Applied AI. We design the interface an AI product is judged on, and ship it as code.

We use AI where it earns its place: in what the product does, and in how our design engineers work. Neither one is a reason to bolt a chat box onto a homepage.

Surface

We shape what the model shows

Layouts that assemble around an answer instead of framing it in a chat bubble. The generated view still has to look like your product.

Behaviour

We design how it acts when unsure

Streaming, steering, retries and the low-confidence path. The states a demo skips are the ones a customer meets on day one.

System

We build what it generates from

Tokens, primitives and patterns your engineers can produce new surfaces from, so the product stays coherent without a ticket to design.

The studio. Design engineers in Stockholm with masters in AI and years of building with it, designing in the browser and committing into your repo.

Book an intro call

Designing AI

Drawn in code

Interfaces built in the browser against real responses, so timing and failure are designed rather than imagined.

Building AI

Wired to your backend

Designed in your repo, against your API. What you approve is already talking to the model and ready to merge.

Working with AI

Inside our own process

Claude and Cursor are part of how we build, on top of years of working with models directly. They keep a small team moving at the pace of a larger one.

Clients. We work with teams putting AI in front of people who rely on it.

Notes. Occasional writing on designing software in code.