The soft link is Everything2's killer app.
E2 is a website concerned predominantly with text: reading, writing and textual interaction. Soft links primarily facilitate the first (and, since guest users may not write or interact, arguably the most important) of these.
It works like this. When you navigate from one node to another node - via hard link, soft link, firm link, or manual search - you leave "footprints". Everybody does, in fact, the whole time they are on this site. E2 records these footprints. The soft links at the bottom of a node are actually a list of the other nodes which have the most footprints leading to them. The first soft link is the most popular destination after reading the node, and the rest become progressively less popular. If there are fewer than the maximum number of soft links then this node is poorly-travelled, and not enough footprints have been made.
By going back and forth between two nodes multiple times, noders make the tracks deeper and the most popular destinations rise to the top of the soft link pile.
Fairly simple stuff, yes?
Okay, let's look at this thing from the point of view of a passive reader. You finish reading the node and finally, at the bottom, you are provided with a list of possible further destinations, with the most popular destinations towards the top. In other words, you get to see an image of the footprints - you see what other people were thinking when they finished the node before you. You can see their collective train of thought.
The resulting soft link pile isn't necessarily coherent, or even relevant to the subject matter, but hell if it doesn't turn up something worth reading, nine times out of ten. And each node after that has more soft links, so naturally you open more browser tabs and keep reading. Presented with such a wealth of interesting things to read, the result is hours spent wandering across E2, reading dozens or hundreds of recursively interesting nodes, getting sucked in - and leaving a trail which, by the very nature of its construction, is likely to suck in others too.
This is the entire secret of how E2 works. No other site has this.
Wikipedia and other wikis allow hard links, which, for factual writeups, are great. A factual writeup (e.g. stars) will almost certainly contain references to other related subjects which an interested reader would wish to pursue (planets, hydrogen, cosmology). These could be justifiably hard linked. But E2 contains much which is not strictly factual: for example, fiction, poetry and daylogs. Somebody reading a poem about a star isn't going to come across the word "cosmology" and think, "Hmm, I want to find out about cosmology", they're going to want to read more poems by the same author, or more writeups on the subject of distance and loneliness. Hard links and pipe links can help with this, but aren't ideal. They can be obstrusive or inappropriate.
Another approach is tag-based systems, such as those used by Digg and Slashdot and, these days, many blogs. These rely on users to manually input metadata which they think is relevant and which they think will lead to more relevant content. But this requires conscious effort and relies on the consistent intelligence and diligence of the users. And it's a pretty tall order for a site with 400,000+ nodes!
With E2, we do still have those options (hard links and a forthcoming tagging system - plus, it's possible to manually tread and re-tread soft links if you can be bothered to do so), but they take a back seat to the natural (organic?) soft linking process, by which noders contribute that metadata subconsciously, according to whatever they were thinking about at the time, and by which E2 is gradually bound together, tighter and tighter, by nothing more than the continous browsing of all its users. Huzzah!