A fully automated villa in JVT, measured for a year

Six rooms held at five temperatures and no others: 23.5 occupied, 25 empty after a fifteen minute wait. The house re-states its intent every fifteen to thirty minutes, so a thermostat's factory schedule can never quietly take a room back. Override it by hand twice and it concedes for an hour.

Three dehumidifiers hold the house just under 50% around the clock; a full tank stops the unit before it overflows and says which room needs emptying. The purifier goes to full power on fine dust and boosts through the cooking hours.

The bathrooms have no switch in practice: a door contact and a presence check bring them up, a timer takes them down. The staircase drops its step lights three minutes after the last movement. The perimeter lights itself at sunset and off at ten. A bulb that loses power returns warm, not bright white.

Curtains open with the morning and close at sunset, and a heat rule closes the living room against direct sun above 34 degrees when nobody is in it. Every morning at seven the house checks the sensors this depends on and reports any that have died.

Vibration sensors on the treadmill know the moment a workout starts. The garage cooling comes on and the master bathroom water heater begins a fifteen minute cycle, so the run ends in a cool room with a hot shower waiting. Nothing to press.

Ring the bell and a snapshot lands on both phones. A vision model reads the frame and writes a one line description of who is standing there, and the visit is logged to the family calendar, so every caller leaves a record. On the wall dashboards the person at the gate appears with an unlock control beside them; one tap and they are in.

A child's room heating past 30 degrees while occupied reaches both parents' phones within fifteen minutes. Paired routines watch rooms believed empty and rooms known occupied, so the gap between them is covered. Alerts are tiered; known faults are logged, never pushed.

Access points and the router restart themselves on evidence from multiple probes, and escalate to a person only when the automatic fix did not work.

After any restart the house waits for all six thermostats, corrects any that came back wrong against the five allowed values, then checks again ten minutes later.

The electricity meter checks itself hourly and undoes false jumps caused by a restart, so the bill estimate stays honest.

When the garage unit was serviced, a fourteen day baseline was captured first. Before: the room reached its target in 2 of 268 readings. After: recovery at 3.59 degrees an hour while it was 43.8 outside, and drift of only 0.25 degrees an hour through peak sun. A service stopped being an invoice and became a measured result.

Compressor hours per cooling degree day is the villa's standing metric: it strips the weather out, so a change to ducting, ventilation or setpoints shows its real effect. The roof fan trial cut it 8.7% and the verdict is still held open, because five days a side is not yet proof.

Per-unit runtime shows which machine works hardest and which room loses its cooling fastest, which is where insulation, shading and duct corrections go first. The next dirham of upgrade spend is pointed by data, not by guesswork.

The inverter's own generation, export and self-consumption records, against the utility's bills, are what sized the array and what will size the next step, whether that is storage, shading or fabric.

Whether the next dirham goes to shading, insulation, storage or another fan bank is decided by the runtime and generation record, not by a salesman's instinct. The roof ventilation trial is exactly this loop running in public.

A service is booked when the numbers drift, and judged by whether they come back. The garage unit's before and after is what every repair in this house looks like.

Where equipment lives today is not where it was first installed. Sensors have moved, ventilation was added, dehumidifiers found the rooms that actually need them: the original drawings were a hypothesis, and the data corrected them.