VECTOR FORGE
Embeddings & semantic search dojo
0 / 900 XP
Zero to production, in order

Nine belts. No assumed knowledge. Real vectors throughout.

Starts with why keyword search fails — no math yet — then builds up through toy 2-number examples, real similarity math, a real embedding model running in your browser, and finally the production tradeoffs (chunking, ANN indexes) that RAG systems actually live or die on.

White → Brown: concept checks & hands-on, all free, all in-browser. Black & 1st Dan: AI-graded reasoning via your budget-capped Worker.

Embedding model not loaded yet — loads on first hands-on belt (~25MB, once).