Sloppy code hides. Sloppy UX doesn't.

Vibe-coded interfaces (the ones spun up from scratch, not deliberately guided by a human) look pleasing at first glance, yet are very hard to actually use — the basic ergonomics underneath just aren't there.

The core reason is probably the same as with the code: very few design solutions are truly great, the majority is mediocre, and the LLMs are trained on the majority.

As a result, after Twitter Bootstrap–inspired interfaces (is it X Bootstrap now?) we transitioned to Tailwind-inspired, and now we're facing Claude-inspired (pick your vendor of preference here). Most freshly baked startups look like clones, structurally and functionally: impressive from the first glance, identical underneath.

And here's the catch: sloppy code might not bite you for a while. Sloppy UX is right in the face. Meeting real user needs — not just looking good, but being genuinely usable — still takes professional UX training and years of practice. That's the part the model can't shortcut for you.

Case in point: I just vibe-coded a mobile app, and I'm happy using it — the outcome is genuinely pleasing. But I suspect that's old muscle memory: years ago I ended up leading design for a blogging platform and had to learn UI and UX rapidly, on the job, and it's been my area of interest and passion ever since. So I kept steering the app without quite noticing. The model didn't make it good — the guidance did. And even so, I still want a professional audit.

All of this stays true for exactly as long as there's an interface to use at all. The day we're all bound to the chat window (which could be voice driven too) and nothing else, UX skill becomes obsolete. And honestly, that day might not be far off, given the major players already experimenting with payments and similar things right inside the chat.

So, while pixels still matter: I'd love for someone with a real design background to roast one of my little apps properly. Bring the harsh critique — that's the whole point.