ChatGPT dots and Space move work beyond the conversation

OpenAI announced dots and ChatGPT Space on 29 September 2026: persistent cloud assistants and a shared place for their work. I see the useful change as continuity. A dot can keep working after your computer shuts down, while Pages give people a common document to inspect and edit. Access depends on plan and region, including a Pro launch restriction across the EU. [1] [2] [4] [5]

What can a ChatGPT dot do while you are away?

A dot can continue an assigned task on its cloud computer while your own computer is off. OpenAI’s documentation describes an assistant that keeps context across projects, can use connected tools and can prepare work for you to review later. Astra supplies the model; the persistent environment supplies somewhere for the work to continue. [1]

That makes a task such as preparing a code change different from a single model answer. The assistant needs to inspect files, run tools, preserve its progress and return a result you can assess. These are also the responsibilities I described in OpenAI’s managed Agents API, although dots expose them as a ChatGPT product rather than an API for your application.

The action boundary matters because persistence is not permission. OpenAI says proactive research uses read-only tools: a dot doing research on its own cannot send messages, change app content or control a browser. Tasks you assign can use authorized actions under the product’s usual rules. I would give it a concrete deliverable and define which connected sources it should use, instead of assuming “keep working” is a sufficiently clear brief. [1]

These are documented launch capabilities, rather than results from my own long-running test. The first practical question is whether the assistant can finish a bounded task and leave a useful record of its decisions.

Who can use dots at launch?

Pro access starts outside the EEA, Switzerland and the UK. Business Premium rolls out across supported ChatGPT regions, while Enterprise access starts as an administrator-enabled beta. OpenAI says rollout can take several days, so plan eligibility does not guarantee that the feature is already visible. [2]

For Pro subscribers in the EU, the initial dots rollout is unavailable. A Business subscription should not be confused with Business Premium eligibility for dots, either. Space has its own plan list, which is why the two products need separate access checks.

Create the first dot on desktop web or in the desktop app, then use the mobile app where supported. Simon Willison ran into that setup requirement during his DevDay live blog when his attempt told him to switch to a computer. He disclosed that OpenAI provided his event ticket and creator-area seat. [2] [6]

What does ChatGPT Space share with teammates?

Space collects files and Pages into a common place to work, replacing Library on the eligible plans. Pages can contain prose, images, charts and interactive tools; teammates with edit access can work on them together, comment and mention ChatGPT or a dot to request revisions. [4] [7]

The useful difference is a shared artifact. A conversation can explain an answer, but a Page lets people inspect and change the actual document. If several agents contribute to a launch brief, a teammate should be able to review the brief without reconstructing every conversation that produced it.

OpenAI separates Page sharing from private chats and personal memory. Sharing a Page exposes its content to collaborators, while linked source files retain their own access permissions. This makes the distinction between working context and published output visible: a source available to one assistant does not automatically become available to everyone opening its Page. [4]

At launch, Page creation and editing work on web and desktop. Mobile supports finding, reading and sharing; collaborative slides and spreadsheets are announced for later. Those limits matter when the intended reviewer works mainly on a phone. [7]

What happens to a dot’s memory and permissions?

Disconnecting an app stops new access but does not remove information the dot has already retained. OpenAI also says individual dot memories cannot yet be inspected or selectively edited. Deleting the dot deletes its own context, while files or chats created elsewhere remain separate. [3]

That is a practical reason to choose access deliberately at setup. Plugin permissions carry across dots, ChatGPT, Work and Codex, and Custom Rules add instructions without replacing the product’s built-in safeguards. Local computer access is optional and starts disabled, so cloud work and access to your own files should be treated as separate choices. [2] [3]

Persistent work still needs deliberate context management, and returning to it still needs reliable session recovery. I would judge dots and Space by how easily I can understand the saved work and continue from it the next morning.