Embeddings — how AI represents meaning

An embedding is a list of numbers — a vector, typically 256 to 4096 dimensions — that represents the meaning of text. Texts with similar meanings get vectors that point in similar directions, so similarity becomes measurable geometry.

Cosine similarity measures the angle between two vectors: 1.0 means they point the same way (very similar meaning), 0 means unrelated. It is the standard way to compare embeddings in semantic search and RAG.

Embedding spaces learn consistent directions for concepts like gender or royalty from co-occurrence patterns in text, so adding and subtracting vectors moves you along those concept directions.