Everyone looks good now
When AI commoditises execution, judgment is all that's left.
When ChatGPT first came out, everything it produced felt original.
I remember that feeling.
You’d type a prompt, get something back, and think this is genuinely good. It felt like a cheat code. Content that would have taken hours, done in minutes. And it read well.
Then a few months passed. And I started noticing something.
Everything sounded the same.
My output. Other people’s output. Blog posts, landing pages, LinkedIn captions. The same cadence, the same structure, the same vaguely confident tone.
It wasn’t bad. It was just... identical.
This shouldn’t have been surprising. Everyone was using the same tools, trained on the same patterns. Of course the output converged. But it took a while to see it because the output was good enough to fool you into thinking it was yours.
I’ve watched this play out with founders too. There’s a moment, usually right after their first week with an LLM, where they enter what I can only describe as a delusional state. They’re convinced they can replace their designer. Fire half the engineering team. Move faster than ever.
Anyone who’s spent real time with these models knows what happens next. The simple stuff works. The complex stuff breaks. The AI-generated marketing copy looks clean but says nothing a competitor’s AI couldn’t also say. The code runs until it doesn’t. The design looks professional in the way that everything looks professional now, which is to say it looks like everything else.
The tools got democratized. The judgment didn’t.
I built a marketing dashboard recently using AI. The kind of thing that used to require weeks of back and forth with a data scientist. Campaign performance, attribution, the specific views I need as someone running multiple channels at once. I built it in a fraction of the time.
I’ve spent years knowing what metrics matter, what views I need at 9am vs what I need in a quarterly review, what signals are noise and what signals mean something changed. The AI built it. My experience designed it. Someone earlier in their career, using the same tool, would have gotten a dashboard. It just wouldn’t have been the right one.
The fairytale is that AI makes craft knowledge less important. Everyone gets access to professional-grade tools, so everyone becomes a professional.
The reality is the opposite. AI makes craft knowledge more important than ever. Because the tool does the execution, and execution is now free. What’s left is knowing what to ask for, why one approach works and another doesn’t, what “good” looks like before you see it.
To prompt AI to create a great video scene, you need to understand how cameras work. How lighting changes mood. How scenes are built. The prompt is only as good as your understanding of the craft behind it.
This is why studying the masters still matters. Kubrick’s symmetry creates tension because he understood exactly how framing controls what you feel. Hitchcock knew that what you don’t show is scarier than what you do. Hemingway hits different because he stripped language to the bone and trusted the reader to fill in the rest. AI can help you edit faster, generate shots, clean up prose. But if you don’t know why Kubrick’s choices worked, you’re just picking from a menu. You’re not directing anything.
The people getting distinctive output from AI are the ones who were already good at something. They’re using the tool to move faster through territory they already understand. Everyone else is generating.
And I think the trait that separates those two groups is curiosity. The ones who use AI well are the ones who can’t stop pulling things apart. How does this work? Why does that look better? What’s the difference between the version that lands and the one that doesn’t? They’re curious across a lot of fields. But they’ve gone deep in at least one.
AI can’t give you that. You have to build it the slow way.
Most AI-generated marketing copy right now is generic. People in the industry can spot the tells. The audience hasn’t caught on yet. They will. And when they do, the companies that sound like everyone else won’t just be ignored. They’ll be invisible.
Three questions worth asking about your own AI output:
Could someone in a completely different industry have produced this with the same prompt? If yes, your craft knowledge isn’t in the work.
Can you explain why you made the choices you made, or did you just accept what the tool gave you? If you can’t defend the decisions, you didn’t make them.
Would this have been better or worse if you’d done it without AI? If the answer is “about the same,” the tool isn’t adding anything. Your knowledge is.
The tools will keep getting better. The output will keep getting cleaner. And the gap between “looks professional” and “is actually good” will keep getting harder to spot from the outside.
From the inside, you’ll know. The question is whether you’re building the judgment to tell the difference, or just trusting the machine to decide for you.
Martin
P.S. This week's soundtrack is "Stressed Out" by Twenty One Pilots.
Human taste and creative judgment 💯