Everyone blames the model. "GPT can't design." Wrong. The model isn't the bottleneck. The pipeline is.
Here's what's actually happening when an AI tool spits out that same purple-gradient, rounded-card, centered-hero layout every single time:
1. They generate pixels, not systems. Most tools prompt a model to output a screen. One screen, in isolation. No shared token layer, no type scale, no spacing system carried across views. So the second screen doesn't know what the first one decided. Consistency is impossible by construction, not by accident.
2. They optimize for "renders without breaking," not "looks intentional." The reward signal is valid output. A layout that compiles. Taste isn't in the loss function anywhere, so the model regresses hard to the mean of its training data. The mean of all web UI is Tailwind defaults and a violet gradient. That's the slop.
Hey PH π
Quick story: we shipped five products together before this. Every time, the hardest part wasn't the code, it was making the thing look like it was designed on purpose.
Then AI design tools showed up, and they all generate the same 3-4 "vibes." Same gradient blobs, same floating pills, same fake trust badges. We got tired of it.
So we built Baroque differently. It pulls live design reference from across the internet, and enforces hundreds of design laws (real rules for typography, spacing, color, components) so every screen looks intentional, not templated.
What's new in v2:
Web agent: doesn't just generate, it references what's actually working on the web right now
Faster, sharper generation
Export to production-ready React or Figma
Free to start, no card. getbaroque.com
Honest question for everyone here: what's the one screen AI tools always get wrong for you? We'll try it live in the comments today.