Focus Flow AI (v7.0) ships unfinished to train open-source focus AI

Focus Flow AI (v7.0) is framed as a reaction to how work is usually tracked: sitting at a screen, moving a mouse every few minutes to keep an icon green, and counting time with a timer. The project calls that pattern an illusion of focus rather than focus itself, and contrasts it with "No Fake Pomodoros" — the argument that time spent does not equal value created. Instead of watching which applications are open, it reads physical signals of concentration.

The technical route is MediaPipe telemetry, which captures gaze dynamics, subtle facial velocity and typing cadence. Privacy is handled by processing motion speed and vector numbers locally rather than video data, under a stated "No Surveillance" principle. The project positions this raw motion physics as a truer reflection of cognitive focus than screen-time logs.

The release is deliberately incomplete. The core engine exists, but the project says human deep work is too unpredictable and personal for any lab algorithm to fully capture, so it is asking creators — people writing code, text and designs — to run the telemetry and supply the data. That makes adoption the open question: whether the beta attracts enough real users for the unfinished engine to become useful, and whether creators accept a tool that watches the body rather than the screen.