Capabilities
We design digital experiences with AI in the loop. This page is the detail: the tools we work in, what each one is for, and where our own judgment takes over.
Where a claim has a project behind it, the link is on this page.
Where we work
- Healthcare and pharma
- Policy and nonprofit
- Political campaigns
- E-commerce and consumer brands
- Real estate
The stack
Models
Claude is our main collaborator, across the Sonnet, Opus and Fable models. It drafts audits, strategy documents, copy and code, and it works inside our Figma files and our site. Gemini is our second opinion: we run recommendations past it and compare the answers, which is cross-model validation. ChatGPT is also in the kit.
Design and prototyping
Figma holds the design system, the components and the screens. Figma Make is where we compose and prototype quickly. We build component libraries with variants, so one change carries through every state, market and breakpoint.
Claude inside Figma
Figma’s MCP (Model Context Protocol) connector lets Claude read the frames and components that already exist and build new screens from them, instead of working from descriptions or screenshots. We write each project’s conventions (layout, type, components, copy rules) into a reusable skill, so new screens match the approved ones.
Code and prototypes
We prototype by describing the behavior and letting Claude draft the code, then refining from there. For client work, Claude writes the Liquid, HTML, CSS and JavaScript, scoped for the client’s developers to integrate, test and deploy.
Images and motion
Nano Banana and Midjourney for image generation. When a wireframe uses generated photography, it is labeled as placeholder imagery. For concept films we script with Claude and work with video collaborators.
Platforms and delivery
Shopify for storefronts and inventory, Squarespace for sites and gated member areas, and Optimizely for CMS-structured experience frameworks. Projects run through Asana, Jira and Miro.
How we use it
The tools change quickly. These habits are the part that stays.
Direct the work
The models draft. We set the strategy, choose which problems are worth solving and decide what ships.
Give the model the real file
A connected Figma file, a skill and a component library beat a screenshot and a description.
Compose, don’t redraw
Screens are assembled from shared components, so a change lands in every state, market and breakpoint.
Check it twice
We cross-validate recommendations against a second model, and we check every price, saving and total against the agreed documents.
Hand off clean
Code is scoped for the people who will run it, and placeholders are labeled as placeholders.
Where it shows up
The clearest examples, with the work behind each claim.