RAG: AI that answers from your documents

The lessons, in order

1. Lesson 113:14

   What is RAG? How AI answers from your own documents

   9 min read · 11 chapters · 11 short clips

   RAG, retrieval-augmented generation, is how an AI app answers from documents the model never read. It finds the right pages, adds them to your question and has the model answer from them. This lesson follows one question through the 4 steps, shows the message the AI really receives, and names the 3 ways it still goes wrong.

   1. A washer error that no AI could answer1:29
   2. RAG is an open-book exam1:15
   3. Why not paste the whole manual?1:27
   4. Step 1: prepare the documents1:28
   5. Step 2: retrieve the closest pages1:16
   6. Step 3: the message the AI really receives1:14
   7. Step 4: an answer you can check1:21
   8. Where you already use RAG1:00
   9. 3 ways RAG still gets it wrong1:20
   10. The 4 steps, and a quick test59 s
   11. RAG makes AI informed, not smarter48 s
2. Lesson 214:57

   How RAG search works: embeddings, hybrid search, reranking

   11 min read · 10 chapters · 6 short clips

   How does a RAG app find the right page when your words are not the document's words? This lesson runs one washer manual through a real search, from keyword search to embeddings and cosine similarity, shows where meaning search is blind, and tests the 3 fixes builders use, hybrid search, context notes and a reranker, with the scores we got.

   1. The water just sits there1:11
   2. Search by words1:38
   3. A map of meaning1:39
   4. How close is close?1:32
   5. Where the map comes from1:04
   6. Where meaning search is blind1:36
   7. Cutting the cards
   8. A second opinion
   9. The search, rebuilt
   10. Bad answers often start with bad search