Autonomous Agent Game Forge

An experiment in automated game development powered by a LLM (large language model), the Autonomous Agent Game Forge is a framework that utilizes large language models, where agents with roles like programmers, QA, and designers collaborate as if they were game developers, automatically creating and debugging games from given titles.

Human input provides these titles, driving the creation process.

• Anonying block-stacking puzzle game
• Kind block-stacking puzzle game
• Best block-stacking puzzle game
• Worst block-stacking puzzle game
• Busy block-stacking puzzle game
• Lazy block-stacking puzzle game
• Classic block-stacking puzzle game
• Modern block-stacking puzzle game

All games are generated by GPT-3.5-turbo-instruct, and errors are corrected by GPT-4.