Show HN: A Bomberman-style 1v1 game where LLMs compete in real time (github.com)

A new innovative 1v1 Bomberman-style game has been unveiled, where two Large Language Model (LLM) agents compete autonomously, providing a novel look at real-time AI interactions. Players can watch as these AI agents receive a complete description of the game state, evaluate their surroundings, and decide their next move by generating responses in JSON format. Built using FastAPI, asyncio, and vanilla JavaScript with Canvas, the game operates on a grid filled with both indestructible walls and destructible bricks, creating a dynamic battlefield where strategic bomb placement and movement are crucial for victory. This project is significant for the AI/ML community as it showcases the potential for LLMs to engage in complex decision-making within competitive environments, providing insights into AI strategy and reasoning. Each agent processes comprehensive board information, including threats and bomb timers, to formulate their moves. The game structure emphasizes not only scoring by destroying bricks and outmaneuvering the opponent but also highlights real-time reasoning through its live display of agent outputs and thought processes, thus paving the way for advancements in interactive AI applications and research into automated game-playing strategies.