Skip to content

Toaster Chef

A simulation platform and incubator for two real-world initiatives: Kitchen Garage (a nonprofit community kitchen) and siMPowerUp (a parameterized simulation of the MPowerUP peer mutual aid network).

Status: Active — Layer 0 complete, Layer 1 next Repo: (not yet created) Org: Big Nerd Idea, LLC


What It Is

Toaster Chef started as a 2D multiplayer community-kitchen game. It has been reframed as a simulation platform and incubator.

The game layer — electric trikes, toasters, a Studio Ghibli-style neighborhood — is a visualization surface. The primary deliverable is the headless analytical simulation: scenario-driven, agent-calibrated, exporting SimulationOutcome JSON that informs real design decisions for MPowerUP and a potential real Kitchen Garage.

Why this matters for MPowerUP: The Red Team Analysis established that MPowerUP's core theories (token economy, mutual aid incentives, SSI benefit interactions) cannot be validated with documentation alone. Simulation is the earliest possible test of whether those theories survive contact with realistic conditions — before real users are in the system.


Incubator Targets

Initiative What it models Why simulation first
Kitchen Garage Nonprofit community kitchen: staffing ratios, food production, trike delivery, economic sustainability Validates operating budget assumptions before committing real money
siMPowerUp MPowerUP Circle/HelpRequest lifecycle, MPWR token economy (Layer 2+), bad-actor facilitator dynamics Narrows [HYPOTHESIS] uncertainty on MPowerUP theories before pilot launch

Architecture

Visualization Layer     Phaser 3 (secondary — interactive display)
Simulation Engine       Colyseus — Kitchen Garage + siMPowerUp, 100ms tick
Analytics Layer         Per-tick outcome accumulation, SimulationOutcome export
Scenario Layer          server/scenarios/*.json — ScenarioConfig per named scenario
Agent Ecosystem         Research agents calibrate parameters with real-world data

Full architecture: docs/architecture.md (repo TBD)


Tech Stack

  • Client: Phaser 3 (browser, TypeScript)
  • Server: Colyseus 0.17.x (Node.js 22, TypeScript) — server-authoritative
  • Shared types: shared/ package — ScenarioConfig, SimulationOutcome, simulation state types
  • Map generation: rot-js, simplex-noise, WFC (server-side only)

Named Scenarios

Three baseline scenarios are defined in server/scenarios/. All have mpwr.enabled: false until Layer 2.

Scenario Purpose Key parameter
baseline.json Tier-2, moderate need — performance floor badActorFacilitatorFraction: 0.0
high-need.json Tier-3 urban, demand > helpers npcCount: 100, responseRate: 0.55
bad-actor.json 15% predatory facilitators badActorFacilitatorFraction: 0.15

Simulation Validation Status

All simulation outputs are tagged [SIMULATION OUTPUT] — a validation marker sitting between [HYPOTHESIS] and [PILOT VALIDATED]. They narrow uncertainty but do not replace field validation. Claims derived from simulation runs cannot be promoted to [PILOT VALIDATED] without real-user data.


Phase Roadmap

Layer Deliverable Status
0 Formal model: types, scenario JSON, doc updates ✅ Complete
1 Simulation tick: Circle/HelpRequest lifecycle, analytics, SimClock Next
2 Scenario system: JSON loader, REST endpoints, headless rooms Upcoming
3 Agent data collection — calibrate parameters with real-world research Ongoing
4 Headless batch simulation + SimulationOutcome export Upcoming
5 Analysis visualization Later

Sub-Agent Ecosystem

Eight research agents are defined in toasterchef/CLAUDE.md:

ToasterChef-specific: Food Systems Research, Nonprofit Operations, Community Need Mapping, Simulation Validation, Scenario Design

siMPowerUp-specific: Mutual Aid Network Research, Token Economy Scenario, Bad Actor & Risk Scenario

Cross-cutting (inherited from big-nerd-idea): Red Team, Regulatory Intelligence, Grant Intelligence


  • MPowerUP — the real-world peer mutual aid app siMPowerUp models
  • MPowerUP Token Economy — MPWR spec; Layer 2 scenario parameters calibrated from this
  • Red Team Analysis — the adversarial findings that motivated the simulation-first approach