"Look! Every test is green! The API returns 200 OK!"

"Relax. The system works perfectly. If the user is just standing there staring at the screen, that's a user problem."

I was two years into my first engineering job, and I had quietly decided my senpais were hopeless. They lived inside dashboards, barely touched the actual product, and got cheerfully drunk on coverage numbers. Their one redeeming quality was that the drunker they got on "the code works," the more pleasant they became.

But code working and a real human getting what they came for are two completely different things. A button can return 200 OK and still leave a person staring at an unchanged screen until they give up and leave.

So one afternoon, instead of arguing, I opened a terminal and built a ~30-line shell spell that finds every UX dead-end without running the app even once. I call it the two-agent static walkthrough.