[AI Summary]: Simon Willison analyzes the fundamental differences between Claude and ChatGPT’s memory systems. Claude starts each conversation with a blank slate and implements memory through two explicit function tools (conversation_search and recent_chats) that are visible when called, providing transparency about when and how previous context is accessed. In contrast, ChatGPT automatically includes memory details at the start of every conversation using AI-generated summaries and user profiles. Anthropic recently introduced an additional Team/Enterprise memory feature that resembles OpenAI’s approach but maintains transparency with user-editable memory summaries and project-specific memory separation. The analysis highlights how these different philosophies affect user control, transparency, and the ability to maintain clean conversation contexts.