Roughly speaking, what this means is that the individual files/concepts/notes (whatever you want to call them) in my PKB are generally about a single concept (as an example, "Amino Acid"), and are about only that concept (Proteins, which consist of Amino Acids, have their own separate note, which links to the Amino Acid note).
However, I am not dogmatic about these rules:
Early in a learning cycle, I will generally have very "unatomic" notes such as "Biology" which will contain a large mess of knowledge.
I prefer to follow the dictum of "Be Useful!" when determining how many links to have or what kind of structure to use (Associative vs. Hierarchical)
I'm also generally pretty accepting of duplication. I've found that the same knowledge can be useful in different places, and it's not necessary to attempt to "unify" them, especially if you are solving diverse and different cognitive problems. Eliminating redundancy is sometimes a novel theoretical insight, and sometimes it's just fake work - over time, I've come to realize that it's more often the latter than I would normally guess.
However, I view those first two axioms as a useful long-term perspective on where I'd like my notes to eventually "end up", as I go through multiple rounds of deepening my knowledge and refining them. The main reason for this is that it allows for accretion: over time, it remains easy to add new concepts, add new knowledge, and draw connections between potentially very disparate concepts.
Another heuristic I try to follow is to make sure that my PKB isn't a bad replica of Wikipedia. Therefore, I frequently link out to external resources, but when I do choose to write about a concept directly in my PKB, I will typically write it with the idea of "refreshing knowledge" in mind.
That is, I ask myself, if I come back to a concept two years from now, having forgotten much of it, what would I want to read in order to help load the most relevant parts of that concept back into my brain as quickly as possible?
Primers. For general concepts that you need to occasionally think about, write a primer that will serve as a rapid refresh of the core ideas and trade-offs that you should consider. Generally speaking, these are most useful when written at the level of concepts, models, and techniques which tend to encourage a foundational understanding of the subject.
The general point being, having a folder called Databases/ that contains everything I've ever learned about databases, while it may be useful for some other cognitive problem, is probably not the best way for me to ensure I'm applying my database knowledge effectively while writing software.