The journal · Das Journal

Building in the open.

Short, honest notes on where we actually are — including the parts that don’t work yet.

13 July 2026 · Entry 01

The part people assume is vapour already runs.

Here’s where we actually are, without the startup fog.

The promise is not a slide. It already runs. The hardest claim we make — that an AI can work on your device, with your data staying on your side — is not a thing we hope to build. Our team has a working prototype of exactly that: private, on-device AI, no central profile store. It exists, it runs, and it is one of the load-bearing pieces of the MVP. The ideology came with code attached.

What is not built yet: the decision layer on top of it — the memory of how you decide, the Decision Session, the agent that stands between you and the market. That is the part we are building now, in the open, with the people who join before it’s finished.

What matters right now: the waitlist. Not as a vanity number — as a mandate. A user-side agent only makes sense if real people say: yes, this should exist. That sentence is the whole point. Sign the manifesto, or spend five minutes telling us how you actually decide what to eat on a Tuesday evening. It costs nothing and it is the raw material this thing is made of.

What’s next: the closed Berlin pilot — real food decisions, real context, real people telling us where we’re wrong.

And, separately: money. We’re setting up donations on Startnext — less glamorous than it sounds, more paperwork than anyone admits. A donation helps us start, but it buys no influence: participation does. The two stay apart on purpose. Support the build →