Inspect the assumptions.

Time and geography

56 days, hourly snapshots, six fictional adults and three schematic settlements. Travel is one hour between Harbor and Mill and two hours on routes involving Ridge. These are not actual places or travel estimates.

Ordinary life

Everyone has weekday work from 09:00 to 13:00 at 8 illustrative currency units (CU) per hour. Household expenses are 24 CU/day. Additional work uses available afternoon hours and reserves time to return home before care or personal commitments. Existing work and household spending are simplified; employment, food and health outcomes are not individually modeled.

Starting resources

Each person has 400 CU. Harbor has 1,800 CU; Mill and Ridge have 1,000 each. The outside economy begins with 100,000 CU and is finite. Ridge has 30 parts, with no automatic restocking. Each settlement starts with 400 litres of stored water and has a 600-litre tank.

Water and guests

Working pumps supply 240 L/day. Harbor requires 220 L/day; each other settlement requires 160 L/day. Supply, use and overflow are recorded at 20:00. Water quality and engineering are not assessed. Harbor hosts two guest nights daily when its modeled water demand is met, at 35 CU/night. Ordinary operating costs are 45 CU/day. Unpaid costs accumulate as a backlog and block new expansion until funded. A completed pilot adds two guest nights and 20 CU/day of upkeep. Demand and prices are assumed, not bookings or forecasts.

Failures and procurement

The seed selects three pump-failure dates around days 9, 19 and 32. All modes face the same stress dates; an already-offline pump keeps its existing repair task. A repair requires one part and five worker hours. Before ordering, a repair also requires 140 CU of labor/inspection reserve; the pilot requires 1,200 CU, including a contingency. This check does not lock those funds away from other spending. Parts cost 40 CU each and take 24–48 hours of supplier processing plus route travel. A supply disruption adds 48 hours to future orders.

Starting workflows

On day 8, the SOS mode proposes 12 hours of setup at 10 CU/hour; a conventional register requires six hours at the same rate. Both reduce Harbor’s assumed dispatch delay from 24 to four hours after completion. The existing-arrangements mode has no setup. These dispatch effects are hypotheses. Payment is booked when an hourly work slot is reserved; completed work is credited at the end of that hour. No model-quality bonus is granted to SOS.

Participation and learning

Three people agree to a bounded SOS trial. Accepted contributions can lead to continued participation; no accepted contribution by day 29 ends a trial. If Leela leaves, Harbor returns to its existing dispatch workflow. Asha, Dev and Nadi retain their outside arrangements. Mill can adopt the shared record after an accepted repair involving a participating Arun. Training requires 12 joint hours with Arun and Mira at Mill, commissioned with 360 CU from Harbor; Mill pays 15 CU/hour to each. This grants only a modeled maintenance capability, not a professional credential.

Inspection and pilot

Repair and retrofit work requires two separately paid hours from Nadi at 20 CU/hour after an inspection arrangement is assumed. She cannot approve work she performed. On day 15 a two-room retrofit is proposed in all modes. It needs hypothetical site review, an inspection arrangement, one accepted Harbor repair, 12 parts and 32 work hours at 20 CU/hour. This is neither the full Tusita resort nor a land purchase or validated construction plan.

Interventions and comparison

Inputs apply at the next selected hour and preserve earlier events. Future scheduled inputs stay in the run unless removed. Changing the starting mode or seed recalculates the entire experiment. The comparison uses the same seed and scheduled inputs with a conventional register. Inputs offered in one run may have different consequences in another.

Graphics contract

The pure engine returns JSON-compatible snapshots, stable entity IDs, activities, destinations, arrival times, events and every successful payment. The map consumes these records. A future 3D renderer can animate the same events, but detailed geometry, collision avoidance, animation and richer individual behavior remain to build.

What must still be learned

Candidate sites, professional design, actual supplier and labor costs, demand, household circumstances, skill assessment, governance and data permissions need real investigation. Energy, food production, ecological limits, institutional disputes and broader migration are future model work. Graphical realism will not remove these uncertainties.

The earlier World Observatory and Tusita design study remain separate experiments with different assumptions. Their numbers are not inputs or validation for this model.