Philosophy

Design starts with intention.

There is so much software and design already out there. Why would you build something from scratch? It doesn't make sense, mostly you would waste your time. There are standards like Apple's Human Interface Guidelines that you can follow to create a consistent and user-friendly experience. When I'm working on something new, I start by understanding the currently existing solutions. You could think it's wasting time, but it's actually the real time saver.

I start by rebuilding someone else’s work. I find the best version of the thing that already exists, take it apart until I understand why it works, and then build my own - again, and again, until a version turns up that’s better than the one I copied.

Taste is reconstructed and invented

I designed the Vortex token selection component. I started by visiting my favorite DeFi platforms and looking at their token selection components. Analysing which one would fit best in the Vortex platform. I chose the Squidrouter's design because it was the most consistent and user-friendly.

Squidrouter’s token selection - the reference.
Vortex - the rebuild.

I knew that our widget is definitely narrower, so I moved the chain selection to the top. You can’t read that from a screenshot; you have to rebuild it and feel where the decisions were. That’s where taste starts. It isn’t where taste ends - a teardown only ever gets you to parity with the thing you took apart.

When there’s nothing to copy

I tried the same thing with the numeric input - the one field in a DeFi app where people type an amount of money. I went through Uniswap, Aave, Curve and Balancer looking for the version worth rebuilding. There wasn’t one. Every team had written its own: Uniswap keeps around 200 lines of regex and locale handling, Aave wraps a formatting library in its own validation, Curve threads the state through a chain of hooks.

They all solve the same problems - comma or dot as the decimal separator, parseFloat quietly losing precision, mobile keyboards inserting characters nobody typed - and each team pays for them again. That’s the opposite of starting from existing work. It’s the same work, re-derived separately every time.

So I wrote Numora once: a zero-dependency TypeScript library that turns any input field into a precise numeric one. It uses type="text" instead of type="number", so you decide exactly what can be typed, pasted and shown. This time the teardown didn’t give me a design to copy. It gave me the list of problems every team ran into.

Exploded drawing of the Numora input as four stacked layers, top to bottom: text input, decimal separator, sanitizer and string value.
Numora taken apart - every layer is a problem each team was solving on its own.

Read the writing