For reference: the bottom of this proposal describes a similar content ranking system, although it doesn't encapsulate brevity and it includes a-0 = "you're stupider for reading it." The full list is below. Each has more description, but I wasn't sure it made sense to quote that much text here.
- a-5: Content is definitional in its field. […]
- a-4: Very high-quality documents. […]
- a-3: Quality presentation of a concept or topic. […]
- a-2: Informative but not exceptional. […]
- a-1: Not obviously false reporting. […]
- a-0: You're stupider for reading it. […]
The author also notes
There's a tremendous amount I don't classify, of course, and some (though not all) of that would fall under a negative level classification.
I've also thought of creating an upper-level classification, a-6, for materials which introduce a (currently) entirely novel concept. Strictly, works such as Newton's Principia, Darwin's Origin, or Claude Shannon's information theory papers should be here, though they now largely belong to history as opposed to current development. I may move them to this realm.
I think that this raises a valuable point, which I would characterize as a distinction between a work that summarizes or explains a concept that has already been presented elsewhere (Arbital's math section would be this) versus a work that explicates something as an authoritative, original source (which would include some of the AI Safety content on Arbital).
This is slightly different than the person I'm quoting here, in that something doesn't have to be defining of an entirely new field in order to itself be a new concept. So maybe there are three categories, with one being "this is an original extension of an existing concept". Much of what I've written falls in this category.
It seems like this kind of metadata might be quite valuable to have on Arbital pages, especially as it grows as a content platform.