What Feng Shui was, before it became decoration

The environment is never silent. It is talking to you at every moment, in light waves, in moving air, in temperature gradients, in the low electromagnetic hum of everything plugged into a wall. Your body has been listening to that conversation for as long as you have been alive. What it has never had is a transcript.

That is the honest way to describe what Feng Shui was. Long before anyone could measure a lux level or a CO2 concentration, people noticed that some rooms made them sleep badly and some made them think clearly, and they wrote down the patterns they saw.

• Don’t sit with your back to the door.
• Don’t put the bed under the beam.
• Let the air move through, but not straight through.
• Face the light in the morning.

Strip away the centuries of decoration that accumulated on top of it — the mirrors, the coins, the ornaments people now sell as Feng Shui — and what remains is a body of empirical observation about how a space behaves. Sightlines. Airflow. Where the morning sun lands. Whether a person can see who is approaching. It was the most primitive underlying logic of space, recorded by the only instrument available at the time, which was a human being paying attention over a very long period.

The problem with a transcript nobody wrote down

Ask why Feng Shui became decoration and the answer is not mysticism. It is data loss.

An observation like “this corner feels stale” carries real information — but it carries it in a form that cannot be verified, cannot be compared against last month, and cannot be handed to whoever manages the building after you leave. So the observation gets repeated as a rule, the rule outlives the reason behind it, and eventually somebody is selling you an ornament.

This is the same failure mode we described in the O&M articles under a different name. A system whose logic exists only in one person’s head is orphaned technology the moment that person changes jobs. A building whose comfort logic exists only in the occupants’ intuition is orphaned in exactly the same way — nobody can act on it, nobody can prove it, and the next tenant starts from nothing.

Four channels, and what each one turns out to be

Put instrumentation into a space and the silent language stops being silent. Every one of the classical concerns turns out to have a measurable counterpart, and once it is measurable it becomes something a system can respond to rather than something you rearrange furniture about.

Orchestration is the part that matters

Reading the environment is the easy half. Anyone can buy sensors.

The difficulty is that these four channels are not independent, and treating them separately produces a building that fights itself. Raise the ventilation rate to bring CO2 down, and you have just changed the thermal load — so the HVAC compensates, so consumption rises, so the energy dashboard flags an anomaly that isn’t one. Drop the colour temperature in the evening for circadian reasons, and the daylight-harvesting logic that was dimming the perimeter fixtures now has a different target to hold against.

flowchart LR
  A["CO2 rises in an
occupied zone"] --> B["Ventilation rate
increases"]
  B --> C["Thermal load
changes"]
  C --> D["HVAC compensates,
consumption rises"]
  D --> E["Energy dashboard flags
an anomaly that isn't one"]

This is where “arranging the elements” stops being a metaphor. A space is a set of coupled systems, and the interesting behaviour is in the coupling.

Handling that coupling is exactly what a unified entity model is for. When lighting from KNX, sensors from Matter, chillers and AHUs from BACnet, and meters from Modbus are all mapped into the same set of entities, a single piece of logic can reason across all four at once — raise ventilation, hold the setpoint, dim the perimeter, log the cost — instead of four subsystems each optimising for its own narrow objective and quietly cancelling each other out.

And because that logic lives in a visual blueprint rather than in a person’s head, it can be read, reviewed, versioned, and rolled back. If an owner adjusts something in month three and the space starts feeling wrong again, there is a previous stable version to return to and a diff to look at. The observation “this room feels off” becomes a question with an answer.

Somebody who has read all of it

There is one more step, and it is the one that turns a monitored building into a responsive one.

A year of environmental data across a floor is not something a human reads. It is far too much of it, and the signal is in the correlations — the meeting room whose CO2 curve rises faster than its occupancy count would justify, the west face where the afternoon load and the lighting schedule quietly overlap, the zone whose comfort complaints cluster on days with a particular wind direction.

With a language model sitting on top of the historian, that year becomes something you can ask questions of in plain language. The system runs the analysis and returns a summary, and the finding that would have taken an engineer a fortnight to dig out arrives in a paragraph.

Where to go from here