Start with a few buildings.

The first ninety days

This is a proposed planning sequence, not a report of secured funding, an operating pilot, or properties available for purchase. A founding group of roughly twelve to twenty people could combine research, housing, building science, finance, care work, local knowledge, apprenticeships, and independent criticism.

1. Write a short charter covering purpose, rights, decisions, financial limits, and closure.
2. Choose one jurisdiction and locality for serious investigation.
3. Build relationships with residents, public services, local government, and existing institutions.
4. Investigate existing buildings that could be adapted without displacement.
5. Record baseline costs, environmental conditions, and participant needs.
6. Publish the uncertainties and the evidence that would change the plan.

A small feasibility budget is preferable to large deposits collected on a distant promise. Substantial recurring work should be budgeted explicitly. Participation does not make labor costless.

One site, bounded responsibilities

Twenty to forty households in existing buildings could test the core institutional proposition. A locally suitable cooperative, purpose-bound company, foundation-related entity, or another lawful structure would hold property. The choice would follow local investigation rather than precede it.

AI would initially advise on retrofit choices, maintenance, shared energy, noisy-work schedules, and understandable budgets. Housing and research participation would remain separable. The system would not rank residents by intellectual worth.

The first site needs ordinary operational competence: repairs, accessible facilities, conflict resolution, reserves, reliable records, and a clear person to contact when something goes wrong. Those responsibilities remain valuable regardless of whether the broader theory of attention proves persuasive.

An illustrative operating model

Consider twenty-four homes in a lower-cost location. All amounts below are hypothetical US-dollar planning assumptions, not verified 2026 market quotations or a financial offer.

The assumed funding mix is $2 million in grants or equivalent non-return-seeking capital, $1 million in patient equity, and $1.5 million in debt. The substantial grant component makes this a subsidized demonstration; the model does not establish ordinary commercial replicability.

Annual assumptions are a $1,100 average monthly housing charge, 95% occupancy, an additional $80,000 institutional lease, hypothetical 4% debt with thirty-year annual amortization, and a capped 3% return on patient equity. Figures are rounded.

Remove the institutional lease and the approximate $44,000 margin becomes an approximate $36,000 shortfall. Higher financing costs, vacancies, or major repairs could also invalidate the model.

Patent windfalls and speculative appreciation are not operating income. The word “community” does not make $1,100 affordable: assistance, scholarships, or cross-subsidy would need an explicit funding source. The end of a research stipend must not automatically mean the loss of a home.

What a first site could establish

• Whether the institution operates safely and fairly.
• Whether residents remain by choice and can resolve conflicts.
• Whether costs, reserves, and maintenance are sustainable.
• Whether privacy survives ordinary operational pressure.
• Whether comfort and wellbeing improve.

A small pilot cannot establish a universal causal theory of flow. Later research across several sites could examine consenting, staggered nonessential retrofits, independent review, and preregistered analysis where appropriate. Essential safety improvements should never be withheld to create an experiment.

Research should include experiences that complicate the initial theory: a resident who prefers another environment, an intervention that adds little value, a recurring cost that was underestimated. Negative results can improve the institution when they are allowed to change its plans. See measurement without intrusion.

2026–2036: expansion must be earned

Qualified AI ownership belongs near the end of this sequence, if it belongs anywhere. Unsafe operations, coercive monitoring, displacement, or unsustainable finances are reasons to pause or stop.

A model that requires silencing its skeptics before it can succeed is not ready to govern a city.