Testing LLM Agents Like Software – Behaviour Driven Evals of AI Systems | Not Hacker News!