கKatralAI

Campus Katral · Lesson 2 · 12 min

Closed book vs open notes

In one sentence

RAG is retrieval-augmented generation: give the model the right page before it answers.

Nila

Nila

Campus mentor

AI is not a person. It is a next-word engine with a very good memory of patterns.

Read the idea, watch the picture, then try it with your hands.

Watch the idea

The concept

Question

What you ask

Retrieve

The right page

Answer or refuse

Never invent a citation

Closed book vs open notes

A base model only has what it was trained on — a closed book, frozen in time. Retrieval-augmented generation (RAG) finds relevant pages first, then asks the model to write using those pages. It is an open-notes exam.

Retrieve, ground, then answer

Search must return the right passage, not a lookalike. Grounding means every claim maps to a span you provided. If retrieval is empty, a good system refuses — it does not invent a citation to look academic.

RAG is not magic memory

Bad chunking, stale indexes, and missing metadata create fluent errors with footnotes. Fine-tuning changes the model's habits. RAG changes what is on the desk today. Most campus and work assistants should start with RAG.

Worked example

What does Katral mean?

Closed book: the model may wander into a river in an old poem. Open notes: your glossary says Katral = the act of learning. The second answer is boring and correct. That is the point.

Picture to keep

Retrieve → Ground → Answer

  1. 1Search your notes
  2. 2Place them in context
  3. 3Answer or refuse