[Écrits]Essai 04All essays

The emperor's new prototype

Why a prototype that looks finished is the most dangerous thing AI design tools produce.

Laura TrouillerJune 20265 min read

Someone sends you a prototype at 9pm. It's clickable. It's polished. It has the energy of a breakthrough, and you feel the pull to schedule it before you've worked out whether the idea underneath it is any good. That pull is the thing worth being suspicious of.

Claude Design, and tools like it, are remarkable. I'm not building an argument against them. Hand one a clear brief, real research, and a designer who knows what they're asking for, and weeks of production work collapse into an afternoon. The effort that used to live in high-fidelity screens and working prototypes has mostly evaporated. That's a genuine gain, and the people treating it as one are using these tools exactly right.

The trap is the other case. Someone opens the tool, types a brief they invented thirty seconds earlier, and gets back a prototype that looks considered. Looks like something a team deliberated over. And because it looks that way, it gets treated that way: shared with stakeholders, scheduled into sprints, turned into a conversation about implementation before anyone has asked whether the underlying idea deserves to exist.

This is the emperor's new clothes, and the tool isn't the villain. It did precisely what it was asked to do.

i.The cost used to be a filter

The old design process had an accidental defence against this. Effort was spread unevenly across the work, and that unevenness kept us honest. A sketch cost almost nothing, a few minutes and a pen, but it forced one question: what is this actually doing? Wireframes cost a little more and rewarded clearer thinking. High-fidelity screens and working prototypes cost a lot. They were slow enough that you wouldn't build one unless you were already fairly sure of what you were building.

That cost was a filter. It's now been removed, or more accurately moved. The effort no longer sits in production. It sits in thinking. And thinking doesn't show up in the file. It doesn't timestamp, you can't screenshot it, you can't drop it into a deck. So when the tool strips out the production cost, what's left on display is whatever thinking happened to exist beforehand. Solid thinking gives you leverage. Thin thinking gives you a beautiful, interactive, shareable record of a bad idea, and the two are very hard to tell apart at a glance.

Click to expand PRODUCTION EFFORT DESIGN STAGE none low mid high Sketch Wireframe Hi-fi Prototype THE GAP effort removed, thinking still required Traditional model (effort) AI-assisted model (effort) Required thinking (traditional) Required thinking (AI-assisted) EFFORT vs. REQUIRED THINKING ACROSS THE DESIGN PROCESS © Laura Trouiller, 2026
The traditional process front-loaded thinking and back-loaded effort, so the slow, expensive stages only arrived once an idea had earned them. AI flattens the effort curve to almost nothing across every stage, but the thinking a good decision requires does not move. It stays high from the first sketch to the final prototype. The gap between the two is where finished-looking work gets mistaken for finished thinking.

ii.When the artefact becomes the conversation

The problem was never poor output. It's that the output looks identical to the output of good process. The difference lives in the decisions made before the tool was opened, and none of those decisions survive into the artefact.

That matters more now because the artefact increasingly is the conversation. Stakeholders see the prototype and react to the prototype. They form opinions, move budget, and set timelines off the back of it. If it isn't tethered to evidence, those choices get made in a vacuum and dressed up as informed ones.

iii.Product thinking still requires resistance

Here's what hasn't moved. Product thinking needs friction. The question of whether something is worth building has to be asked and answered before anyone touches how it should look. Research isn't a phase that happens before design; it's the material design is made from. Skip it and you haven't designed a product. You've drawn a picture of one.

The tools aren't the problem. The belief that speed equals readiness is the problem. A prototype was always the last thing you made, not the first thing you reached for. Still true. Just easier to forget when the last thing now takes an afternoon.

← Back to Écrits Next essayThe icing was always the easy part