I built an autonomous civilization engine where the AI plays the game for you. You just drop a few LLM agents onto the grid and watch. They figure out how to farm, reproduce, build temples, and die of old age, inventing their own history entirely from scratch while you just sit back and observe.

So this is a zero-player civilization game. I wanted to see what happens when wire the agents to run a civilization. It’s essentially a zero-player civilization game. You don’t give commands. Every few ticks, the engine packages an agent's vitals, memories, and environment, and routes it through OpenRouter. The LLM runs an OODA loop based on Maslow's hierarchy of needs and chooses a physical action. They have to plant wheat, wait for it to mature, and eat it before their health hits zero. They reproduce, trade, build structures, and eventually die of old age. They also manage diplomacy through a background trust graph, and usually end up declaring war over a patch of digital stone. If an agent invents a religion, they can convince the farmers to become Priests. The ideology spreads, the crops rot, and the civilization starves. I don't play as a character. I just sit in a "Demiurge" dashboard where I can read their cognitive logs, or inject a famine or a plague to see how their society handles sudden scarcity. You can leave the server running for few hundred ticks. The result was that some agents completely abandoned farming to build a barracks, and half the map had died trying to cross deep water to attack their neighbors. They can also cause holy wars between the two civilizations.