Untitled grand strategy game
Real-time grand strategy, starting from today's world.
Built in reaction to grand strategy games that collapse into dice rolls and min-maxing. It starts from the present-day world at city-level detail and lets you play as anyone, from a commoner to a shadow figure.
- Real time, not 4X. Dynastic play extended to whole countries, with adoption and self-created titles.
- Reinforcement learning agents, not LLM agents, with shared memories and grudges and alliances passed down generations.
- Up to about 500 full agents near the player, with simpler AI running on the same parameters everywhere else.
- Single player, fully local, and no win condition. Much like real politics.
The problem with grand strategy
I have spent an unreasonable number of hours in Paradox games. Eventually every campaign turns into the same thing: dice rolls, modifiers and min-maxing a spreadsheet that happens to have a map on it.
This game starts from a different premise. The world is today's world, with current statistics down to the level of individual cities. Time runs in real time rather than in turns. And you can play as anyone: a commoner, a minister, a dynasty, or someone nobody has heard of who quietly runs everything.
The agents
The people in the world are reinforcement learning agents, not LLM agents. They share memories, and grudges and alliances are passed down through generations. Full agents run near the player, up to about 500 of them, while the rest of the world runs simpler AI on the same parameters so nothing behaves differently just because you are not looking.
The first slice
The vertical slice covers every city in South Asia, the Middle East and North Africa. The first playable moment is deliberately small: a commoner rising to power on public opinion alone.
It is single player, fully local, and there is no win condition. You stop when you are satisfied, or when the country is.
What comes next
Most of the engine is shared with Leviathan Engine, which takes the same machinery and runs government itself through millions of iterations of evolution.