It’s how llms work… The points in vector space are basically tokens. The distance is correlated to aspects of things/concepts

So you can use geometry to search the vector space and get very good results, it’s mainly used to store/index information for later retrieval