Pherowar – Ant Colony RTS game for developers (github.com)

Pherowar is an RTS sandbox that challenges developers to program the AI of an ant colony: design agents that explore with pheromone trails, gather and return resources, and outmaneuver rival colonies. Each competitor supplies a compiled .so library that runs inside an isolated container (built and managed with Podman), ensuring safe, reproducible, and fair head-to-head matches. The repo includes build.sh to construct the app and player wrapper image and pherowar.sh to launch matches, making it straightforward to get started. The game is significant for AI/ML practitioners because it provides a lightweight, competitive testbed for swarm intelligence, decentralized decision-making, and multi-agent coordination. Play revolves around emergent behavior driven by pheromone-based signaling, so algorithms that excel will blend local heuristics, collective coordination, and robustness to noisy, adversarial environments. Pherowar is well suited for benchmarking reinforcement learning, evolutionary strategies, distributed control, and hybrid rule-based/learned systems under realistic constraints (compiled agents, container isolation). Its emphasis on reproducibility and fairness makes it useful for research, pedagogy, and competitions focused on emergent multi-agent behaviors.