I recently wanted to find some information about how-the-site-works (because I was trying to use the UI to do something[1] and not getting the results I expected) and could not find any easy way to look up this kind of information. Information is distributed across many posts of varying ages, and I can only determine whether something is still accurate by finding more recent information that contradicts it. Old posts may have screenshots depicting an older version of the site, where even if the same functionality is still available now, it doesn’t necessarily work quite the same. I think documentation of site features needs to be available in one place that is kept up to date, such as a wiki.
- ^
I didn’t understand until recently how to filter Latest Posts by tags. As it turns out, the “Add Tag Filter” button does not, by itself, filter anything: it only adds to the list of tags on which I can then apply a filter, and selecting a tag that is already on this list (which the menu by default invites me to do, offering a list of the core tags) does nothing (giving me the least possible information about what the feature is supposed to do). Hovering on a tag name, where I would have located the answer if I had paid close enough attention, brings up something that looks almost exactly like a non-interactable link preview but for one extra line of mostly low-contrast text, and after failing to notice this important distinction I went on a long detour around the site looking at old feature announcement posts.
Has anyone tried to map relationships between (at least some) LessWrong posts?
What I’m looking for would be some kind of overview of what connects to what that could be parsed without clicking through links recursively. If I assume a chronologically earlier post cannot refer to a later one, I expect this type of structure to have interesting properties. A practical application that comes to mind is to inform an algorithm to decide what to give attention.
To define a relationship, links from one post to another would be a good if imperfect metric (not all links represent the same type of relationship, and not all relationships are represented by explicit links).