When I wrote SdiDesk I explicitly decided that I wanted many small, hand-layed-out diagrams which are hyperlinked together, rather than automatically generating a large diagram. (Of course, it helped that that was the lazy thing to do, in terms of coding 🙂
Manually drawing diagrams doesn’t help with *discovery* of new connections or of other structure but, for me, it’s the right way to capture some of the structure in *my* thinking. I tend to create diagrams with around 10 (15 max) nodes where I might be deliberately representing a hierarchy or a supply-chain or some other small chunk of the world. And I can arrange the nodes in a way that makes sense for that kind of thing. If I really need more nodes, I probably want to break it up into an overview page and sub-pages with details. Of course, SdiDesk nets are more primitive than even I want, visually (so yeah, I am thinking about how to improve on that in my next opus) But I prefer them to Buzan-style mind-maps where I find most software forces me to build one “butterfly” shaped hierarchy. Or to something like TouchGraph which is very clever but actually not much use when you want to get acquainted with the shape of your information.