I built a 3D visualiser that overlays semantic topics and sentences of documents with different "embedding types and orders"[1] using Natural Language Processing.
It colours each document with the same colour. You can see how two documents overlap, semantically - it's pretty awesome for Job Role/CV overlays for example, or educational resources and exams - a mix merging of colour shows both equally discuss something... missing colour means one document doesn't.
Since it's semantic, depending on the embedding, the gaps in-between spots make a lot of sense intuitively, and you can sometimes even see how the conclusion of a document ends up in a different semantic space to the start as such (even though there is no time data, you just notice the later topics are semantically in a different space to the earlier ones for the same document)
[1] Similarity based, NLI based, GPT-raw etc.