Blog
Standortbestimmungen, Engineering-Notizen, Gedanken aus der Praxis. Geschrieben von einem Arzt, der gerade eine App baut.
Du bist das Protokoll
Niemand atmet falsch. Ein stiller Essay über Anleitungen, Protokolle und die Erlaubnis, auf deine eigene Weise zu atmen. Folge nicht dem Protokoll — du bist das Protokoll.
You Are the Protocol
No one breathes wrong. A quiet essay on instructions, protocols, and the permission to breathe your own way. Do not follow the protocol — you are the protocol.
ATMOS Interface
A transparent command layer for biofeedback. Most breathing apps hide the machine — ATMOS shows input, measurement, output, error, and reset. A monochrome BIOS-style interface for breathing and control.
Why a healthy breath is a piece of engineering we don't appreciate
A primary-care doctor on the paradox in every breath, and what intensive care medicine teaches about it. Ventilation, oxygenation, ventilator-induced lung injury, and why the quiet breath turns out to be the masterpiece.
Was wir bauen, was wir nicht bauen, und warum
Eine Standortbestimmung in fünf Ankern. Über Reizarmut, physiologische Sorgfalt, den Unterschied zwischen Messen und Anweisen, und einen Arbeitsbegriff für das, was wir machen: Stilles Biofeedback.
What we are building, what we are not, and why
A statement of position in five anchors. On low stimulus, anatomical care, the difference between measuring and instructing, and a working term for what we do: Silent Biofeedback.
From RMS thresholds to a rhythm-aware state machine
Lessons from shipping real-time breath-phase detection on mobile audio. Three engineering failures, their fixes, and where machine learning honestly fits. Plus the two-stage research framing: validate the deterministic detector first, then use its qualified labels to train personalised ML.
Beyond the timer: how biofeedback changes everything about breathwork apps
Why your heart rate tells you more about your breathing than your lungs do, and what shii·haa's biofeedback actually measures that generic timer apps can't.
From real-time biofeedback to personalised ML
Four months after shipping real-time audio biofeedback on iOS, the next step: turning qualified labels from the deterministic detector into training data for personalised on-device ML.
How we solved iOS audio analysis
The technical story behind our open-source Capacitor plugin, and why Apple's WebView broke our microphone. Now on npm as @shiihaa/capacitor-audio-analysis, MIT-licensed.