NURA · The app

The Presence Engine: an observer of the phone, never a witness to the person

The Presence engine describes how the day actually went — roughly how long you rested, where you stopped, and whether you appear to have reached an appointment — from what the iPhone's own sensors can honestly report. Its core is a set of pure functions with no device, no network, and no clock of their own: every input is passed in, so every rule can be proved in a test rather than trusted.

One rule the whole engine exists to enforce

A phone is not a person. Everything here is an inference about a device, and the device is only sometimes a good proxy for its owner. So no function returns a bare fact like "you slept" or "you attended." Each returns a verdict with a confidence and the reason behind it, and the vocabulary is deliberately hedged — "looks like," "probably," "I can't tell." There are four verdicts: likely, possible, unlikely, and unknown — and unknown is a first-class answer, not a failure. The error this engine is written to refuse is telling someone they skipped an appointment they actually attended, or that they slept eight hours when they were awake and the phone was on the kitchen counter.

Sleep, from stillness — and why not GPS

NURA estimates rest from spans in which the phone did not move. It deliberately does not use location: indoors, GPS is accurate only to 10–50 metres and drifts even while a phone sits perfectly still, so a motionless phone appears to wander. The accelerometer answers the question actually being asked — "did this object move?" — works indoors, and costs almost no battery. A span must run at least three hours to count, and it must genuinely overlap the local night window (roughly 21:00 to 11:00, generous for late risers). The overlap is measured by walking the span in five-minute steps in local time, which is immune to the night window wrapping midnight and to daylight-saving hours appearing twice or not at all. Several hours of overnight stillness reads as likely; the same stillness in the afternoon is reported as the weaker possible, because a phone left on a desk is indistinguishable, to any sensor, from a phone beside a sleeping person. NURA never shows a confidence percentage — the number exists only to order results.

No background mode — a history query instead

NURA does not declare a background motion mode. Live motion sensing is foreground-only, because a background mode is both an App Review rejection and a battery sink. Instead, the morning after, NURA reads the motion coprocessor's roughly seven-day hardware history buffer for the night's stationary spans — so a single query recovers the whole night even though the app was suspended through it. Inside the iOS app this comes from CoreMotion via the native Presence plugin, which requires no permission prompt to read motion; a browser build falls back to the web motion event, which needs the iOS permission. Any failure returns an empty result rather than throwing — a bad night of data is never worth taking a screen down.

Attendance, without accusation

NURA can ask whether the phone went where an appointment was, from iOS geofence enter and exit crossings — never a stored location trail. It pairs each entry with the next exit and measures the overlap with the appointment's window (with a 30-minute grace either side). A solid overlap reads as likely you were there; a brief clip of the edge is the weaker possible. Crucially, an absence is never "you missed it" — it is "I never saw your phone arrive, but it may simply not have been with you." Presence of the phone is not attendance of the person, and absence of the phone is not absence of the person.

Places stay anonymous

The engine turns arrival and departure visits into a plain-language day, and it never returns coordinates. An unnamed place stays "somewhere you've been before" until you give it a name yourself; NURA does not reverse-geocode and does not guess that a place is "work." Location permission is requested in two honest stages — While-Using first, then a provisional Always — and NURA listens for iOS's later deferred prompt and for any change you make in Settings. A "where am I right now" request resolves to a place name and street, and returns nothing rather than a wrong guess when it cannot answer.

How it is used

Your presence data — the sleep estimate, place names, and visits — is stored on your device, not on our servers; there is no NuroNova backend for the core app. The optional AI reads all three systems through a single get_presence tool, and it is instructed to repeat the hedging, never to harden it.

The Presence engine is built by NuroNova Genix Co. (Delaware; headquartered in Houston, Texas). It runs on-device with no account. The free app is coming soon to iPhone.