I'll admit it upfront: setting up proper automated UI testing was one of those tasks I kept bumping down the list. It's not fun in the way building a new feature is fun. Nobody sees it. It doesn't move the app forward, it just quietly stops it from moving backward, which is a much harder thing to feel good about at the end of a day.
I finally sat down and built out a handful of automated flows: the app launching, logging in, getting the wrong password on purpose, that sort of thing. Nothing exotic. The kind of basic coverage that should have existed months earlier, honestly.
Almost immediately, one of those boring little tests found something I'd missed while using the app: a mistake in the web address I'd configured was preventing it from reaching my server properly.
I traced it back to the app's rename, months earlier. I'd updated the address but left a small formatting mistake behind. I corrected it and ran the test again.
Here's the part that actually got to me: I'd been testing manually the whole time. The failure looked plausible enough that I hadn't checked whether the app was reaching the right place. I'd accepted what looked like an ordinary login problem instead of checking the assumption underneath it.
It's a good, slightly humbling reminder about the limits of just "using the thing to test it." You test the paths you expect. A machine, dumbly running the same four steps over and over without getting bored or making assumptions, found in about ten minutes what I'd managed to miss for months.
I'm not going to pretend I now love writing tests. But I've stopped putting it at the bottom of the list.
