yolo4lyfe (because some poseur already took #yolo4life) and before you hit 140 characters your interests have been balkanized beyond all recognition.

Things would be better if Reddit allowed spreading of content across multiple subreddits (something I started discussing with the Old Guard back in 2009). That fucks with their power structure, however, and will be even more crippling to their servers. Hubski has an advantage here in that we've got up to 3 tags - something I lobbied for hard. It's imperfect, though, because the only way you know that #writing is a wasteland and the good stuff is in #writebetterdammit is through folklore.

The library of babel has no card catalog; it just has gurus on stacks of books pointing you in the general direction of Canticle for Liebowitz.

HERE'S THE SOLUTION

Any aggregator needs a plaintext, WYSIWYG, idiot-grade search and collation engine so that if I submit an article about Vonnegut, the system KNOWS that it might go in #goodlongread, it might go in #writebetterdammit, it might go in #writing, it might go in #humor. The system then PRESENTS enough information about any possible tagging that the user can correctly assign one or multiple tags to the content such that it's a non-action for him. Additionally, the system presents other users with the content AND other places where the content belongs so that other information can be presented. Finally, the system connects the content with those interested in the discussion such that they perfect the web.

Basically, pearltrees without the java-heavy fail.

Hubski works because it's small. It still doesn't scale and the way it's going, it won't. Reddit ceased to work as soon as they hit ~100k real users (best guess is their numbers are off by a factor of 10). Hubski may scale to usefulness if it can grow slowly but every catastrophic injection of users from Reddit is a body-blow.

You do not grow the library by dumping books on the loading dock. You grow the library by building shelves, hiring librarians and investing in the infrastructure so that your members can always find what they need, when they need it, where they want it.

Reddit is now and has always been about OH SHIT WE GOTTA LOTTA BOOX and fuck all the rest. Hubski thinks of itself as a charming little book club in the corner of the coffee shop, while also insisting that there's room for everyone. I'm here to tell ya - if you don't figure out how to parse a large dataset FIRST, you're going to hit the wall as soon as your schema fails.

Everything else is hand-waving.