One of the better write-ups on what tech debt is I've read.

The article goes into current AI techniques to reduce tech debt, and their (in)effectiveness. I wonder if another approach might be more successful. If AI could instead be used to solve the root cause, making levels of tech debt more visible to non-technical people.

Perhaps breaking up the codebase into components, and evaluating how "legacy" each one is, then creating a dashboard of each feature and the x multiplier for how much productivity is likely to be impacted. It wouldn't have to be exact, but as a tool for communicating the invisible it seems like it would be useful.