Engines, not one-off pages
Systems that serve several brands from the same codebase, each with its own identity, language and deployment. Adding a company means adding a configuration, not rewriting the site.
AI product studio
We build software with AI and leave it running in production.
Product, engineering and deployment · One person to talk to · Every project measured from day one.
See the projects- Our own code
- In production, not a mockup
- Measured, not guessed
What we do
From an idea to an address that works.
Agents for the repetitive work
Routines that prepare each day's material — images, copy, links — and leave it ready. Nothing goes out to anyone else until a person has read it and signed it off.
Bilingual by contract
Both languages are validated at build time: untranslated text breaks the build. That way a gap never turns up on screen in front of a client, which is where nobody wants to find it.
Handed over running, not as a zip
Deployment, domain, certificate, security headers and browser tests against the real build. What you get is an address you can open.
Projects
What's already running.
AI phone receptionIn testing with clients
A voice agent that takes the call, knows the clinical vocabulary and books the consultation. When it hangs up it writes down who called, which treatment they asked about, whether an appointment was booked and whether they asked about pricing. And what isn't its call — a complication, a post-op question, a complaint — it flags for the medical team instead of improvising an answer.
How we work
Three rules, and the first one is to cut.
We decide what not to build
The first pass on any project is subtraction. What can't be maintained doesn't get built, and what the business doesn't need yet gets written down as pending, not as code.
We build what can be tested
Every rule that actually matters — money in whole cents, dates in UTC, destinations checked against an allowlist — has a test holding it down. An intention without a test is an intention.
We deploy and we measure
It ships to production with its domain and its headers, and from day one you can see what's happening inside. A dashboard with real numbers beats a status meeting.
Questions
What people ask before starting.
What do you do with the data of people who visit this page?
Nothing that leaves here, and you can check by opening the repository. There's no analytics and no pixels: the fonts are served from this same domain, and the site's content policy forbids asking anything of an outside server. The only thing stored is the language you picked. That's also why there's no form.
How do you decide price and timeline?
By scope, and after the first conversation. There's no published rate here and no number of weeks, because a timeline written on a public page is a contract signed before anyone knows what's being built. What does get settled before starting is what's in and what's out.
What exactly do I get at the end?
An address you can open: deployment, domain, certificate, security headers and browser tests against the real build. Not a zip file, and not a demo you have to start up on another machine.
How are the two languages handled?
They're validated at build time. Untranslated text breaks the build, so a gap never turns up on screen in front of a client, which is where nobody wants to find it. This page is served that way itself.
The numbers
How many hours go each year.
480hours a year
48 working weeks, and that's where the figure comes from. It says what that work costs you today, not what you'd save: how much of it can be automated depends on what it is, and that gets looked at before anything is promised.
Got something to build?
Tell me about it and I'll say, in one conversation and with nothing owed either way, whether it makes sense to build and what the smallest version that would already help you looks like.