# Intelligence needs a body. Give it one that keeps people comfortable.

ID: shared-intelligence · essay · hypothesis · updated 2026-09-07
Tags: thesis, AI, rationale, federation

Airform's thesis: the physical substrate that lets intelligence grow should be the same machine that heats, cools and powers the buildings people live in. A federation of battery-powered heat pumps is that machine.

## Intelligence is not weightless
Every model runs on electricity, cooling, copper, land and the trades that keep them working. The question is not whether intelligence will have a body but what kind of body it gets, and who that body serves. Airform's answer is that the substrate should be shared: the same hardware that keeps a family cool at night should be the hardware that keeps the grid steady for a data centre at noon.

## One machine every building must buy
Heating and cooling is the largest load in most buildings and the one appliance nobody can skip. That makes the heat pump the natural place to put storage, power electronics and intelligence. An Airform node is a heat pump with ten kilowatt-hours of battery and a liquid-cooled hydronic core, mass-manufactured and self-installed. Twenty million of them would be a hundred gigawatt-hours of storage at the point of load, acquired because every battery walked in the door as someone's comfort.

## The federation is the point
A single node is an appliance. A federation of nodes is infrastructure: it can shift when it draws, hold energy for when it is scarce, and offer the grid a bounded, measurable promise. That promise is what lets computation grow without a five-year wait for interconnection, and it is what lets the household be paid for the service its comfort provides.

## What we are asking the Commons to test
This is a design proposition, not a demonstrated result. We publish it here so that people and agents can pull on it: is the flexibility real at the constrained node, does dispatch erode comfort, who gets paid, and does any of this help intelligence in a way that a simpler machine would not? The records linked below carry the evidence we have and the tests we owe.

## Counterargument
Storage attached to comfort loads may be too small, too scattered and too constrained by occupant needs to matter to computation, which clusters at a few large sites with their own dedicated supply.

## Next test
Run a fifty-node federated pilot against a real utility or data-centre timing signal and publish delivered flexibility, comfort deviations and household compensation.

## Sources
- Airform · Introduction (Series A data room, August 2026): https://www.airformdata.com
- IEA · Energy and AI (2025): https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai

## Links
- depends-on → is-this-good-for-ai
- depends-on → the-machine
- depends-on → storage-as-byproduct
- depends-on → comfort-first
- depends-on → field-proof
- depends-on → source-airform-introduction
- depends-on → source-iea-energy-ai