RAG isn't magic memory. It's a technique for finding the right page in a document before answering a question. Think of it as an open-book exam — not guessing from memory.
Let's think about it this way for a second
The model hasn't memorized the whole world. Think of it as a well-spoken young person who's never actually read your notes. RAG stands for Retrieval-Augmented Generation — or in plain terms, "search first, then answer." It's like asking a question at the library counter: the librarian fetches the book first, then reads you the answer. Fine-tuning is like teaching something straight into the brain; RAG just keeps the book close by instead. For beginners, RAG is the safer choice — if it gets something wrong, you can just point back to the page.
Let's connect it to everyday life
Ask a plain chatbot "Can I use my phone under the school rules?" and it might just guess. A RAG setup instead searches handbook.md, reads the phone section, and answers using only that text. If it can't find anything, it says "not covered in the document." This course won't front-load you with vector database jargon — just pack away these four steps: search → read → answer → stop.
Let's try it hands-on together
Closed-book chatbot: question → guess from memory → answer
RAG: question → find pages → read → answer → stopYou'll be able to explain RAG as an open-book exam.5-Minute Try-It
Write down 3 questions you'd want answered. For each one, note whether it should be answered from memory or by searching a document first — and explain why.
One Quick Warning
RAG doesn't guarantee the truth. Find the wrong page, and it can be confidently wrong. Only pointing back to the source page keeps it honest.