Show HN: StyleSeed – a design-rules engine so AI agents stop building generic UI

AnthonyR

While I empathize a lot with the issue of agents building super generic "vibe-coded" UIs, the webpage linked from the Github provides a Showcase with presets in the style of Notion, Vercel, Linear etc. which just change the color / theme not the actual design to match that of these companies which I thought was the point of the design skills?

james2doyle

There are actually quite a few elements failing the Lighthouse accessibility audit. This page also has 26 emdashes on it. For a page that isn’t an article, that seems pretty high.

jjcm

I’d recommend a different approach if you’re looking for unique, non-LLM looking styles. Consider trying diffusion as a starting point, then feeding that into an LLM to build. Ie here are some results: https://html.non.io/tarot https://html.non.io/hydroponics https://html.non.io/solara All of these started with diffusion renders as a starting point. LLMs have intrinsic problems when it comes to design. They write code to represent design, but are trained to write consistent code. Consistent code is great if you’re writing a backend function that handles financial, but for design you end up with similar looks. You can use LLMs to police themselves, but it’s still like using backend engineers to review other backend engineers’ designs. I left Figma to go build a UI design tool based off of diffusion because of this issue. I have a full walkthrough of how I build that tarot page here: https://x.com/pwnies/status/2076755344471289898?s=46&t=bwJTI...

jjcm

This is really cool. What learnings would you say you could extract on the design side from the Watergate scandal, and how do you think StyleSeed applies to those learnings?

vivzkestrel

- LLMs are not good for design - LLMs lack so far in anything that involves actual human creativity - Sure they can build the most generic prototype of a webapp that is needed but they sure as hell wont make a unique one

_boffin_

I'm liking the idea of the project, but you're spending too many tokens on critic prompts for things that should be deterministic.