ChatGPT now updates its own memory between chats

Hook

Your ChatGPT is still acting on things that stopped being true months ago. The trip you already took. The hire you already made. OpenAI just shipped the fix, and it's coming to free accounts too.

The news

On June 4, OpenAI started rolling out a new memory system it calls "dreaming," which synthesizes and revises what ChatGPT knows about you over time instead of just storing what you typed. The example OpenAI gives: it rewrites "You're going to Singapore in July" into "You went to Singapore in July 2026" once the trip is over, then goes back to tailoring recommendations to your home and time zone. It's live for Plus and Pro users in the US today, with Free and Go users following over the coming weeks after OpenAI cut the compute needed to run it by roughly 5x. (OpenAI's announcement is here.)

On your desk

Stale memory is the quiet tax on using one assistant for everything: you correct it, or worse, you don't notice it's working off last quarter's facts. Dreaming goes at that directly by keeping the record current on its own. The lever that matters for an operator is the memory summary page, where you can read exactly what ChatGPT thinks it knows about you and fix anything wrong. Treat it like a CRM you actually audit. It's the same direction your own persistent memory setup already points: context that compounds instead of going stale.

What to do next

Open ChatGPT's memory summary page, read what it thinks it knows about you, and correct anything that's gone stale.