R. Almeida
Model and inference
Spent six years making recommendation systems and wanted to build something that recommends less.
About
We build the quietest software we can.
Kip started because the three of us kept uninstalling the same category of app for the same reason: it wanted more of our attention than it gave back.
Position
Attention is not a metric.
Most focus tools are really measurement tools. They count your hours, chart them, and hand the chart back to you as though the problem were that you didn't know. You knew. Everyone knows.
What actually helps is much smaller: something noticing, at the moment you drift, that you drifted, and then saying so once, quietly, and never mentioning it again. That is the entire product. Everything we didn't build is deliberate.
There is no streak, because streaks punish rest. There is no leaderboard, because focus is not a competition with your colleagues. There is no daily email, because a daily email is a context switch wearing a helpful costume.
History
How we got here.
2024
A shared complaint
Three people at the same studio, all having quietly uninstalled the same four apps that month.
2025
The first model
A weekend prototype that watched application-switch frequency and nothing else. It was more useful than any of the four.
2026
Local-first, on purpose
We moved the whole model on-device, which cost us a year and every growth metric a normal company would want. It was the right call.
People
All three of us.
T. Sørensen
Interface
Believes the best interface for a focus tool is one you forget is running.
M. Kaur
Systems and privacy
Argued for on-device inference before it was convenient, and won.