A new wave of tools promises to turn a sentence into a finished screen. Some of that promise is marketing, and some of it genuinely changes how the first draft arrives.
The question worth asking is not whether a model can produce a layout — it clearly can — but which parts of the work quietly move somewhere else when it does.
So where does the designer actually add value once the pixels are cheap to make?
The honest answer starts with knowing what you are even trying to build.
Judgement still leads
Before you accept a generated screen, ask the boring question first: what is this screen for, and what should a person be able to do the moment they see it?
A model will happily hand you twenty variations in the time it takes to read one. The scarce skill is no longer producing options; it is the taste to throw nineteen of them away, and the words to explain why the last one stays.
That editing pass is where craft hides now. The tool widens the funnel of ideas, and you are the thing that narrows it back down to something honest and usable.
In practice the loop feels less like drawing and more like giving direction.
- You set the constraints, then read what comes back against them.
- You correct in plain language and let the tool redraw — the prompt is the brief, and the brief is still the hardest part to get right.
None of this removes the designer; it just relocates them to the edges of the work.
At the front sits the brief — the framing of the problem that no generator can invent for you. At the back sits the judgement: the decision about what is good enough to ship, and what should quietly be cut before anyone else ever sees it.
It helps to treat the model as a fast, tireless junior who never gets bored redoing the same screen, and never once tells you when the underlying idea is wrong.
That gap — enthusiasm without judgement — is exactly the space a designer is paid to stand in, and it does not appear to be closing very quickly.
Prompt as a brief
A vague prompt gets you a vague screen, dressed up well enough to look finished. The clearer your intent, the more the tool feels like leverage instead of noise.
Small specifics move it a long way. Ask for one primary action, name the empty state, and the output stops guessing and starts answering the real question.
The best briefs read like constraints, not wishes. They say what must be true, then leave the tool room to be surprising inside those walls.
When a result feels wrong, the fix is usually upstream. Rather than nudging pixels, go back and sharpen the sentence that produced them; a better question almost always beats a better retouch, and it costs a fraction of the time.
This is why writing is quietly becoming a design skill. The people who get the most from these tools are rarely the fastest clickers; they are the ones who can describe an experience precisely enough that a machine, or a teammate, can act on it.
It is worth keeping a short vocabulary for the tool. Terms like dense, calm, or one screen do more work than a paragraph of adjectives, because they map to decisions rather than moods — and a model can act on a decision.
Once the brief is sharp, most of the remaining work is taste applied quickly — and taste, unlike production, has not been automated yet.
Taste at the edges
Speed is only a gift if you already know what to point it at.
A tool that produces ten screens an hour is a liability in the hands of someone with no opinion; it simply manufactures more to review, and review is the slow part.
Paired with a clear point of view, the same tool feels like a studio of one — you decide, it renders, and the distance between an idea and its artifact nearly disappears.
A 200ms opinion beats a 400ms deliberation when the stakes are a single button.
As a rule of thumb, keep the whole review loop under one coffee.
The failure mode is not that the tools are bad; it is that they make it effortless to produce plausible work, and plausible is the enemy of good. Without a strong editor at the end, the volume rises while the quality quietly flattens out.
So the discipline shifts from making more to choosing better.
Where the craft still lives
The point was never to generate for its own sake; it was to build things people actually reach for, sometimes many times a day. Occasionally that needs a fresh screen, and just as often the best move is to add nothing at all.
Knowing which is which is still a human judgement, and probably the most valuable one you have. If you want to go deeper on how these tools reshape the day-to-day of design, I wrote a longer companion piece on exactly that: