Source URL: https://simonwillison.net/2025/Sep/12/claude-memory/#atom-everything
Source: Simon Willison’s Weblog
Title: Comparing the memory implementations of Claude and ChatGPT
Feedly Summary: Claude Memory: A Different Philosophy
Shlok Khemani has been doing excellent work reverse-engineering LLM systems and documenting his discoveries.
Last week he wrote about ChatGPT memory. This week it’s Claude.
Claude’s memory system has two fundamental characteristics. First, it starts every conversation with a blank slate, without any preloaded user profiles or conversation history. Memory only activates when you explicitly invoke it. Second, Claude recalls by only referring to your raw conversation history. There are no AI-generated summaries or compressed profiles—just real-time searches through your actual past chats.
Claude’s memory is implemented as two new function tools that are made available for a Claude to call. I confirmed this myself with the prompt “Show me a list of tools that you have available to you, duplicating their original names and descriptions" which gave me back these:
conversation_search: Search through past user conversations to find relevant context and information
recent_chats: Retrieve recent chat conversations with customizable sort order (chronological or reverse chronological), optional pagination using ‘before’ and ‘after’ datetime filters, and project filtering
The good news here is transparency – Claude’s memory feature is implemented as visible tool calls, which means you can see exactly when and how it is accessing previous context.
This helps address my big complaint about ChatGPT memory (see I really don’t like ChatGPT’s new memory dossier back in May) – I like to understand as much as possible about what’s going into my context so I can better anticipate how it is likely to affect the model.
The OpenAI system is very different: rather than letting the model decide when to access memory via tools, OpenAI instead automatically include details of previous conversations at the start of every conversation.
Shlok’s notes on ChatGPT’s memory did include one detail that I had previously missed that I find reassuring:
Recent Conversation Content is a history of your latest conversations with ChatGPT, each timestamped with topic and selected messages. […] Interestingly, only the user’s messages are surfaced, not the assistant’s responses.
One of my big worries about memory was that it could harm my "clean slate" approach to chats: if I’m working on code and the model starts going down the wrong path (getting stuck in a bug loop for example) I’ll start a fresh chat to wipe that rotten context away. I had worried that ChatGPT memory would bring that bad context along to the next chat, but omitting the LLM responses makes that much less of a risk than I had anticipated.
Via Hacker News
Tags: ai, openai, generative-ai, chatgpt, llms, anthropic, claude, llm-tool-use
AI Summary and Description: Yes
Summary: The text discusses Claude’s memory system in contrast to ChatGPT, highlighting its unique approach to managing user conversation histories. This is particularly relevant for AI and LLM professionals focusing on user privacy and transparency in AI interactions.
Detailed Description:
The text explores the memory system of Claude, an LLM (large language model) developed by Anthropic, emphasizing its distinct architecture when compared to ChatGPT’s memory model. Key features of Claude’s memory system are analyzed, which can impact the design and implementation strategies for AI applications where user interaction and data privacy are critical.
– **Key Characteristics of Claude’s Memory System:**
– **Blank Slate Approach**: Each conversation starts without preloaded data about the user or past interactions. Memory is only triggered when directly invoked by a user.
– **Raw Conversation Access**: Unlike ChatGPT, which processes and summarizes interactions, Claude accesses only the actual conversation history in real-time. This allows users to see exactly how their past interactions are being used.
– **New Function Tools**:
– **conversation_search**: A tool for searching through past conversations to extract relevant context.
– **recent_chats**: A tool that retrieves recent conversations with sorting and filtering options.
– **Transparency and User Control**:
– Claude’s memory features are visible through tool calls, enabling users to understand when their historical data is being accessed.
– This contrasts with ChatGPT’s automated inclusion of past conversation details, which can obscure how context is applied and affect user experience.
– **User Concerns**:
– The text addresses concerns about memory systems potentially carrying bad conversation contexts over to new chats. Claude mitigates this risk by not surfacing the assistant’s responses, thereby allowing users a clearer reset when starting new interactions.
Overall, the insights from the document are crucial for AI developers and security professionals, as they highlight an innovative approach in LLM design that prioritizes user oversight and control over data, contributing positively to privacy and information security discussions within the AI landscape.