Asimov's Zeroth Law of Robotics: testing and observing AI (ExpoQA 2026)

Nicole van der Hoeven explains how AI evaluation terminology (evals, graders, LLM-as-judge) maps directly to established testing concepts, arguing that observability is the prerequisite zeroth law for AI systems. The talk includes a live demo testing an AI Dungeon Master using k6 and OpenTelemetry in Grafana, plus coverage of specific AI failure modes that require testing reasoning processes, not just outcomes. Targeted at QA and testing professionals looking to apply existing skills to AI systems.