uoaui.ai

Most component libraries lock you into one design system. Switching tokens, themes, or moving between React and Angular usually means rebuilding. uoaui is the tool I’m building to solve that for myself and a small group of testers: a UI Kit across five systems, plus a Builder that turns prompts into editable UIs with compare, tune, and code export.

The name "uoaui" wraps the letters U and I, a small nod to the people the tool is built for. The intent is the same. Tools that amplify human creativity rather than replace it.

I’m building uoaui solo as a side project: interface, front-end, brand, product direction, and tester feedback loops. I use AI coding agents in terminal, but the product decisions, system model, and visual direction are mine.

I worked through five phases: Empathise, Define, Ideate, Prototype, Test. A small group of designer friends tested between phases, which helped me close one question before opening the next.

The uoaui brand is built dark mode first, with deep backgrounds and vibrant aurora accents. The visual language balances expressive colour with restraint, letting content breathe.

uoaui supports five design systems: Salt, Material 3, Fluent 2, Carbon, and the uoaui system. Each uses its own tokens and component behaviours, with theme switching and code export across all of them. More details available in a private conversation.

Decision 1. Component composition with AI assistance, not generation from scratch.

Options considered: AI generates whole pages from a prompt; user composes from a fixed library; AI assists in component composition and theme decisions.

What I chose: AI as an assistant inside a component composition tool, not a generator.

Why: Pure generative AI tools tend to produce code that breaks the moment a designer wants to make a change. I wanted the user to stay in control. Pick components from real design systems, ask AI for help with composition or theming, but never lose the ability to edit any component directly.

Decision 2. Dark-mode first, not an afterthought.

Options considered: light-first with dark toggle, dark-first with light toggle, system-preference only.

What I chose: dark-mode-first with deep backgrounds and aurora accents.

Why: Design tools get used for hours at a time. Dark mode reduces eye strain in long sessions and lets the vibrant brand accents carry real meaning. Light-first would have meant the dark mode felt secondary, which is the opposite of what the user base wants.

Decision 3. Five Design Thinking phases, not lean iteration.

Options considered: lean UX with continuous iteration, Double Diamond, five-phase Design Thinking, no explicit framework.

What I chose: five phases (Empathise, Define, Ideate, Prototype, Test) with user validation at each.

Why: A solo founder can drift into premature commitment without structure. Discrete phases force me to close one question before opening the next, and real-user testing per phase prevents the system from converging on my assumptions.