—
Let me tell you about a screensaver. Not a particularly beautiful one, but a functional one, because I was building an AI tool for design education and I needed something to show in the background while the thing processed, and so I fed the AI a prompt and it gave me an image, and the image was, to be honest, exactly what you’d expect from a machine that has consumed the entire internet and averaged it all into one unremarkable visual gesture. It was broad. It was generalized. Nobody’s specific thought was anywhere inside it. I’ve been calling this “AI slop” for a couple of years now and I know that phrase is getting some mileage in the discourse, but the first time I used it, in my classroom, I meant it precisely: not bad the way interesting things are bad, not broken in a way that tells you something, just average, just the statistical middle of everything that has ever existed on the subject, compressed into one forgettable image.
So I didn’t stop there. I ran it through a round of critique, gave it twenty-five specific improvement suggestions, ran it again, gave it twenty-five more, and kept going, and you know what happened around round eight or ten? Something shifted. Not because the AI got creative, let me be clear about that, but because my questions got specific enough to stop asking for the average and start asking for something particular. By round thirteen, the thing had actual creative judgment embedded in it, and that judgment was mine, pushed through the machine one precise question at a time. There’s a line I keep coming back to, one I’ve said in classrooms for years: the quality of your life is in direct proportion to the quality of the questions you ask. I did not think, when I first said that, that I’d be proving it on a screensaver at fifty-something years old.
Here’s the thing about that experiment, though. The lesson isn’t about AI. The lesson is about round one. Because round one is always slop, whether the tool is artificial intelligence or a junior designer or, honestly, me on a bad day. The first answer comes from the statistical average of everything you’ve absorbed, every reference you’ve seen, every safe choice that worked before. The first answer is competent. The first answer is the predictable middle. And the predictable middle, I’ve come to believe, is the most dangerous creative space you can occupy.
A meta-analysis of 106 experiments on human and AI creativity just confirmed what I’ve been watching in my classroom for twenty-five years. AI doesn’t threaten the edges. The truly original work, the deeply personal work, the work that could only come from a specific person who has lived a specific life and has something particular to say about it, that work is not going anywhere. The deeply bad work, the genuinely experimental failures that teach you something real, that’s not going anywhere either. What AI threatens, what it is already displacing, is the competent middle. The polished but empty. The technically correct but nobody’s thought. The work that looks finished before it’s begun.
I watch students do it in real time and I’ll be direct with you: it stops me cold every time. They open a tool, type a vague prompt, get a vague answer, and they submit it. Not because they’re lazy, I want to say that clearly, most of my students are not lazy, they’re just doing what the medium invites them to do. The medium says: here is a finished-looking thing, and they hear that as here is a finished thing, and those two sentences are separated by an enormous distance that nobody warned them about. So I stop them and I say, you asked AI one question and you accepted the first answer. That is not design. Design is thinking. The tools are just how you execute. But AI makes it possible, for the first time in the history of creative tools, to skip the thinking entirely and arrive at something that looks like an answer before you’ve actually asked a real question.
And this is the part that keeps me up at night, if I’m honest with you. Bad work gets caught. A truly bad design fails visibly, someone notices, you fix it or you learn from it. But mediocre work that looks polished slides through. It passes review. It ships. It fills the world with slightly more of exactly what already exists, and nobody sounds an alarm, because nothing is technically wrong. The predictable middle is invisible precisely because it looks like everything else. It hides in plain sight.
Here’s where I’ll tell you about the other experiment I’ve been running, because this one changes the frame a little. I have 650 structured creativity exercises, years of them, built from research on how creative thinking actually works, not as magic, not as inspiration, but as a trainable skill, a discipline, something you condition through practice the same way you condition anything else. And I’ve been feeding them to AI, one at a time, with accumulating context, watching whether the practice changes the output. My students over twenty-five years have already proved the hypothesis on the human side: structured creative practice, repeated, with constraints that narrow the search space away from the obvious, changes what people make. The work gets specific. It gets personal. It stops looking like the average of everything they’ve consumed and starts looking like them. So the question I’m sitting with is whether the same conditioning process works on machines, and the early evidence suggests it does, not because AI becomes creative in any meaningful sense, but because accumulated constraints narrow the search space from the statistical average to something particular, and particular is where the good work lives.
That’s what I want you to hear underneath all of this. AI is not coming for the creative edges. It is coming for the safe middle, and it should, because we shouldn’t be spending human creative energy on the safe middle anyway. The real danger isn’t replacement. The real danger is permission: permission to stay in round one, to accept the first answer, to produce something that looks finished before you’ve done the thinking that makes finishing mean anything. Every tool in history has offered that permission and the best practitioners have always refused it. The printing press didn’t kill writing, it killed the monks who were only copying, and the writers who had something to say wrote more, reached more people, changed more minds. I believe that’s the shape of this too, but I don’t think it happens automatically. I think it requires the same thing it has always required: someone who cares enough to ask the next question, and the one after that, and to keep going until the work stops looking like everything else and starts looking like the truth of a specific person thinking hard about a specific thing.
You know what that sounds like to me? Round eight. Maybe round ten. Definitely not round one.
—
This post is part of the “How Creativity Actually Works” series. If this connected with something you’re working through, the next step is simpler than you think: take whatever you’re making right now and ask one more question about it before you call it done.
—
