Grafana Labs is proposing a familiar reliability engineering approach for one of the less familiar problems in modern software: how to measure whether an AI agent is actually behaving well. By Craig Risi

Source: InfoQ