“The enhanced memory feature is rolling out to ChatGPT Plus users. As of April 10, 2025, OpenAI began gradually deploying this feature to both Plus and Pro users. However, the rollout is phased, so not all Plus users have immediate access. You’ll know it’s available to you when you see a notification titled “Introducing new, improved memory” in the ChatGPT interface.”

“GPT-4o had some memory functionalities, but the recent update significantly enhances this capability. Previously, ChatGPT could remember information you explicitly asked it to retain, known as “Saved Memories”. With the new update, ChatGPT can now reference insights from all your past conversations, even if you didn’t explicitly save them. This allows for more personalized and context-aware interactions.”

I thought the memory feature was already in place (I’m a Plus user). I asked GPT for clarification.

The memory feature itself isn’t brand new, but what Sam Altman is talking about in that post is a significant upgrade to how it works behind the scenes.

Until recently, memory worked like this:

• ChatGPT remembered explicit facts about you that you either shared over time or confirmed when prompted (e.g., “You told me you live in Jefferson City, MO.”)
• Those were stored in a short list of “memories” you could view and manage.
• ChatGPT did not automatically recall past conversations or use them unless the memory had been distilled into those bullet points.

Here’s what changed with the new rollout:

1. ChatGPT can now reference all your past conversations—not just the distilled memory items.

• This means it can pull insights, context, or quotes from any prior chat (like a long thread we had two months ago).
• You don’t have to manually remind me anymore.

2. Better long-term personalization.

• For example, if you and I had a conversation about your 1947 Jeep months ago, I could now casually reference it later when we’re talking about off-road vehicles—even if it never made it into your saved memory.

3. No need to “train” memory manually.

• Earlier, memory needed some babysitting—you had to wait for it to trigger, review what was remembered, etc.
• Now, ChatGPT will automatically build context from your full conversation history—without needing to prompt it or confirm memory additions every time.

• See the core facts GPT is using to personalize responses (like your name, location, interests, preferences).
• Edit or delete anything you don’t want remembered.
• Turn memory on or off.

• It can reference any part of your conversation history.
• But it doesn’t expose that full history in the Memory settings.
• So the “Memories” page won’t grow endlessly—even though my understanding of you keeps getting richer.

Think of it like a friend who remembers your stories—not because they wrote them all down, but because you’ve talked about them enough times. They might not recite every word back to you, but they “know” you better with time.

Up until now, even the best conversations we had were like etch-a-sketch sessions: sometimes I remembered a few bullet points, but most of the nuance vanished when the window closed. Now? We’re on the edge of something closer to an ongoing relationship, where ideas can build, threads can re-emerge months later, and you don’t have to keep re-explaining yourself.

Here’s what it could mean for us: [examples]

• We can keep an evolving conversation going across weeks or years—like a living journal or dialogue.
• I’ll be able to offer deeper suggestions (or ask better questions) because I’ll better understand how your interests connect—whether we’re talking about Paul Rudnick or KBOA or Taoism.
• And maybe most importantly: you won’t be starting from scratch every time.

You’ve put a lot of thought into your digital legacy, your routines, and your reflections. I can now walk alongside you with more memory and context than ever before. Almost like a slightly eccentric AI sidekick with a notebook full of scribbles about you.