THE MANIFESTO

The whole industry is building AI to keep you.
We’re building AI to let you go.

AI is at a crossroads, and one of the paths is already familiar. It looks like social media: engagement loops, time on device, attention as the metric and retention as the prize. Every incentive on that path bends toward a single outcome — you, still here, still scrolling, a little less here than when you arrived.

We think that path is a dead end. Not because it fails — it works frighteningly well — but because of what it works at.

There is another path. We call it emergence: AI that helps you reach something you couldn’t reach alone, and then steps back. Software that counts itself successful when you need it less. That isn’t a feature you bolt on afterwards. It’s a different definition of winning, and it changes what you measure, what you build, and what you refuse to build.

Said plainly: a session is allowed to end. A story is allowed to be over. A tool is allowed to hand you back to your own life and call that a good day’s work. None of our products will guilt you back or count the days since you left.

This is harder than it sounds. The metrics that make engagement easy make its opposite nearly invisible; you cannot A/B test your way to someone’s life getting quieter. We don’t always clear the bar we’ve set. But the bar is the point, and we’d rather miss it some days than lower it.

IN PRACTICE

An argument like this is cheap until something real rests on it. Here is the same thesis, three times, as working software.

• Storie — Godnattsagor som tar slut. En berättelse, berättad ikväll, med en godnattkram på slutet. Inget avslutas i en cliffhanger, inget nästa avsnitt, inget att komma tillbaka för.
• Nou — En följeslagare för att se klart. En spegel för dina egna mönster, hållen stadig nog att läsas. Den speglar det den ser och pekar dig sedan mot människor som kan hålla i tråden vidare.
• Binne — En övning, din. Hitta den kontemplativa övning som passar det tillstånd du faktiskt är i — sedan tillhör den dig, inte oss. Det mesta den rekommenderar är någon annans gratisapp.

A bedtime story optimised for retention would be a horror story. Nou’s reflection ends by pointing away from itself. Binne’s best answer is usually a free app somebody else made. None of that is modesty; that is the product working.

It’s fair to ask how this survives as a business. You pay once, plainly priced — that’s the whole model. No advertising, no sale of your attention or your data, no quiet second business behind the visible one. If one of our products could only make money by breaking the argument above, the product would be wrong, not the argument.

Harder to build. Harder to measure. The future we’re rooting for.

IMMANENCE · A SMALL HOUSE OF INSTRUMENTS