AI gives us speed. Quality comes from taste and judgment, and that stays with us.
Eight principles we hold to on every project, from the first interview to production code.
On every project AI does the heavy lifting: the research summary, the directions, the prototype, the code. What stays, we choose, once users have tested it. Where AI belongs in your product, we design it around the user and test it before you build.
Inputs work, errors appear, logic runs. So users react to the idea, not to the gaps, and a test tells you something about the real product.
With tools like Claude Code we develop much of the product ourselves, from interface to data and integrations, and bring in a technical partner where the work calls for it. AI does the heavy lifting; we keep it true to the design.
A model makes what is statistically common. It doesn't know your users, goals or brand. Rising above that average is design work.
AI goes straight to output. A designer first asks which problem we're solving, tests with real people and adjusts. That loop is where a product gets good.
AI sees the task you give it. We see the product around it: vision, context and priorities. And we answer for what ships.
We agree per project what may go into which tool. Nothing of yours trains a public model, and what we deliver is yours, with or without AI.
We develop our own tools in the studio, with AI. That's how we know what it does well, where it slips and what to trust it with in your project.
In every service AI does part of the work and we make the choices. Here is which part, per service.
- What we do
- We write the guide, ask the questions and listen.
- How AI helps us
- AI synthesizes the conversations in hours.
- What we do
- We walk through the product as your users do and decide what comes first.
- How AI helps us
- AI checks the screens against usability heuristics and accessibility rules and lists what it finds.
- What we do
- In the room we facilitate, so the quiet expert gets heard too.
- How AI helps us
- AI lines up what has already been said and written and drafts versions of the vision to react to.
- What we do
- The problem, the choice and what the test means stay with our designers and your team.
- How AI helps us
- AI makes the prototype on the third day.
- What we do
- We choose what fits your users and design it through to the edge cases.
- How AI helps us
- AI makes the first designs and explores more directions.
- What we do
- We decide what becomes the default and keep your brand in it.
- How AI helps us
- AI writes the first components, tokens and documentation.
- What we do
- We judge whether it still feels like your brand.
- How AI helps us
- AI tries your color and typography on real screens in many variations.
- What we do
- We keep it true to the design and decide what it has to answer.
- How AI helps us
- AI makes the prototype in little time, so people react to the product instead of a placeholder.
- What we do
- We direct it and keep it true to the design, from interface to data.
- How AI helps us
- AI writes much of the code, with tools like Claude Code.
- What we do
- We watch, listen and keep a bottleneck apart from an opinion.
- How AI helps us
- AI sets up the test script and clusters the feedback in hours.
- What we do
- How much of that came from your change and how much from, say, the season, we judge.
- How AI helps us
- AI lines up the numbers and flags what changed.
With only designers, only developers or only AI, you always miss a part. We bring design, development and AI together in one team.
Designers
- Strong in UX and UI
- An eye for detail
- Don't build it themselves
- Design clashes with tech
- Handover costs time
Developers
- Build what's specced
- Technically solid
- Code over users
- No strategy or design
- Little validation
AI
- Something on screen instantly
- Almost free
- An average result
- No coherence
- Choices without context
Strakzat
- Design and code in one team
- AI's speed, our judgment
- Tested before it's built
- No loose screens
- Choices with context
Once AI agents write the code, the work that matters moves to three places. That is where we come in, beside your team.
Deciding what is worth building
Before an agent writes a line, we set with you what users need and what it should deliver.
Context the agents can read
Your design system written down as files, so what the agents make follows your rules and your brand.
Checking what comes out
Real users test what the agents made, and once it is live we watch for the signs they get stuck.
Where AI belongs in your product, we design it around the user, with fixed rules for how it suggests and how you steer it.
AI suggests, the user decides
A suggestion can always be confirmed, changed or ignored.
You see what it rests on
With an answer we show where it comes from and how sure it is.
You know what is happening
Listening, thinking, working, done. Every state is visible.
When it does not know, it says so
Then it offers a next step instead of a guess.
Tested first
With real users, before it goes live.
What we do and do not do with AI, on every project.
A person signs off
Nothing goes live until one of us has seen it and can explain it.
AI where it helps, not everywhere
If AI does not make a product better, we leave it out.
Within the European rules
We work within the GDPR and the AI Act, and say so when an application does not fit them.
What we make is yours
Made with or without AI, the results pass to you, as in our terms.
To make our own process faster and better, and to learn where AI's limits are. So we know what to trust it with in your project.
Meridian
Shared Claude Code configuration, loaded into every project and updating itself. Our standards, inside the tool that writes the code.
Constellation
A tool for alignment diagrams such as bubble charts.
Orbit
Annotation and feedback engine shared across client prototypes, one package for many projects.
Prism
Palette generator for data visualization, built on Tailwind hues and information-design principles.
Make this the main action
Sent to Notion