ChatGPT is smarter now that it's learned to forget - a huge memory upgrade is coming (www.techradar.com)

OpenAI has upgraded ChatGPT’s memory system with automatic memory prioritization: instead of filling up or forcing users to manually prune stored facts, the model now ranks memories by recency and frequency so relevant items stay “top of mind” while less-used details fade. The change doesn’t delete data automatically—ChatGPT reweights what it surfaces—but users can still view, search, sort, re-prioritize, roll back, delete memories or disable memory entirely. The feature rolls out to Plus and Pro web users globally starting October 15, 2025. Technically simple but impactful, automatic prioritization reduces maintenance overhead and gives ChatGPT persistent, context-aware continuity across sessions, moving it closer to a long-term assistant rather than a purely reactive tool. For the AI/ML community this matters because memory management is a core scaling problem for personalized LLMs: heuristics like recency/frequency can keep context relevant without unbounded storage or user burden. It also raises familiar trade-offs—automation can misprioritize, so manual inspection and privacy controls remain important—but overall this is a pragmatic step toward sustainable, user-controllable personalization in deployed LLM systems.