Inspiration

Traditional history textbooks and static board games often fail to capture the dynamic, chaotic, and deeply strategic nature of ancient and modern warfare. We were inspired to build Tacitus AI: Historical War Simulator to bridge the gap between historical education, interactive simulation, and modern AI. By leveraging interactive maps and multi-agent LLMs, we wanted to let users not just read about history, but simulate and alter the strategic decisions of legendary commanders in conflicts like World War II or the Punic Wars between Rome and Carthage.

What it does

Tacitus AI is an interactive war simulation platform that uses map interfaces and LLM agents to model historical campaigns.

AI-Driven Commanders: LLM agents simulate generals and leaders, making tactical decisions based on historical doctrines, terrain constraints, and real-time theater conditions.
Dynamic Map Integration: Visualizes troop movements, supply lines, and frontline shifts dynamically on interactive maps.
Scenario Replay: Allows users to set up historical scenarios—such as Hannibal's invasion of Italy or key European theaters of WWII—and observe or intervene in the simulation.

How we built it

Frontend & Map Interface: Built using TypeScript and JavaScript alongside modern mapping libraries to render terrain and unit movements.
LLM Agent Orchestration: Core simulation logic is driven by advanced LLM agents to process tactical commands, evaluate risks, and update game states.
Development Workflow: Developed rapidly by structuring clear requirements upfront and using AI coding assistants (like Kiro) to iterate through physics, map rendering, and agent logic.

Challenges we ran into

Balancing historical determinism with emergent AI behavior so that simulations remain plausible yet unpredictable.
Optimizing map rendering performance while handling asynchronous agent decision loops for multiple battalions or factions.

What we learned

We gained deep insights into multi-agent coordination architectures and how to effectively translate abstract historical parameters into programmatic rules and prompt-driven behaviors.

What's next for Tacitus AI

Expand the historical database to include more campaigns (e.g., the Napoleonic Wars and the American Civil War).
Introduce a human-in-the-loop "commander mode" where you can issue direct orders to counter the AI opponent.