Country Simulation: one country, one year
Playable at cookwala.ai/sim/country (code: [sim/country/](../sim/country)).
Meridia is a fictional country of 27.5 million people in six provinces: a capital district, a northern port province, a farming heartland, a southern coast, low-income eastern highlands and a desert province. It has 15 cities and rural regions, of three types: urban, suburban and rural. Each province adopts robot cooks and the Cookwala protocol at its own pace, and every city and region is compared with itself without robots or the protocol: same people, same months, same seasons.
Method: three linked scales #
- Home simulator (SIMULATION.md): one kitchen, every protocol step, to find gaps in the spec.
- City simulator (CITY-SIMULATION.md): individual homes, stores, wholesale, farms, vehicles, garbage and energy for a month. Calibrated to UNEP (household waste ≈0.22 kg/person/day) and FAO (≈13% loss before retail).
- Country simulator (this one): [
calibrate.mjs](../sim/country/calibrate.mjs) runs the city simulator for urban, suburban and rural settings at 0, 25, 50, 75 and 100% robot homes, with and without the protocol (2 seeds each, 60 runs). It stores per-person daily rates in [calibration.js](../sim/country/calibration.js). The country engine applies those rates month by month:- to each region's population;
- with S-curve adoption;
- with the protocol switching on in its start month;
- with seasonal effects: winter cooking, a festival month, summer heat, the harvest glut and December holidays.
What calibration revealed about places #
| Setting | Household waste g/person-day (none → all robots + protocol) | Energy kWh/person-day | With all robots but no protocol |
|---|---|---|---|
| Urban | 223 → 121 | 1.62 → 1.55 (−4%) | 1.70 (+5%, robot electricity, no logistics gains) |
| Suburban | 223 → 121 | 1.60 → 1.37 (−14%) | 1.60 (±0%) |
| Rural | 223 → 121 | 1.81 → 1.20 (−34%) | 1.59 (−13%) |
The protocol's energy and traffic benefits are largest in rural areas, where long shopping drives are replaced by consolidated deliveries. In dense cities, the robots' own electricity offsets most of the kitchen savings.
Results: mixed national rollout (default), whole year #
| With rollout vs without | |
|---|---|
| Household food waste | −13.0% (2,088 vs 2,400 kt; 312 kt saved) |
| Food lost before homes (farms, wholesale, stores) | −9.5% (246 kt saved) |
| Garbage collected | −9.0% (322 kt) |
| Natural gas for cooking | −4.7% (17 million m³) |
| Total energy (cooking, robots, transport fuel, cold chain) | −2.5% (423 GWh), after +273 GWh of robot electricity |
| Vehicle fuel / food vehicle-km | −4.9% / −9.4% |
| Greenhouse gases | −13.9% |
| Household food spend | −3.9% (≈$1.26 billion) |
| Meals rescued from store surplus | 161 million, ≈4.7% of what Meridia's 3.1 million food-insecure people need |
| Cooking & shopping time | −20.5% (≈250 million hours freed) |
By province #
| Province | Rollout | Household waste | Lost before homes | Garbage | Natural gas | Energy | CO2e | Meals rescued (M) |
|---|---|---|---|---|---|---|---|---|
| Capital District | Incentives + protocol from January, 20 → 55% | −19.2% | −21.0% | −15.9% | −7.7% | −3.3% | −22.7% | 68.6 |
| Northshore | Protocol from March, 10 → 45% | −14.4% | −15.3% | −11.9% | −5.8% | −3.9% | −17.1% | 33.6 |
| Central Plains | Protocol from July, 3 → 30% | −8.9% | −7.9% | −6.7% | −3.6% | −3.3% | −10.3% | 24.0 |
| Southern Coast | Robots 12 → 40%, no protocol | −13.7% | +9.3% ⚠ | −1.4% | −3.0% | −0.2% | −8.2% | 0 |
| Eastern Highlands | Relief program runs the protocol from January; robots 2 → 15% | −4.6% | −10.4% | −7.4% | −2.1% | −2.3% | −6.8% | 34.7 |
| Western Desert | No adoption (control) | 0% | 0% | 0% | 0% | 0% | 0% | 0 |
Scenarios (whole year, country vs its own baseline) #
| Scenario | Household waste | Lost before homes | Garbage | Natural gas | Total energy | CO2e | Meals rescued | Time |
|---|---|---|---|---|---|---|---|---|
| Mixed national rollout | −13.0% | −9.5% | −9.0% | −4.7% | −2.5% | −13.9% | 161 M | −20.5% |
| Nationwide rollout with the protocol (60% robot homes by Dec) | −18.5% | −20.6% | −15.7% | −7.1% | −5.6% | −22.3% | 229 M | −30.0% |
| Robots everywhere, no protocol anywhere | −18.5% | +12.8% ⚠ | −1.8% | −3.4% | −0.7% | −10.6% | 0 | −30.0% |
| Protocol everywhere, very few robots | −1.9% | −8.1% | −5.9% | −0.9% | −0.6% | −3.3% | 363 M | −2.8% |
What it shows #
- Robots plus the protocol beats either alone. Robots alone help homes but push waste upstream: the national bullwhip is +12.8% more food lost before homes. The protocol alone mostly redistributes surplus. Together they cut waste at every stage.
- Where the protocol runs, every province improves upstream. Where robots sell without it (Southern Coast), the supply chain gets worse. That's an argument for making the protocol part of any national robot-cooking policy, not only the robots.
- Geography matters. Rural regions gain most in energy and traffic. Cities gain most in tonnes of waste avoided (more people).
- Seasons matter. The biggest monthly gains come in summer heat, the September–October harvest glut (demand shaping toward abundant produce) and December holidays.
- The hunger contribution is real but partial. Rescued surplus covers about 5% of the food-insecure population's needs in the mixed rollout. Rescue matters most where supply chains are still wasteful, so the "protocol, few robots" scenario rescues more because more is wasted upstream. Ending hunger needs Cookwala's relief layer together with funding, production and policy.
- Energy is a modest win nationally (−2.5% to −5.6%), larger in rural areas. Robot electricity (+273 to +389 GWh a year) has to be planned for. Natural gas falls more than total energy.
Limitations #
- Illustrative, not a forecast. Rates are interpolated linearly between calibrated adoption levels. Seasonal multipliers are assumptions.
- No robot costs, no food service or restaurants, no cross-region trade, no price feedback or rebound effects, no home fridge energy.
- Food-insecurity shares per region are illustrative inputs.
- Next steps:
- multi-seed uncertainty bands;
- a guided-human-planning home type (protocol without robots, via apps);
- region-to-region food flows;
- policy levers (subsidies, mandates) as scenario inputs.
Protocol on/off for the whole country #
The Cookwala protocol selector offers three modes:
- as planned per province;
- off everywhere (robots alone);
- on everywhere from January.
The robot adoption curves stay the same in all three. A four-way table compares them with the no-robot baseline. In the mixed rollout, food lost before homes is 2,827 kt with robots alone and 2,348 kt with the protocol as planned (−16.9%). Switching the protocol on everywhere from January cuts it further.