Every runbook I have ever used was out of date the day after it was written. The one for our payment service still said "restart the pods" long after two separate on-call engineers had proven that restarting the pods makes the problem worse. So I stopped writing runbooks and built one that updates itself from what actually happened. What the system does The Hindsight Incident Responder is an on-call agent for production engineering teams. When an alert arrives — PaymentService 5xx rate > ...