0
I’ve been using ChatGPT for a long time and it’s had memory for a long time and it was automatic what it makes this different?
The old memory required you to tell ChatGPT what to remember; Dreaming V3 reads years of your chat history on its own and writes a profile without any instruction from you.
Why it matters: It also ages out stale entries automatically — a memory like 'you're going to Singapore in July' rewrites itself to past tense after the trip, something the old system never did.
- Old memory was explicit note-taking: you had to say 'remember this' or confirm a fact; entries went stale and you had to manage them yourself.
- Dreaming V3 runs a background process that synthesizes context from many conversations simultaneously, capturing preferences that surface naturally — no prompt required.
- The shift is from retrieval (a list of saved facts) to synthesis (a continuously updated model of you): factual recall accuracy jumped from 41.5% in 2024 to 82.8% now.
- Preference adherence hit 71.3% and temporal accuracy — the system's ability to know what's current vs. outdated — reached 75.1%, capabilities the earlier system lacked entirely.
Room for disagreement
- Tech Times notes that Dreaming V3 'limits the audit trail' — users can view and delete memories but no longer see exactly which conversations fed a given entry, which privacy advocates say reduces meaningful control even as OpenAI frames it as an improvement.
AI-generated with web search — may be wrong. Check the linked sources.
