Retrieval-Augmented Generation

Retrieval-Augmented Generation is the contrast class used in Brevvie Python + AI — Section 13: Build Your Own LLM Wiki (Karpathy / Obsidian). The source describes RAG as a system where documents stay in a source folder or vector index and relevant chunks are retrieved at question time.

Contrast with LLM Wiki

• RAG re-synthesizes answers from retrieved chunks each time
• an LLM wiki converts sources into persistent cleaned-up pages ahead of time
• RAG retrieves; the wiki accumulates

Why it matters

The section does not reject RAG. It argues that a personal wiki becomes more valuable when knowledge is stored as maintained notes instead of rediscovered from scratch on every query.