# RAG: Your AI Has Your Files. Can It Find the Answer? Canonical: https://brew.new/browse/templates/email/pt1_k97yqpmpfagvr1ndpmace3f7vs8e2rdp Brand: substack.com Category: newsletter ![Preview of RAG: Your AI Has Your Files. Can It Find the Answer?](https://cdn.brew.new/email-preview-2948c5ffca8ae990-tracking_r57h64mtt1mmtb25s4n39n83as8e0jy8-1788974078252.png) ## Email content Forwarded this email? Subscribe here for more RAG: Your AI Has Your Files. Can It Find the Answer? Build your first RAG assistant, test it with 30 questions and check the evidence behind its answers and publish on GitHub. ChatGPT Sep 8 ∙ Preview READ IN APP You can save every important detail and still spend the next evening explaining the same project again. The information is somewhere in your files. You remember writing it down. But when you ask AI a question, you are back to opening documents, copying passages and deciding what to paste. That is the next problem in our series. In Memory Engineering, we explored how to preserve useful information. This fifth article asks what happens afterwards: how does the right information reach the model when you need an answer? Think about the workshop example we have been using. One record contains the approved budget. Another contains a supplier quote. An old draft carries a different number. Saving all three is straightforward. Finding the evidence that should govern today’s answer takes more care. Here is something you can check immediately. The next time AI answers from your documents, ask it to show the passage supporting the answer. Read that passage yourself. Does it establish the claim, or does it merely discuss the same subject? That small distinction matters more than how confident the answer sounds. RAG, short for retrieval-augmented generation, adds a search step before the answer. It brings selected evidence into the model’s working context. It can also retrieve the wrong evidence, which is why our project includes questions designed to expose that failure. If you are joining us here, AI Foundations explains the underlying system. Prompt Engineering and Context Engineering explain the instructions and information we will put to work. You can follow today’s project without having the earlier files. In this article, we will build a searchable workshop assistant using supplied files, a runnable starter and thirty test questions. You will inspect its evidence, repair a retrieval failure, update a source and publish a project you can explain on GitHub... Keep reading with a 7-day free trial Subscribe to ChatGPT to keep reading this post and get 7 days of free access to the full post archives. Start trial A subscription gets you: Subscriber-only posts and full archive Post comments and join the community Like Comment Restack © 2026 ChatGPT 548 Market Street PMB 72296, San Francisco, CA 94104 Unsubscribe Get the appStart writing [Open and remix this design](https://brew.new/browse/templates/email/pt1_k97yqpmpfagvr1ndpmace3f7vs8e2rdp) [Browse email designs](https://brew.new/browse/templates)