A brand system your engineers can ship
A brand that lives in a slide deck is a brand that slowly stops being true.
19 May 2026
Plenty of identity work is beautiful and completely undeployable. The logo has twelve clear-space rules and no favicon. The palette is defined in Pantone with no accessible pairings. The type is a foundry licence nobody bought a webfont for. Six months later the product looks nothing like the guidelines, because the guidelines were never buildable.
Deliver the system as code
- Colour as tokens — semantic names (surface, ink, accent) mapped to values, with the contrast pairs already checked to WCAG AA.
- Type as a scale — a defined set of sizes and weights with real webfont files and fallback stacks, not 'use the display font, big'.
- Components, not screenshots — the button, the card, the input, with their hover, focus, disabled and error states, in the framework the product actually uses.
- Motion as constants — easing curves and durations as named values, so every animation on the site feels like it came from the same place.
The identity should survive contact with the codebase
The test of a brand system isn't the reveal deck. It's what the product looks like a year after launch, when a dozen features have shipped and nobody re-opened the guidelines. If the system is tokens and components, the answer is 'still on-brand'. If it's a PDF, the answer is 'drifting'.
Because our designers and engineers are the same team, we don't hand over a brand and hope. We ship it into the build and maintain it there.
Read nextInstrument from the first commit
Bring us the problem behind the reading.