Taste is the work now
“Building in code” is a great idea. It solves speed. It solves iteration. You can prototype in an afternoon what used to take a sprint.
But it doesn’t solve quality. It doesn’t solve taste. It doesn’t solve innovation. Those happen before the code, and outside of it.
So when AI writes most of the code and makes design dramatically faster, it’s tempting to conclude that product designers aren’t needed anymore. That’s a mistake.
What designers actually do
Designers were never just the people who pushed pixels. We’re problem solvers. We’re the ones who notice what’s off, who know why one layout feels calm and another feels cheap, who can tell the difference between an idea that’s new and one that’s merely different.
We’re curators. We collect, filter and choose. And we run all of that through the one thing that matters most: what the people using the product actually need.
AI doesn’t replace that. It makes it more important.
AI fills in the blanks
Here’s what happens when you start too soon, with too little intention.
You open a chat, type “a clean, modern landing page”, and get something that looks fine. Plausible. Familiar. Then you spend the next hour nudging it. Rounder. Less purple. No, the other font. Every round, you’re steering away from a generalized idea of what’s good, towards something you never quite defined.
That’s because AI fills in whatever you leave blank. And it fills it with the average.
The average is never what you meant. It’s what you’d have accepted if you hadn’t stopped to think.
The upfront work matters more than ever
The real leverage has moved. It’s no longer in the hours spent producing; it’s in the minutes spent deciding.
- Curate. Gather the references that shape your taste, not a mood board of everything that looks nice, but the specific things you’d actually borrow.
- Direct. Say what matters in each one. Not “I like this”, but “the radius on these buttons”, “the way these cards overlap”, “not the purple tint”.
- Communicate precisely. Point at the detail. Number it. Give the model something to hold on to, so it doesn’t have to guess.
Ten minutes of this up front saves the hour of back-and-forth after. Fewer rounds, fewer tokens, and a first draft that starts from your taste instead of the internet’s.
Don’t stop thinking
Embrace AI. Use it for everything it’s great at: speed, breadth, the hundred variations you’d never have time to draw.
But don’t hand it your judgment. The experience, the eye, the intention, the attention to detail: that’s the part that makes the work yours, and it’s the part no model can supply for you.
Don’t stop thinking for yourself. Ever.