AI agents on-call could trigger wild reliability incidents

Blog post argues AI agents should act as first responders in on-call rotations, escalating to humans only when they hit something novel. Warns that AI agents are the most complex software systems ever built, making their behavior hard to reason about.