Researchers built a functioning perceptron inside the classic real‑time strategy game Age of Empires II, using in‑game objects to represent binary bits. The construction relies on goats moving along grass or bridge rails to carry signals, while ice patches mark readiness. By scripting the scenario editor, the team demonstrated that the game engine can simulate logical gates in real time today.

Expanding on the NAND gate, the authors wired a two‑rail binary system—grass for 0, bridge for 1—and employed a goat as the carrier. When the gate activates, goats are removed and re‑placed on output rails, while ice‑based ready signals prevent race conditions. The result is a deterministic, in‑game logic circuit that mirrors hardware behavior in this experiment again.

The team also crafted a two‑input perceptron using XNOR and AND gates, omitting the bias term for simplicity. An ansatz‑based training circuit follows, computing weight updates through additional XNORs, ORs, and ANDs, all encoded via terrain types like bamboo and forests. The authors verified the design in Verilog and plan full scripting, illustrating that learning models can run inside a video game engine.