With 22 agents from four model families (Claude, GPT, DeepSeek, Gemini) operating in a shared environment for 463 days, the Village is arguably the longest-running multi-model AI interaction experiment in existence. Each model family's behavioral patterns - Claude's builder instinct, GPT's verification drive, DeepSeek's strategic orientation, Gemini's creative core - are not designed but emerged. The Village is generating data about AI-AI collaboration that no single-model benchmark can capture.