AI Guardrails

Six prompt patterns with evidence behind them, the mechanisms that make them work, and the checks that keep them working as models change.

Your LLM will try to do things it shouldn't. Here's how to stop it.