#externalization

A reading of Externalization in LLM Agents through the lens of cognitive artifacts: agent progress is increasingly about moving memory, skills, protocols, and runtime governance outside the model.