💠on tool assisted brainstorming Creating new entities just to make relationships is dangerous because it makes the graph less dense rather than more. Idea adjusted into
- Can create a relationship with a collection, which could be with the collection itself or if the collection isn't populated could make an item within it specifically based around the relationship context. Arguably doesn't cost since the collection would be populated anyway.
- When creating a new node, before writing the general node content, could first pick the 'random' other nodes to form relationships with. Can then write the content with those nodes in mind, and finally get the semantically related nodes, then finalize the relationships. This just makes sense I think since the node ought to be modified by its unexpected relationships. If a merchant is related to some tiny distant nation, that should be part of its general impact content rather than a side note.
I kind of don't see any issues with this approach. I think I can just implement it. I wonder if relationships should be pages so that they're easily read from the context of any of their related nodes