01
Routine anomaly
The action was missed.
You usually switch the kitchen light off around 19:15. The window passed, and it is still on.
Live in silent validation
Living Intelligence
Silent pilot live
It learns what usually happens.It notices when it doesn’t.
Two learning loops turn ordinary device events into a home that understands its rhythms—then explains the exceptions and asks before it acts.
07–14
days to learn
15 min
state portraits
Ask first
before action
Two kinds of truth
A habit is not the same as a state.
Most “learning homes” watch transitions and stop there. Living Intelligence keeps both: what the family usually does, and what the whole home usually looks like at that moment.
Example model
Daily habit
It learns the action.
The same device. The same action. The same part of the day. Once a routine repeats across enough distinct days, Living Intelligence can recognise the habit—and notice when it is missing.
Learns
Transitions across 7–14 days
Detects
A usual action did not happen
Missed habit
“The porch light usually turns on near sunset. Tonight, it did not.”
Example evidence
Repeated8 of 10 evenings
Usual window18:54–19:11
ContextSomeone is home
What “unusual” means
Three exceptions. Three different responses.
02
State anomaly
The portrait looks different.
At 22:30 the hall is normally dark. Tonight its state does not match the home’s learned portrait.
Portraits maturing
03
Device-health anomaly
The machine changed.
A pump runs longer than its normal range, or a sensor repeatedly stops answering. This belongs to the self-healing layer.
Next-stage intelligence
Its pattern vocabulary
A home has more rhythm than a clock.
01
Clock habits
True routines tied to a time: study light at eight, gate at seven.
02
Sun-relative
Porch and landscape lighting that follows dusk, even as seasons shift.
03
Scenes
Devices that move together become one meaningful scene—not six noisy cards.
04
Week rhythm
Weekday and weekend routines stay separate when the family lives differently.
05
Presence context
A habit can belong only to “someone home” or only to an empty house.
06
Comfort setpoints
Recurring AC temperatures, brightness levels and fan speeds are learned too.
Evidence before automation
The home offers. You decide.
Confidence alone is never permission. Every suggestion should explain what repeated, when it happened and which context mattered—then wait for an explicit choice.
Show the evidence
Support days, time window and context stay visible.
Silence is allowed
A hard suggestion cap protects the family’s attention.
Every mode is reversible
Apply, remind, dismiss—and later, undo.
Iniyal noticed a pattern
Product preview · no live action
86% confidence
Keep the evening arrival scene?
On eight recent evenings, the porch light warmed and the gate closed within the same 17-minute window while someone was home.
Seen
8 days
Window
18:54–19:11
Mode
Ask first
Truthful rollout
Learning is live. Attention is still gated.
We separate what is running, what is being validated and what comes next. A roadmap should never be dressed up as a feature already in your home.
Live now01
The engine is learning silently
Home-only collection, pattern mining, 15-minute snapshots and the watcher are running in validation. No customer notification. No automatic change.
Validating02
Precision before attention
Missed-habit and state-mismatch quality are reviewed before anything is allowed to reach a family.
Next03
Approval inside the Living Home app
Suggestions, evidence, remind-only, dismiss and explicit apply—one calm decision surface for every learned pattern.
Private by architecture. Careful by behaviour.
Runs where the home lives
Pattern data is processed on the Living Home hub, not by watching camera or microphone content.
Suggest-first
No silent automation write. Apply is an explicit family decision.
Deterministic safety path
Statistical rules handle the pattern and anomaly path without depending on an LLM.
Forget and re-learn
Life changes. Old patterns can expire instead of becoming permanent mistakes.
Questions, plainly answered
What families should know.
01
How long does a home take to learn?
Usually 7 to 14 days of clean observation. A habit must repeat across enough distinct days before it becomes a candidate. The home is allowed to conclude that there is no useful pattern and stay silent.
02
What are the two learning methods?
Method A learns repeated actions, such as turning a light off around the same time, and can flag a missed habit. Method B captures device states every 15 minutes, learns the expected portrait for each time slot, and can flag a state mismatch such as a hall light that is usually off but remains on.
03
Will it keep sending suggestions?
No. The product is designed around a small weekly attention budget, currently capped at two new suggestions. Dismissed suggestions are recorded and are never applied. Silence is a valid and preferred result when confidence is weak.
04
Can it change my home without asking?
No. Living Intelligence is suggest-first. Pattern confidence is evidence, not permission. A learned pattern can become an automation only after explicit approval through the product apply flow.
05
What is it actually watching?
Device events and device states: which switch changed, when it changed, and what controllable devices looked like at each 15-minute slot. Routine learning does not need camera footage or microphone content, and the pattern path is designed to run on the Living Home hub.
06
Is Living Intelligence available today?
The engine is live in a silent validation pilot: home-only collection, pattern mining, 15-minute snapshots and anomaly watching are running without customer notifications or automatic changes. The customer approval experience is the next rollout stage.
A home that learns carefully is a home worth trusting.
See the Living Intelligence layer inside an Onwords home—and the infrastructure that lets it keep learning after handover.
Living Intelligence is one part of the Onwords AIoT layer.