Blog

Testing AI systems
notes from the course

Short, practical pieces on the things testers actually get stuck on: how to tell a hallucination from a retrieval failure, what to assert when there is no single right answer, and which controls are worth having.

2026-08-19 How to detect LLM hallucinations in testing Four layered checks that catch a confident, well-written, false answer: grounding, self-consistency, LLM-as-judge, and the unknown-entity probe.