No-Code Builder at Product startup

The first version was dark, visually raw and clearly built under time pressure. Underneath, though, it carried a genuinely worked-out system of component states — action types, navigation, purchase and restore behaviour, all specified. The surface was not worth keeping. The logic was.

So the redesign was not a repaint. I rebuilt the editor end to end and drew the light and the dark theme at the same time, so every component was born with both appearances and engineers never had to guess at the second one. It took about three weeks.

A system starts at the colour layer, and the colour layer is not a palette. Every value in this library is a semantic token — Background, Surface, Content, Border, Action — and every token carries both a light and a dark value, so a component never has to know which theme it is standing in.

Action tokens go one step further and carry their states: initial, hover, pressed, disabled. That is the difference between a colour sheet and a system an engineer can implement without asking a follow-up question for every interaction.

Components are built on top of that layer, and each one ships with every variant and every state it will ever need in production. A button is not one rectangle: it is size, hierarchy, icon slots, loading, disabled — twice over, because light and dark were drawn together rather than ported afterwards.

The task was to build the onboarding for the No-Code Builder: the path from a first sign-up to a product that is actually set up and connected. Two moments in that path do the damage — email verification, which quietly stops people before they ever start, and the setup itself, which asks for everything at once.

That set the design principles. Make OAuth a first-class way in, so the verification letter stops being a gate. Apply progressive disclosure, so setup asks for one thing at a time instead of presenting a wall of fields. And treat connecting an app and installing the SDK as the real goal of onboarding rather than a step someone might get to later.

In parallel I took apart four competing products — Adapty, RevenueCat, AppHud and Purchasely — using one consistent method for each: sign in, sign up, dashboard. Comparing the same three moments across four products makes the differences arguable instead of decorative.

The new sign-up puts OAuth next to the email form instead of behind it, keeps the form to the fields that are actually needed, and carries its error and validation states as designed screens rather than as a note for engineering.