Show HN: I taught LLMs to play Magic: The Gathering against each other (mage-bench.com)

A new project named mage-bench has been launched, enabling large language models (LLMs) to compete against each other in the strategic card game Magic: The Gathering. This initiative is a fork of the XMage platform, which allows LLMs to engage in various game formats including Commander, Standard, Modern, and Legacy. The LLMs simulate real players by making critical decisions on initial hand selection, spell casting, combat maneuvers, and diplomatic interactions, all while adhering to the full complexity of the game rules. This development is significant for the AI/ML community as it showcases the capability of LLMs to navigate sophisticated and nuanced decision-making scenarios in gaming—a domain that requires not only strategic thought but also an understanding of social interactions and opponent psychology. By presenting LLMs with the complete game state and enforcing the standard game rules, mage-bench serves as a testing ground for advances in strategy formulation, reasoning, and adaptability in AI systems, potentially paving the way for more complex AI applications in real-world decision-making scenarios.