The Confabulist
SubstackEssays on working with AI collaborators, covering Claude's document editing, agent security analogies and AI memory systems. The author writes from hands-on practice and applies ideas from ontology and biology to explain how agents behave.