RAG

Retrieval-Augmented Generation (RAG): LLMs retrieve relevant chunks from uploaded documents at each query.The Kaparthy LLM Wiki

Characteristics

• Knowledge rediscovered from scratch every question.
• No accumulation or persistence.
• Examples: NotebookLM, ChatGPT file uploads, most RAG systems.

Contrast with LLM Wiki

• RAG: Ephemeral retrieval.
• LLM Wiki: Proactive maintenance of persistent wiki.