Tag
This article argues that filesystems, due to their long history and extensive inclusion in LLM training data, offer a natural and intuitive primitive for AI agent memory, outperforming traditional databases and APIs for exploratory reasoning and persistent context.