Somebody I respect posted a version of it again last week: it blows their mind that anyone with an established creative life would even consider using AI for anything. Not just imagery. Anything. A grocery list. The replies agreed, at volume, and the shortest one was just: use your brain.

I read every reply. I always do on these. And I should be upfront about why, because it isn’t anthropology… they’re talking about me. Twenty years in, established practice, and I use this stuff more than anyone I know personally. Daily. If you drew a picture of the person that post is about, he’d have my haircut.

So believe me, I know what this looks like.

I’ve been trying to write an answer for a while now, and every draft failed the same way: it came out as a defense, and a defense wasn’t the assignment. Somewhere around the third try I figured out why nothing worked. I kept answering “the argument,” and there is no “the argument.”

There are four. They just travel in one post.

Reposted everywhere 2/2

“It blows my mind that any real creative would touch this stuff. You’re skipping the whole point of making things, you’ve traded your craft for a shortcut, it’s all built on stolen work, and you’re making yourself less human. Use your brain.”

READS AS ONE ARGUMENT
Fig. 01 One post, four arguments. They don't want the same answer.

Pull the post apart and four separate accusations fall out. You’re skipping the part that makes the work meaningful. You’ve traded craft for convenience. You’re using something built on work nobody agreed to give. And you’re making yourself less human. Two of those are about the craft. One’s about the world. One’s about me, personally. They don’t want the same answer, and honestly, my answers don’t all point the same direction.

The craft ones

The picture, if you hate AI, is a guy typing “make me a logo,” posting whatever comes out, and calling himself a designer. That guy exists. I’ve seen his work. So have you.

But here’s an actual Tuesday. I needed to send one email. The machine drafted it in about two minutes, and then I spent forty-five changing four words, because the rhythm was off and it didn’t sound like me. Reading it out loud, nudging a phrase, reading it again. Forty-five minutes, four words. Nobody made me do that. The standard did.

That ratio is the honest shape of how I work now. The generating takes minutes and the editing takes the afternoon. The covers on this journal are drawn by hand, every one. Partly for the look, sure, but mostly because drawing is where I catch what’s wrong. I’ve thrown out six finished hours because something felt off and I couldn’t name it yet.

And I should say the uncomfortable part out loud: a lot of AI work looks exactly like what those posts describe. The lazy version is everywhere, it’s visible, and I flinch at it too. I’ve shipped some of it myself, machine habits I should have caught, and every time that happens the fault is mine, not the model’s. I wrote a whole piece about hunting those down in my own work.

What people picture Timesheet
  • Type the prompt0:12
  • Post it0:03
Total 15 seconds
15 SECONDS · NO EYE INVOLVED
Fig. 02 One email, timed. The machine's share is the small number.

What actually changed is narrower than the posts assume, and I’ve made this case before, so I’ll point instead of repeating it: production got cheap, deciding didn’t. I never stopped doing the part that made me a designer. I stopped doing the typing.

Actually, let me say that more carefully, because it sounds glib. I trained as a creative director, and that job was never about producing every artifact with your own hands. You hand a rough idea to a room of art directors, they run it out in six directions, and your work is knowing which one is right. I don’t have that room anymore. It’s me at a desk. So when I pass a half-formed idea to the machine to flesh out and smooth over, I’m not doing less of the job I trained for. I’m doing the same job with the only art department I’ve got.

The theft one

This is the one where I’m supposed to have a take, and I don’t have a clean one. These systems were trained on work by people who never said yes. Mine is probably in there. Twenty years of portfolio sites, and nobody asked me either. I benefit from that every day. The consent question is real, the compensation question is real, and I’m not going to pretend I can settle either one from my desk. I’d side-eye any designer who claims they have.

So here’s the honest version. I benefit from a system I didn’t design and wouldn’t have designed. I’ve tried finishing that sentence a dozen ways that make it more comfortable, and none of them survive a reread. It doesn’t make me innocent, and I’m not going to argue it does. And the obvious counter-move, just refusing to touch the stuff… I’ve watched refusal before. It didn’t reform the phone supply chain and it didn’t get musicians paid. It’s clean for the person refusing. It’s just never once been the lever.

Maybe that’s a rationalization. Genuinely, maybe. I’ve turned it over enough times to know I can’t fully trust my own read from in here. But it’s why I’ve landed where I’ve landed: inside, with standards, instead of outside, with a post.

Because inside is where standards do anything at all. The professionals who know these tools well enough to know what they shouldn’t touch are the same ones using them every day, and that second part is what clients are actually asking for now. Not “make it with AI.” The other question: should we? Working designers get to answer that one with a straight face. No, the machine shouldn’t write the apology your customers are owed. An apology only counts if a person wrote it. No, we shouldn’t generate photos of smiling “customers” who don’t exist. That one isn’t even an AI question, it’s just lying with better production values. Those nos land because they come from fluency. The same no from someone who’s never touched the thing gets waved off as fear.

Drawn lines Rev A
  • The typingThe machine
  • First draftsThe machine
  • Production passesShared
  • The coversBy hand
  • The words under my nameMine
  • The decidingMine. Never delegated
The list moves. If it never does, I stopped paying attention.
Fig. 03 The lines, as currently drawn. Rev B expected.

Those lines aren’t a policy document. They’re just… where I’ve drawn them so far. They’ll move as I learn more. And if a client asks what the machine touched, they get the real list, every time. That’s kind of the point of standing where I can learn.

The human one

Then there’s the last argument, and I want to be fair to it, because it’s the one that stays with me after I close the app. One reply said the boring bits are what make us human. That they’re the difference between a good day and a bad one, and that people like me are quietly losing the humanity in our work and our lives.

They’re not wrong, exactly. Some of the boring bits were load-bearing. The reps I did in 2005, foam core and spray mount at 2 a.m., are where my eye came from, and I’d be lying if I said I knew what a career of skipped reps does to a 25-year-old’s eye. I don’t. Nobody does yet.

What I can do is watch for it in myself. My test hasn’t changed in twenty years: can I still catch the wrong thing without help? Which of the twenty options is off, and why, out loud? The day I can’t do that anymore, those posts stop being about some other guy with my haircut and start being true. I don’t think that day is close. I check anyway. The checking is the forty-five minutes on four words. It’s the six hours in the trash. It’s a hand-drawn cover on a piece about machines.

That test scales past me, by the way. If you work anywhere near this trade, try the blackout version. Every LLM goes dark tomorrow: ChatGPT stops answering, Claude stops Claude-ing, the Geminis all go quiet. Could you still do the work? Do you know how the house is built? Could you stand in a room and say why each decision on the page got made? If you can, the machine’s a tool on your desk, and the boring-bits argument has less on you than it thinks. If you can’t… well. Then the posts aren’t talking about somebody else.

So, yeah. From the outside this looks like a shortcut. Up close it’s the least shortcut-shaped thing I’ve done in my career, and you’ve already seen the timesheet. I don’t have answers to the big questions, and I won’t fake them for a tidy ending. What I have is a working position. Stay inside. Keep the lines where everyone can see them. Make the work good enough that the question changes, from “how could you use it” to “okay, how are YOU using it?”

That second question isn’t really about AI. It’s the one I’ve been answering my whole career, about every tool on my desk. So that one I’ll answer all day.