Guides
Guides to verifying AI output
How to check what a language model produces before anything acts on it: what each kind of check catches, what it cannot, and how to fail safely.

7 guides cover
- What each kind of check catches
- What it cannot
- How to fail safely
- How to verify LLM outputWhat each kind of check catches
- What fail-closed meansFor AI checks, and when fail-open is fine
- Deterministic guardrailsRules, parsers and solvers for agents
- Can LLMs do maths?How to check AI calculations
- JSON Schema for LLM outputChecking shape and values
- AI-generated SQLChecking a query before it runs
- Audit trailsWhat each decision record holds