What an embedding actually encodes

An embedding maps text to a point in space where nearby means related. It captures topic and usage, not truth, and definitely not recency.

What this lesson covers

• Read an embedding as a position in semantic space
• Know what embeddings fail to capture
• Choose an embedding model for your content

This lesson sits in Embeddings and Vector Search, part of Retrieval-Augmented Generation from Scratch. It assumes what came before it and leads directly into the next lesson in the module.