Vector Search, Explained Visually — How Databases Find a Needle in 5 Million Vectors

Vector search across millions of embeddings is explained using a supermarket analogy — brute-force search is a random pile on the floor, while ANN (Approximate Nearest Neighbour) search uses "aisles" via clustering and centroids. FAISS implements this as IVF, trading perfect accuracy for speed. Real apps also need persistence, metadata filters, and live updates — which is why vector databases exist beyond raw search libraries.