Assumption Is the Mother of All Fuckups

Years ago, in my formative years as a tester, my manager said something to me that never left. “Assumption is the mother of all fuckups.” He probably said it in passing and didn’t think twice about it. I’ve thought about it more times than I can count since.

That sentence describes much of QA. Every bug I’ve found started with an assumption that turned out to be wrong. The API would return that field. The user would follow that flow. The edge case wouldn’t happen in production. Those assumptions are where things break.

AI complicates this job in a way I didn’t expect.

AI assumes too

AI hallucinations come from the same process that produces useful answers. The tool fills gaps in what you meant, the context you omitted, and your definition of “good.” Fluent writing makes those guesses look like facts.

The real problem isn’t that AI assumes. Every tool does some version of that. The real problem is we trust it too much to notice.

A harder conversation than catching a bug

Leading a team means challenging my assumptions and the ones other people bring to the work.

Someone hands me their work. It sounds confident. It looks clean. And I have to stop and ask a question that didn’t used to exist in this job: did you check this, or did the AI just say it was fine?

That’s a harder conversation than catching a bug. Catching a bug is impersonal, you found something wrong in the work, that’s just Tuesday. This is different. You’re not questioning the work anymore. You’re questioning the trust someone placed in a tool, and by extension, whether they did their part of the job or handed it off. Nobody loves being asked that. I don’t love asking it either.

Same job, different layer

QA’s job was always to verify assumptions before they became problems. That hasn’t changed. What’s changed is where the assumptions are hiding now. They used to live in a developer’s head, in a requirements doc, in a decision nobody wrote down. Now some of them live inside a tool that sounds certain about everything, all the time, whether it’s earned that certainty or not.

The question my manager gave me still holds. I’ve just had to start asking it one level up. Not “what did you assume,” but “what did you let the AI assume for you, and did you check it before you called it done.”