The SQS and SNS we saw on Day 1 and Day 2 both deal with "messages I explicitly send." The producer calls the SendMessage/Publish API and decides in code where it goes. But most events that occur in a modern cloud system aren't like that. The moment a file lands in S3, the moment an EC2 instance changes state, the moment a CloudFormation stack fails to deploy, the moment AWS Health signals a region outage — these aren't "things I explicitly send" but events the AWS infrastructure generates. Add SaaS events like a Datadog alarm, a Zendesk ticket creation, an Auth0 login failure, or a GitHub PR merge, and a single company has dozens to hundreds of event sources to handle.