Each point is a philosopher. Its position reflects semantic distance between ideas — computed from 3,072-dimensional vectors generated by OpenAI's text-embedding-3-large.

For each of the 1,000 philosophers, gpt-5.4 produced two texts: an intellectual profile — dense, in the tradition's own vocabulary — and a philosophical stance — a period-agnostic cross-tradition analysis. These texts, together with key concepts, were embedded into high-dimensional vectors.

UMAP projects those 3,072 dimensions onto a 2D plane while preserving local distances: semantically similar philosophers appear clustered together. The KNN network built on these vectors achieves Precision@K=5 of 0.827 — over 8 in 10 nearest neighbors share the same period or school of thought.

Colors indicate the historical period. Use the settings icon to adjust repulsion strength — higher values push different philosophical communities further apart — node appearance, and switch between 3d and 2d terrain mode.