Descriptions:
Igor of The AI Advantage breaks down two significant OpenAI product moves announced around the June 2nd ‘Intelligence at Work’ event: a comprehensive overhaul of ChatGPT’s memory system and the announced merger of the Codex agentic app into the main ChatGPT interface. The new memory format moves ChatGPT away from simple extracted lines toward a structured summary report — covering work context, personal context, top-of-mind items, and recent history — closely mirroring the format Claude has offered for several months. Igor walks through how to navigate, correct, and update these memories in real time, with a clear warning that inaccurate entries will silently skew every subsequent conversation.
The episode includes a live side-by-side comparison of ChatGPT and Claude’s memory implementations, noting that Claude’s version additionally captures standing tasks and output format rules. Igor also explains the underlying ‘dreaming’ pattern — where the system periodically reviews past conversations and consolidates them into the memory store — tracing its origins to standalone agents like Open and Hermes before it migrated into consumer AI platforms.
The new ChatGPT memory is live for paid US accounts at time of recording, with a two-week rollout window to free accounts globally. For power users who rely on persistent context across AI tools, this episode is a practical guide to auditing and optimizing ChatGPT’s updated memory layer and understanding where it currently stands relative to Claude.
📺 Source: The AI Advantage · Published June 05, 2026
🏷️ Format: Tutorial Demo
