Dreaming: Better memory for a more helpful ChatGPT

2026-06-03 · Source: OpenAI News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, extended

Summary

OpenAI announced on June 4, 2026, a significant update to ChatGPT's memory system, introducing a more capable and compute-efficient architecture built on "dreaming." This new system, Dreaming V3, addresses challenges of staleness, correctness, and scalability observed with previous memory features, including the "saved memories" launched in April 2024 and an earlier "dreaming" version from April 2025. Designed for hundreds of millions of users and multi-year horizons, the update optimizes memory synthesis for freshness, continuity, and relevance. It automatically curates memories from chat history, allowing ChatGPT to carry forward useful context, follow user preferences (like wildlife photography or strong AC needs), and stay current over time by updating location-based information. Recent improvements reduced compute requirements for Free users by approximately 5x, enabling a broader rollout to Plus and Pro users in the US today, and to additional countries and Free/Go users in coming weeks.

Key takeaway

For AI Product Managers developing conversational AI, you should prioritize memory systems that automatically adapt and update user context over time. This "dreaming" architecture demonstrates that continuous, background memory synthesis is crucial for maintaining relevance and personalization at scale, preventing staleness in user interactions. Evaluate your current memory solutions for their ability to dynamically refresh information and integrate user preferences without explicit prompts, ensuring your product remains helpful and current across long-term engagements.

Key insights

ChatGPT's "dreaming" memory system automatically curates and updates user context for personalized, current, and relevant interactions.

Principles

• Memory systems must evolve with user context.
• Automatic curation enhances relevance.
• Time-awareness prevents staleness.

Method

The "dreaming" method leverages a background process to synthesize ChatGPT's memory state from chat history, automatically curating and updating information to provide fresh, relevant context.

In practice

• Tailor recommendations using user history.
• Update location-based context automatically.
• Personalize responses based on preferences.

Topics

• ChatGPT Memory
• Conversational AI
• AI Personalization
• Memory Synthesis
• OpenAI
• System Scalability

Best for: AI Engineer, AI Product Manager, Director of AI/ML

Related on AIssential

• OpenAI upgrades ChatGPT memory with a new personalization system — ChatGPT's memory evolves from explicit user prompts to an automated, synthesised, and user-editable system for enhanced personalization.
• Your AI shouldn’t forget your recovery journey every time the app updates — AI memory should be user-owned and persistent, not platform-dependent, to maintain critical conversational context.
• How Chatbots Remember Who You Are — Chatbot memory, both short-term and long-term, is essential for maintaining conversational context and enhancing user experience.
• ChatGPT's new memory upgrade is powerful - and could poison every answer it gives you — ChatGPT's advanced memory system, while technically efficient, creates an often inaccurate and unmanageable user profile that can bias responses.
• FBI: AI Fraud = Its Own Category — AI agents can achieve persistent memory and enhanced intelligence through file system-like memory stores and an asynchronous "dreaming" optimization process.
• When Should Memory Stay Silent: Measuring Memory-Use Boundaries in Memory-Augmented Conversational Agents — LLMs often inappropriately integrate sensitive user memory, requiring explicit boundary management at retrieval and generation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by OpenAI News.