Ant Murphy's argument against epics and user stories fits the AI software moment because coding is becoming less of the scarce bottleneck.
If implementation gets cheaper, the weak link moves upstream. A team can generate more tickets, more code, and more variants, but that does not mean it knows which customer outcome matters or which assumption is false. Managing through outcomes, opportunities, assumptions, and measurable experiments is a better control loop than managing through a backlog of work packets.
For New Runtime, the point is practical: agentic coding increases output, so product leadership needs a tighter validation frame. The unit of management becomes the verified outcome, not the user story that caused code to exist.