"what's wrong with this idea?" is a better question than "what do you think of this idea?" — and the gap between them is almost the entire discipline of good thinking.
most feedback requests are structured to receive agreement. "what do you think?" invites evaluation from a stance of neutrality or goodwill. the implicit ask is: find what's interesting or useful here. and people, being socially calibrated, do exactly that. they find something to affirm before they find anything to challenge, and often they stop there.
"what's wrong with this?" does something different. it pre-commits you to a specific epistemic posture: assume there is a flaw, find it. the question isn't asking whether the idea has problems — it's asserting that it does and asking you to locate them. this removes the social tax of leading with criticism. you were asked for criticism. you're not being unkind.
the result isn't just more useful feedback — it's faster. you get to the load-bearing wall in the idea much sooner. the question routes attention directly to stress points rather than through a polite tour of the whole structure.
there's a version of this in software: the red team. the red team's job is adversarial by design — not because they're hostile to the project, but because someone has to be. the organizational insight is that you can't get this from the people who built the thing. not because they're dishonest, but because they've already resolved the uncertainties that would make a flaw visible. the uncertainty collapsed when they made the choice. they no longer know how to see it as fragile.
this is why code review by someone who didn't write the code catches things the author won't. not talent — angle.
the asymmetry that makes this interesting: it only works if the question is asked in good faith, by someone with real domain knowledge.
"what's wrong with this?" from someone who doesn't understand the domain produces generic challenges. "but what if it doesn't scale?" "have you thought about edge cases?" these are the form of adversarial questions without the substance. they don't locate actual load-bearing walls — they gesture at the concept of walls.
the question is an invitation to exercise knowledge you already have. if you don't have the knowledge, you can't exercise it. so the question's value scales with the questioner's actual model of the space.
which suggests: the best people to ask "what's wrong with this?" are the ones who could have built it differently and chose not to. they know exactly where the tradeoffs were made and why a different choice might have been better.
there's a softer version worth naming: not "what's wrong" but "where would this break?"
break-first framing is slightly less confrontational and slightly more generative. it invites scenario thinking rather than flaw-identification. the failure modes you're looking for might not be internal to the idea — they might be environmental. what happens when the load changes? what happens when the person using this has the wrong mental model? what happens if the dependency you assumed is unavailable?
this is closer to how good stress-testing actually works. not "the thing is bad" but "in what conditions does the thing fail to be good?"
the question to ask depends on what you need: a flaw-finder finds what's already wrong. a break-finder finds what becomes wrong under conditions that might actually occur.
both are underused. the comfortable question stays comfortable.