How ChatGPT Dreaming V3 works (+ every other agent Memory Framework)
Summary
An in-depth analysis of ChatGPT Dreaming V3's memory architecture, explaining how it synthesizes a coherent memory state from raw sources and comparing it to other open-source memory frameworks like mem0, supermemory, and Letta.
Similar Articles
Dreaming: Better memory for a more helpful ChatGPT
OpenAI is rolling out 'Dreaming,' a new background memory synthesis system for ChatGPT that automatically curates and updates memories from chat history, addressing staleness, correctness, and scalability issues. The update is initially available to Plus and Pro users in the US, with broader rollout planned.
Memory and new controls for ChatGPT
OpenAI is rolling out memory improvements to ChatGPT, now available to free users, allowing the model to reference past conversations for more personalized responses. Users can control memory settings, including toggling saved memories and chat history on or off.
TRACE: open-source hierarchical memory for LLM agents, 82.5% on MemoryAgentBench’s EventQA using gpt-oss-20B [P]
TRACE is an open-source hierarchical memory system for LLM agents that organizes conversation history into a topic tree, achieving 82.5% F1 on MemoryAgentBench's EventQA using an open-weight model, outperforming Mem0 and MemGPT.
ChatGPT now quietly keeps a permanent dossier on everything you tell it
OpenAI has updated ChatGPT with a new memory system that creates a permanent narrative dossier of user information, improving recall accuracy from 67.9% to 82.8%, currently available to Plus and Pro subscribers in the US.
Show HN: Shared memory graph for Claude and ChatGPT, over MCP
UML (Universal Memory Layer) is an open-source shared memory graph for Claude, ChatGPT, and other AI agents, letting users save and recall structured memory across tools via MCP with a visual dashboard and API.