Why RAG?

The Three Problems RAG Solves

1. Stale Knowledge

LLMs have a training cutoff. GPT-4o's knowledge stops around late 2023. Any newer information is unknown.

2. Hallucination

LLMs are optimized for coherence, not truth. They confidently make things up when uncertain.

3. Private Data

Your internal docs, codebase, customer data — the model has never seen them.

The RAG Solution

Query -> [RETRIEVE: Search vector DB] -> [AUGMENT: Add to prompt] -> [GENERATE: LLM with context]

When to Use RAG vs Alternatives

Your Turn!

Identify the best approach for each scenario:

scenarios = [
    'A legal AI that must cite specific court cases',
    'A creative writing assistant for poetry',
    'A math tutor for solving equations',
]