latentspace.

latentspace.


I hope this blog itself can function as a latent space, where each note or project acts as a latent vector—and through their combinations, both similar and unexpected notions may emerge, revealing something interesting. In a latent space, nearby points share underlying structure; points far apart may still be interpolated, producing meaningful transitions that were never explicitly written. The same should hold here.
Each post is embedded into a high-dimensional vector based on its content, then projected down to two dimensions. Posts with similar content land near each other. The edges between nodes represent related posts, generated by AI based on conceptual dependency and natural reading flow between posts. Node colors are assigned by category, derived from each post's emoji label.

There is no prescribed order here. Pick any node, follow an edge, or jump across the map—the connections you find may depend as much on what you bring as on what was written.