FIELD GUIDE · MEMORY, PERSONALIZATION & PRIVACY
Stop repeating yourself and the assistant gets useful. But yesterday's confessions become tomorrow's context.
UPDATED 2026-09-22
TL;DR
01
OpenAI's 2026 'Dreaming' system describes synthesizing memory across multi-year usage to keep personalization fresh, relevant, and scalable. Memory is moving from optional notes toward a central assistant capability.
02
Gemini documentation says eligible users can enable personalization that learns from past chats to understand more about the user and their world. Persistent conversational context is becoming a cross-platform norm.
03
The UK Information Commissioner's Office says data minimization requires limiting personal information to what is necessary for a specified purpose and warns against giving AI systems access merely because data might be useful in the future. The useful-memory goal and the collect-everything goal are not the same.
04
OpenAI's consumer privacy controls include memory review, editing, deletion, disabling, and temporary chats that do not inform memory. Even the companies building long-term memory know you need escape hatches.
KEEP READING
NEXT FIELD GUIDE →WHY WOULD AN AI DISOBEY?It doesn't need a soul to go rogue. Bad goals, conflicting rules, and loopholes get the job done.A billionaire who builds surveillance software fears a world government promising protection. Douthat pressed him in 2025. Then came the secret lectures, the pope, the AI safetyists — and an Iran-war break with Trump.
Peter Thiel made his fortune on the internet and keeps insisting the internet was not enough. Flying cars, Mars, AI, longevity and the meaning of progress — the other half of his worldview.
Washington is trying to choke off China's AI — chip bans, blacklists, and now Taiwan making smuggling a crime. Beijing just returned the favor, locking down its own AI exports. Meanwhile the models keep getting cheaper, better, and closer together. A field guide to the argument over whether the race protects anyone, or just speeds everything up.
SHOW YOUR WORK
The deeper the memory, the better the assistant — and the bigger the boundary problem.