Bad Ideas Don’t Give You Permission. They Give You a Map.

Bad Ideas Don’t Give You Permission. They Give You a Map.

I give students an exercise and I say “design the worst possible version of this, make it deliberately, committedly terrible, put real intention into the awfulness,” and the thing that always happens is that arriving students go into total paralysis because they’ve never been given permission to fail this explicitly, they’ve been trained for 20 years to reach for the acceptable and the right, and the moment you tell them to reach for the wrong on purpose something shuts down. But there’s always a moment in the middle of that paralysis where something shifts, where they realize that the worst possible version actually requires more thinking than the best possible version because the best possible version has a template you can follow but the worst possible version has to be specifically, committedly bad, and once they start generating they can’t stop and what emerges is something that reveals what they actually think design shouldn’t do.

The exercise is diagnostic; it’s showing you the territory where you’re afraid to go, and if you can map that territory by committing to it fully, by making something that is deliberately wrong according to your own standards, then you know where your limits are and you can start to question whether those limits are actually there for a reason or whether they’re just inherited from someone else’s idea of what’s correct. You can’t question what you won’t enter, and if you’re afraid of the bad idea, you’ll never understand what makes it bad, so you’ll never understand what makes something good; you’ll just know that you’re supposed to follow the rules but not why the rules exist.

When I gave the worst-possible-version exercise to an AI system one submission at a time, early responses were interesting in the way they were uninteresting; the AI produced bad design by producing generic design, by taking the safe choices and deploying them incorrectly, by making something that was technically wrong in ways that felt theoretical. I said “that’s not bad enough; commit to it; make something that’s actively wrong, not accidentally wrong,” and I kept iterating and something started to happen where the AI understood that the exercise was asking for specificity and commitment, not just deviation from the template.

As the sessions accumulated context and exercises built on each other, what came back started getting stranger and more committed and less predictable; the AI started producing bad design that was specific and intentional and genuinely violating in its refusal of convention. And that’s when I realized what the exercise actually reveals: it shows the difference between deviation that’s random and deviation that’s committed, between something that’s wrong and something that’s wrong in an interesting way, between a violation that’s accidental and a violation that’s deliberate. At that point you can work with the bad idea because you understand the mechanics of it.

Every project that broke loose for me started with a version that would have embarrassed me to show anyone, and I don’t mean version 2 or version 3, I mean version 1, the first attempt at the thing, the proof of concept that was so crude and obvious about what it was trying to do that it seemed impossible it could ever become something real. The first draft of this blog pipeline was a mess, all the thinking was visible and unrefined and the process was showing everywhere, and it was perfect for what it needed to be which was a map of how the thinking actually worked before it got polished. The first iteration of the curriculum was overcomplicated and unfocused and it tried to do everything at once before I understood what it was actually trying to do. The first attempt at the creativity experiment had no mechanism for capturing what was happening, it was just me running exercises and noting vaguely that something was working.

Every one of those failures was a map, and the map showed me exactly where I needed to simplify, where I was overcomplicating things, where the real insight was hiding underneath all the theory, what was actually worth keeping and what was just noise. If I’d waited to start until I could do it perfectly, I’d still be waiting, and the thing I’d be waiting to do would have died while I was waiting. The bad idea is only bad in retrospect; in the moment it’s the only idea you have access to, and the only way to get to the good idea is to make the bad idea and learn what it’s trying to teach you.

This is why I tell designers and students and anyone making something: commit to the bad idea completely, give it real intention, make something that’s deliberately wrong according to your own standards, because that commitment is what reveals the standards and that revelation is what lets you work with them instead of just being controlled by them. The bad idea doesn’t give you permission to make bad things; it gives you a map of what you actually believe about how things should work, and once you have that map you can navigate the territory with intention.

What I’ve learned from the exercises and from submitting them to an AI system and from watching what emerges is that the committed bad idea is more valuable than the uncommitted good idea, because the committed bad idea is built on thinking and the uncommitted good idea is just following the template. If you can make something that’s bad in a committed way, you’ve made something that has structure and intention and understanding underneath it, and understanding is the thing that lets you iterate toward something that works. The good thing without understanding is just beautiful mistake; the bad thing with understanding is a tool that sharpens your thinking every time you use it.