Earlier quoted context omitted.
The details of what I record (in another comment[1]) might be idiosyncratic and uninteresting to you. Those details aren't important. The important point is that so much information belongs in multiple places. A single statement, let alone article or book, might regard all of economics, justice, common sense and beauty. It might be about parsing, Haskell, and keeping data types flexible. I'd like to be able to search…
How often do you look stuff up that's older than say a week? Personally, I'm kind of glad to soak up information as I go and being able to just forget everything else. As opposed to constantly feeling the need to tag, sort, archive, reference, cross-reference, keep track of information, that future self will to 99.9% never actually need. Yet, I wonder whether use cases start to materialize, once you start using it...
Federation could reduce the hurdle -- both reducing the need to ingest, and the difficulty of doing so. Borichevskiy (author of article that generated this thread) nods at the idea of using someone else's index of books. Such labor-sharing could be much more general.
I don't want to suggest, though, that ingesting something is a waste of time if you don't review it. That's true of boilerplate metadata like ISBN numbers. But for original ideas, or for paraphrased insights gleaned from reading online, I find that sticking it into the graph produces value at the moment it's done. The rigor of compressing an idea to a few words, and determining which adjacent concepts it bears on, and how, has helped me understand things.
(Teachers often ask their students to state things in their own words. It's like that, but faster and harder. Like punk rock for essays.)