Everyone’s rushing to plug AI into their workflow like it’s a life jacket. Prompting their way to deliverables. Generating their way to “done.“ Mistaking speed for skill and output for craft.

AI is a multiplier, not a foundation. And multiplying zero still gives you zero.

I’ve been designing for over twenty years. I’ve coded, illustrated, built design systems at enterprise scale, and written children’s books rooted in folklore my grandparents carried across oceans. I didn’t learn taste from a tool. I earned it from failure, from studying things that had nothing to do with design, from caring deeply about work that most people would never notice.

That’s what taste actually is — the accumulation of a thousand quiet decisions made when nobody was watching. It’s knowing why something feels off before you can name it. It’s the instinct that stops you from shipping the thing that technically works but spiritually doesn’t.

---

There’s a popular idea floating around right now that goes something like this: teach AI who you are, and it will amplify your taste. Give it your preferences, your references, your criteria for what good looks like — and watch it extend your creative voice at scale.

Honestly? That’s not wrong. But it’s only half the story.

Because before you can teach AI your taste, you have to have taste. And that part — the uncomfortable, unglamorous, years-long part — doesn’t show up in anyone’s workflow diagram.

---

"But isn’t taste just pattern recognition? And isn’t that exactly what AI does?“

Fair point. Human brains are also pattern recognition machines. When you call something beautiful, you’re firing neurons shaped by years of accumulated exposure. AI trained on a trillion images does something similar — and sometimes it fools even experts. Art competitions have been won by generated work. Galleries have been deceived. If trained human eyes can’t tell the difference, does the origin of taste actually matter?

Here’s where that argument quietly falls apart: pattern recognition explains preference. It doesn’t explain conviction.

Taste without stakes isn’t taste — it’s statistics. When you push back on a stakeholder decision you know is wrong at 11pm before a deadline, that’s not pattern matching. When you scrap three hours of work because something feels spiritually off even though the client would never notice — that’s not an algorithm. That’s judgment forged from actually caring about something beyond the output.

AI has no skin in the game. Ever.

It doesn’t lose sleep over a bad decision. It doesn’t carry the weight of a community’s stories into a design choice. It doesn’t know what it feels like to ship something you’re not proud of.

How do you manage to do so much in a single day? I do everything very poorly. — Brazilian saying

That’s AI in a nutshell. Remarkable speed. Remarkable volume. And a stunning indifference to whether the work is actually right. It’s the most productive collaborator you’ve ever had — and the one least qualified to tell you if what you made is worth keeping.

---

"But taste is subjective anyway — it’s just shifting cultural consensus.“

Sure. And AI can absolutely learn and replicate consensus. It’s genuinely impressive at giving you what most people would call acceptable. But acceptable isn’t a vision. Consensus isn’t a point of view. The work that actually moves culture forward has never come from the middle of the bell curve.

And here’s the thing about that bell curve — AI will always regress toward its center. When it pushes back on your unconventional idea, it’s not evaluating your idea. It’s telling you your idea doesn’t resemble anything it was trained on. That’s not judgment. That’s pattern-matching with a confidence problem.

There’s another angle making the rounds — that taste is now a bottleneck, not a foundation. That people with high taste but low technical skill finally have leverage because AI handles execution. And sure, that’s liberating in a real way. But it quietly sidesteps the harder question: how did those high-taste people get there? Nobody wakes up with good judgment. They earned it by doing the technical work, struggling through the craft, failing publicly, and caring enough to figure out why.

Taste built without struggle is just preference with good PR.

---

I’ll be honest. I’ve been building my own AI-powered product for nearly two years now. It still hasn’t shipped. Not because the technology failed me — AI has been a remarkable collaborator. It hasn’t shipped because my own taste kept blocking me. Something would be 90% there and I’d know — the way you just know — that it wasn’t right yet. No amount of swearing at an AI agent fixes that. Trust me, I tried.

That’s the part nobody puts in their workflow post. The part where your own standards become the obstacle. Where the very thing that makes your work worth doing is also the thing that won’t let you finish.

The friction is the feature. The discomfort of “not yet” is taste doing its job. AI can generate a hundred variations overnight. It cannot tell you which one is worth shipping. Only you can do that. Not knowing the difference isn’t an AI problem — it’s homework that still needs doing.

---

AI can execute. It cannot care.

And caring — about the user, about the craft, about whether this thing actually helps a real human being on a bad Tuesday — is the whole job.

The designers who will thrive with AI aren’t the ones who use it most. They’re the ones who know exactly when not to. Who can look at generated output and say “close, but no” — and actually know why.

That discernment isn’t a feature you can enable. It’s built slowly, over years, through curiosity and humility and genuine investment in people who aren’t you.

Before you evangelize your AI workflow to the world, ask yourself honestly: what are you bringing to the table that the prompt didn’t provide?

If the answer is just the prompt — we need to talk.

AI is a remarkable tool in the hands of someone with something to say. In the hands of someone still figuring that out, it’s just a faster way to sound like everyone else.

Develop your taste. Do the work. Then let AI help you do more of it.

The slop doesn’t come from the machine. It comes from not knowing the difference.