Most incident chaos is missing ownership, not missing tools. ## Ship the operating contract Every new service needs an owner, a paging path, and a minimum runbook before production traffic. ## Prefer boring alerts Alert on user-facing symptoms and SLO burn. Noise trains people to ignore pages. ## Practice recovery Game days beat documentation that nobody has opened during a real outage.