I hear you, but I've also been around the block enough times in my career to see the awful consequences of designing some feature, breaking it into a few Agile

I hear you, but I've also been around the block enough times in my career to see the awful consequences of designing some feature, breaking it into a few Agile "stories", coding it up, then unlocking an enormous reservoir of down stream complexity that only becomes apparent later... And no degree of making new stories or bugs can put a lid on it because there's no model underneath it -- its just an increasing amount of ad hoc code (however well-programmed: unit tested, peer reviewed, CI/CD-ed). In almost all these cases, developing a simple model grounded in some kind of common CS formalism, playing with the idea, and then explicitly constructing the system around that model would have made the application much more understandable, with a lot fewer mysteries, and reliably used by the customer in much less time. Even small systems containing simple logic, with a few input and outputs, can yield extremely complex emergent behavior (Game of Life being first obvious example).