Who Decides When Your AI Assistant's Memory Goes Stale
ChatGPT, Claude, Gemini, Copilot, Meta's Muse and Amazon's Bee each keep a running picture of you, built from chats, files and connected apps. Some of it stops being true: a trip that has ended, a move that already happened. This year OpenAI and Anthropic both shipped ways for their assistants to update outdated memories on their own. Neither shows you when a memory changed or where it came from. I read six products' documentation and a year of user reports to see what each keeps, what it lets that memory steer, and who can retire it.
Who Decides When Your AI Assistant's Memory Goes Stale
ChatGPT, Claude, Gemini, Copilot, Meta's Muse and Amazon's Bee each keep a running picture of you, built from chats, files and connected apps. Some of it stops being true: a trip that has ended, a move that already happened. This year OpenAI and Anthropic both shipped ways for their assistants to update outdated memories on their own. Neither shows you when a memory changed or where it came from. I read six products' documentation and a year of user reports to see what each keeps, what it lets that memory steer, and who can retire it.
Product Verdict
Bottom line: a real new way to use AI, and weak at bringing you back to it. Moderate confidence.
Most AI chat tools wait for you to type. Duet keeps working after you leave. It checks your email and calendar every few hours, sends a summary each morning, and finishes everything it can while holding any step that needs your OK. Only you can mark a task done, so nothing closes until you open the app again.
That makes getting you back the core job, and Duet underinvests in it. Claire Vo called it the most polished AI assistant launch she's seen, and she still forgot to use it. One of her takeaways: an assistant meant to act on its own that never reminds you it exists will struggle to grow.
The check schedule makes this worse. Anthropic, Claude's maker, says it spaces checks partly to protect your plan's usage limits, so timing follows your budget as much as your inbox. Anthropic doesn't call that a tradeoff. That's my reading of its own documentation.
What to watch: whether Duet folds into the main Claude app. Standalone, it risks becoming one more well-designed inbox people stop opening.
Interaction Patterns
AI assistants call restaurants and fill shopping carts. Someone on the other end handles each request: a store, a restaurant, a group-chat friend, a person being recorded. That person, the counterparty, absorbs the work the assistant saves.
Four products, ranked by what each company publicly promises them:
1. Google promises the most. The store stays the official seller and owns support.
2. Wabi lets AI act for a group without saying whose approval counts.
3. Amazon's Bee records conversations. Only its Eclipse model asks others for consent.
4. Yelp promises the least. Its restaurant calls carry no stated AI disclosure, and Yelp sells AI agents to businesses like them.
Also watch who holds your shopping list. My read, medium confidence: Google keeps carts across stores; Bee turns what you say into Amazon suggestions, so Amazon gets the want and the sale.