Heist Architect
A game that uses AI pathfinding to let players plan and execute heist strategies.
About the Project
A turn-based asymmetric strategy game built to make multi-agent AI planning legible. You plan a crew on a clear 2D tactical map; the AI plans against you.
The whole point was explainable AI — you can see why the agents move the way they do.
What I Built
The planner uses A* pathfinding and Conflict-Based Search for multi-agent coordination, CSP-style temporal constraints for objective ordering, and Bayesian tracking that models guard suspicion as a probability heatmap.
Built in TypeScript, with an AI-vs-AI spectator mode and an interactive tutorial — the algorithms are far more interesting when you can watch them think.
Why It Matters
Search algorithms are usually taught as pseudocode on a whiteboard. Making them playable turns Conflict-Based Search and Bayesian inference into something you can build intuition for.
