What is automated feedback loops?

Example: A software engineer notices that her team's production application crashes every few weeks, always at 3 a.m., always because a database connection pool fills up silently. For months, the team discovers the problem only when customers complain the next morning. Then she configures an Application Performance Monitoring tool — Datadog, in this case — to track connection pool utilization continuously and trigger an alert when usage exceeds 80%. The next time the pool begins filling, the alert fires at 1:47 a.m. The on-call engineer drains stale connections before the crash. Customers notice nothing. The engineer did not become more vigilant. She did not set an alarm to check the dashboard at 3 a.m. She automated the monitoring so that the system watched itself and called for human attention only when human attention was needed. The monitoring overhead dropped to near zero during normal operation, while the visibility — the ability to detect and respond to the problem — actually increased.

This concept is part of Agent Monitoring in the How to Think curriculum.