Design Playground / Home

Design with Agents
for Humans

Design Playground is where our product ideas come alive. Every prototype is composed in code by our agents from a living design system, shared for feedback with tools that feel like magic, and reviewed the only honest way — by playing with it. No mock-ups. No screenshots. No imagining. Just the work, working.

Browse the prototypes Open a design system
3 design systems
46 prototypes
5 collections

Counted from the explorer’s manifest, generated by walking the repository — so the number on this page is the number of folders that exist. Last generated 2026-09-18, 15:24:20 UTC.

Start here

Three concepts organise everything: the language, the work, and the collections they belong to.

Design systems

/design-systems/
The language, one per brand. A real component library — tokens, components, behaviour — shown as a living gallery. Every bound prototype builds from one.

Prototypes

/prototypes/
The work. Each prototype folder holds its page, its brief, its source and its handoff — which is what makes an approved folder a spec rather than a picture.

Collections

/collections/
Named groups of prototypes with a shared purpose — a product surface, or an idea that is not bound to any design system. Every collection is a peer; folders stay flat.

How a piece of work moves

Agents execute; humans govern. Three steps, and the same three every time.

  1. Compose from a design system

    An agent reads the gallery, then builds from the system’s source — pinned to a release, never copied and never overridden.

  2. Review at a mounted link

    The mounted URL goes into the Slack thread and the Notion embed. Iteration rebuilds the same folder, so the link never rots.

  3. Hand off from working code

    Approval freezes the folder. The handoff carries the full story to the agent that builds the real thing.