A Microsoft AI researcher, Adrian de Wynter, used the game *Age of Empires II* to build a neural network out of goats to critique anthropomorphism in large language models (LLMs). After peer-reviewing hundreds of computer science papers over the past two years, de Wynter found that 57% started with the assumption that LLMs have human-like traits, and 77% of papers specifically studying those traits concluded they exist. In his paper “If LLMs Have Human-Like Attributes, Then So Does Age of Empires II,” de Wynter used the game’s scenario editor to wire logic gates where a goat on grass is a 0 and a goat on a bridge is a 1. He applied the same anthropomorphic language used for AI to his goat-powered computer, describing goats as “agentic” and cooperative, though they only implement basic logic gates. De Wynter told 404 Media that “absurdism is pretty standard in philosophy and theoretical computer science,” arguing that if “agentic” goats seem silly, then words like “intention,” “cooperation” and “reasoning” deserve more scrutiny when applied to language models.