Learn · Embeddings
How embeddings work
AI turns words into vectors — lists of numbers — so that similar meanings sit close together in space. Click a word to see its nearest neighbours, or run an analogy and watch the maths happen. This is a hand-built 2-D map; real models learn hundreds of dimensions from data, but the ideas are exactly the same.
Click any word. Distance ≈ similarity — closer means more alike.
Words → numbers
Each word becomes a vector. Here it's 2 numbers (x, y); real models use hundreds, learned from huge amounts of text.
Close = similar
Meaning becomes geometry: related words cluster together, so "distance" is a measure of similarity (real models use cosine similarity).
Meaning is directional
Directions carry meaning too — the step from man→woman is the same as king→queen, so you can do arithmetic with meaning.
This map is hand-built to illustrate the concepts. Real embeddings are learned from data and live in hundreds of dimensions — we're projecting the idea down to 2-D. Live embeddings for any word (and true semantic search) are coming with our backend work. Want the theory? See the 3Blue1Brown and word2vec videos on our Learn shelf.