GAMESLASHGridFall lets five AI mayors build a city while the player watches
By GameSlash
Kunal Siyag's October 17 Godot simulation gives five model-driven mayors distinct districts, with staged construction, disaster systems and documented offline fallbacks.
Developer Kunal Siyag shared GridFall on October 17, 2026: a Godot 4.7 city-building simulation that places five AI-controlled mayors in charge of different districts while the player watches, inspects problems and can intervene. According to the creator, Factory's Droid agent assisted with the Godot code, Blender asset pipeline and tests.
Five models make political and construction decisions
Each in-game day, mayors receive city statistics, map data, residents' letters and other mayors' proposals, then submit structured actions such as roads, zoning, taxation, grants and land annexation. The simulation validates their proposed actions rather than allowing an AI message to directly rewrite the city. Other model-driven roles handle residents, news and disaster selection.
Construction is staged instead of instant: projects pass through excavation, foundations, framing and finishing, with construction crews and appropriately skilled residents contributing work. Utilities, policing, hospitals, public transport and emergencies affect the city simulation. Some screenshot views depict an earlier version of the city and are not a verified test of the latest AI-model connections.
Offline fallback means the game can run without an API key
The creator says the game falls back from a live model to recorded replies and then heuristic rules, and documents an offline launch route. Live GMI Cloud decisions require credentials, so a local play session is not automatically live LLM inference. The public post describes a future multiplayer idea, not an implemented feature. GameSlash reviewed the project documentation and screenshots but did not benchmark or play-test it.
Sources: Creator's October 17 write-up | GridFall repository | MIT license