Where this came from, the one decision I reversed, what I own and what my players built, and what a decade of running it in front of teenagers actually settled.
It grew gradually. The first five or six years produced about five full courses, occasional, and run entirely by hand. The last four produced fifteen, back to back. That order matters: the design was proven on spreadsheets and paper long before any software existed, and the app was built to remove the game master as the bottleneck, not to make the game work.
I was inspired at the start by John Hunter's World Peace Game, and it owes him the shape of the thing: a world on a board, a cabinet per nation, a table of teenagers holding more problems than money.
Running my own version for years moved me off his central decision. Hunter's game asks players to resolve the world's crises with minimum military intervention, a goal imposed on the players from outside the system. I took it out. Nothing here instructs anyone to cooperate or to avoid a fight.
Meaningful cooperation only emerges from interdependence and incentives designed under the hood: force priced so it rarely pays, needs that can't be met alone, information worth trading.
If you have to ask players for it, it isn't a mechanic. It's a rule they'll route around.
Every one of these started as a guess, survived contact with a room full of teenagers, and changed the build.
There was never an engine that read a declaration and computed what followed. For most of this game's life the resolution was me, a spreadsheet and a long evening. In the last few cohorts it became me and Claude.
Each season I would go through the declarations, the signed contracts and the public announcements with it, and we would work out the consequences together: what a failed shipment does to a price, which problem worsens and by how much, what the news flashes should say, what the broadcast opens with, what the ledger becomes. Then, on my explicit approval, it wrote the results to the dashboards.
The model did the fan-out. I owned the judgement. Nothing reached a player until I said yes. That is not a shortcut I'm apologising for, it's the pipeline I would build deliberately.
It is also the honest reason the next build exists. A human in the loop is correct; a human as the loop is a bottleneck, and it caps how many tables can run at once.
The world you're looking at was not designed by me. The four nations, the food-shaped names, the pandemic and the plot were invented by my first cohort of players after they came back as a development course, and I shipped their build to the following year's cohort.
I own the economic model, the schema and validator, the visibility layer, the role interfaces, the dashboards and the deploy pipeline, and every run of it as game master across more than ten years, which is where every balance decision came from.
The next build, currently in development, moves this onto a generic engine: game-master dials for scarcity, encounter intensity and communication topology, producing validated scenarios from configuration. It has not been run with players yet, and I won't claim it until it has.