Adversarial Review
Evaluate something by actively trying to break it, not by checking whether it usually works
Adversarial review is a way of evaluating something by trying to make it fail. Instead of checking whether it handles normal situations correctly, you construct the situations most likely to cause a problem and see what happens.
Standard review asks: does this work? Adversarial review asks: how does this break? The two questions are not the same. Something can handle every typical case and still have a reliable failure mode that an unlucky situation will eventually find. Ordinary testing tells you about the average case. Adversarial review tells you about the worst case.
The mindset shift is to stop thinking like someone hoping the thing works, and start thinking like someone trying to beat it. What does it assume that isn’t always true? Where does it behave predictably enough to be exploited? What situation would cause it to do something it wasn’t supposed to?
The trade-off is false positives. When the task is to find problems, you will find them — including ones that aren’t serious. Every finding still needs judgment. Adversarial review surfaces candidates; a human decides which ones matter.
You’re running a fire drill at an office. The standard check confirms the alarm works and everyone knows where the exits are. The adversarial version changes the conditions: the main stairwell is blocked, it’s a Monday morning when the building is full, two people are working late and didn’t get the memo. You’re not testing whether the drill works when everything goes right. You’re testing what happens when things go wrong in the ways they actually go wrong. The failures you find that way are the ones that matter — because they’re the ones that show up when it isn’t a drill.