A House That Survives Its Brain: The reversibility premium of agentic private property — Swarm intelligence, physical AI, agentic architecture, reconfigurable matter, and the future of housing

Abstract

Two houses can hold the same area, the same address and the same comfort today, and hold entirely different futures. This paper builds one of them and measures the difference. A house is described as modules, relations, activity contracts and a safety constitution; a runtime reads rooms off the relations rather than from a plan; and intelligence is removed one layer at a time while the house is asked what it can still do. Across 122 modules and 124 relations, the two architectures are identical with every brain running, both carrying an activity capacity of 0.722. Removing the house brain leaves a house whose intelligence is spread through its modules at 0.676 and a house whose intelligence sits in one node at 0.000. With every brain gone, the spread house still carries 0.326, because its warmth, its air, its water and its doors arrive through a thermosiphon, a passive stack, mains pressure and a hinge. Reversibility is measured separately: a house whose joints release by hand and a house whose joints are bonded deliver the same activities today and differ by a factor of sixty-seven in a candidate reading of operational property, and by a factor of one hundred and sixty in the hours needed to take them apart. Fifty-two counterfactuals, twelve experiments, a buildable room-scale demonstrator, a runtime, an observatory page and a bill of materials travel with the paper. The composition throughout is multiplicative: a house whose structure stands, whose air moves, whose water runs and whose only exit is blocked is uninhabitable, and a weighted sum would keep that fact hidden. Author ORCID: 0009-0008-6255-7724. License: Creative Commons Attribution 4.0 International.