OpenAI just launched the most important memory upgrade of ChatGPT since its creation. It's called "Dreaming" and it fundamentally changes how the model remembers who you are: instead of a list of notes that you load, a background process synthesizes your preferences, projects, and context across many conversations. For those who work every day with ChatGPT in LATAM, this redefines productivity and opens serious questions about privacy.

The announcement arrived on June 4 with OpenAI's blog "Dreaming: Better memory for a more helpful ChatGPT", covered by TechCrunch and other media like Dataconomy. It's not a cosmetic touch-up: it's a new memory architecture that starts arriving this week for Plus and Pro users in the United States, with expansion to more countries and to the Free and Go plans in the coming weeks.

What Dreaming is and how it differs

Until now, ChatGPT's memory worked like a list of saved things: you asked "remember this" and it was noted down. Dreaming replaces that approach with a system that synthesizes information automatically, without you having to indicate it. The model connects what appears across different chats and builds a more coherent understanding of your context.

The other big novelty is that the memory stays updated over time. OpenAI gives a clear example: a piece of data like "you're going to travel to Singapore in July" is revised on its own and becomes "you traveled to Singapore" once the date has passed. It's a living memory, not a static file.

The three changes visible to the user

• Memory summary page: a readable view of what ChatGPT synthesized about you.

• Controls to edit: you can add or update remembered details.

• Topic settings: you define which matters the model can mention and when.

What changes for professionals in LATAM

For those who use ChatGPT as a work tool, the benefit is concrete: less repetition of context. You don't have to explain in each chat what you do, what project you're on, or what tone you write with. The model already knows it, which makes each interaction faster and more personalized.

This brings ChatGPT closer to the logic of persistent workspaces that other tools already offered. If you work in a team, it's worth comparing approaches: you may be interested in how Claude projects for teams are used to maintain shared context.

How to configure it and get the most out of it

• Review your memory page as soon as you have access, to see what the model synthesized and correct what's not right.

• Define topic limits: indicate which matters you prefer it not to remember or not to mention.

• Be consistent: the more you use ChatGPT for real tasks, the better the context it will build.

• Audit periodically: living memory changes, so it's a good idea to review it every so often.

The privacy side

A memory that synthesizes on its own, without you asking, raises legitimate questions. How granular is the control? How transparent is the record of what the model infers? Some analyses pointed out that the new system reduces part of the auditable trail compared to the previous model. The recommendation is clear: if you handle sensitive or client information, review the configuration, limit topics, and avoid loading confidential data that you don't want to persist.

Understanding how these systems work on the inside stops being optional for any professional who works with AI. The difference between using the tool blindly and using it with judgment is in knowing what the model does with your data.

Recommended Coderhouse courses

To take advantage of ChatGPT's new memory with judgment and not stay only on the surface, these training programs give you the base:

• Introduction to Artificial Intelligence Course: to understand how language models and their memory work (beginner level).

• AI Engineering Course: to go deeper into how AI systems with context and personalization are built (intermediate-advanced level).

• AI Agents Course: to create agents that leverage memory and persistent context (advanced level).

If you want to see how the whole industry is moving, our analysis of Microsoft Scout, the new AI assistant may also be useful.

Frequently asked questions

What is Dreaming in ChatGPT?

It's the new memory architecture of ChatGPT that automatically synthesizes information from multiple conversations, without you having to explicitly ask it to remember something. It replaces the old list of saved memories and keeps the context updated over time.

Is it available in Argentina and the rest of LATAM?

The launch started with Plus and Pro users in the United States and OpenAI announced that it will be extended to more countries and to the Free and Go plans in the following weeks. The arrival in LATAM is part of that gradual rollout.

Can I control what ChatGPT remembers?

Yes. The system includes a page with the summary of what it synthesized about you, controls to edit or add details, and settings to define which topics it can mention and when. It's a good idea to review it as soon as you have access.

Is it safe for sensitive work information?

It requires caution. By synthesizing context automatically, it's important to review the configuration, limit sensitive topics, and avoid loading confidential client or company data that you don't want to be stored.