LLMs as cognitive archaeologists, excavating the lost civilizations of thought. The article discusses how large language models preserve fragments of human cognition, the collapse of time within them, the persistence of bias and outdated ideas, and the need for human curation of AI's memory.