A Designer's Rules, Built Into Your AI
Emil Kowalski encoded his UI design principles into a file that teaches AI coding agents to build interfaces that actually look considered.
A chef's recipe book, but for taste in software
Emil Kowalski is one of those designers-who-also-codes that other designers-who-also-code quietly admire. He's built some of the most-used UI components on the web — the little drawer that slides up on mobile, the toast notification that fades in the corner. Clean, considered stuff.
This month he released something different: a file that encodes his personal design principles so that AI coding agents can follow them automatically. Think of it like handing your AI a style guide written by someone with genuinely good taste. The agent reads it once, and from then on it builds interfaces the way Emil would — not the way a forgettable tech demo would.
It hit 15,000 stars on GitHub almost instantly. That's developers voting with their attention.
Why this matters if you're building a product
If you've been using AI tools to put together a website or app, you've probably noticed the results can look a bit... generic. Technically functional, aesthetically forgettable. This kind of skill file is the fix. Drop it into your setup and the AI stops defaulting to safe-but-dull.
You don't need to explain good taste to it every time. It already knows.
Words worth knowing
Skill file — A set of written instructions you give an AI agent that shape how it behaves. Like a briefing document it keeps on its desk.
UI (User Interface) — Everything a person sees and touches on a screen. Buttons, spacing, how a menu opens. The feel of a digital product.
GitHub stars — A rough measure of how much the developer community respects something. 15,000 is a lot.
Worth asking yourself: does your current digital product look like it was built by someone with taste, or just someone with access to AI?