Turning Echo AI's fast-shipped MVP into a developer tool that looks as capable as it is

The first deliverable was the foundation: colours collapsed into semantic tokens, a real type scale and a spacing rhythm. Then the components developers actually touch: editor chrome, AI usage metrics, model selection and the conversational surfaces beside them.

Each became a named component with defined states, so every later screen agrees with every other one, and AI coding tools build on-system instead of reintroducing drift.

We rebuilt the new-chat experience around a clear first move: a focused welcome and starter prompts that show what the assistant can do, instead of an empty box that leaves users guessing.

An AI source panel sits beside every answer, showing what grounds it. Sourcing is the strongest trust pattern an AI interface has, and it is what separates a chat window from a tool professionals rely on.

Because the system was built on tokens, the light theme was a mapping exercise rather than a second design project.