Developer documents AI design patterns for grand strategy game mods built in a scripting language without arrays

Three AI mods were created for Paradox games: Anbeeld's Custom AI (ACAI) for Stellaris, Anbeeld's Revision of AI (ARoAI) for Victoria 3, and AI in Imperator: Invictus for Imperator: Rome. Each mod was built using Clausewitz script, a language with only events, triggers, effects, and data structures as primitives.

The design approach distributes computation over time and uses classification before optimization, rather than attempting full search or multi-step planning. The article contrasts this with common game AI architectures like utility systems (used in The Sims) and GOAP (used in F.E.A.R. and Killzone 2), which require floating-point math and symbolic state representations that Paradox script cannot provide.

Key techniques include: creating a unified world picture across subsystems, taking over productive decisions to break mid-game stagnation, using implicit bootstrapping for cold starts, and designing graduated failures instead of cliff edges. The mods target specific failure modes in vanilla AI rather than attempting abstract optimization.