OpenAI introduced Dots on September 29 as a persistent assistant inside ChatGPT. A dot can take a goal, work through several steps on its own cloud computer, use connected apps, and return with work for review. Unlike a single chat exchange, it can keep an ongoing project in view between conversations. OpenAI says the system runs on GPT-6 Astra and is beginning a gradual rollout to eligible paid users. The Verge's launch report places the announcement at DevDay 2026.
The practical distinction is responsibility over time. A regular question might produce a summary now. A dot can follow a changing source, notice new information, update a draft, and ask for a decision when the next step needs one. It is an AI agent , meaning software that pursues a goal through actions rather than only returning a response. OpenAI's announcement describes the product and its intended use. Its examples are demonstrations and early-user accounts, not measured evidence that every task will work reliably.
What a dot can do
Each dot has a separate cloud computer with a browser and tools. With access the user grants, it can examine material in connected apps, create files, analyze data, and work on more than one project. The user can inspect the dot's computer and activity, add instructions, or redirect its work. Connecting a personal computer is optional and starts off. OpenAI says dots can also be reached through ChatGPT, Slack, and Teams, although its setup guidance warns that some launch features are arriving gradually.
Dots also perform what OpenAI calls proactive research. In that mode, they can read information from permitted connected sources and save private notes without a new request. Those background research tools cannot themselves send messages, change connected content, or control a browser or computer. A subsequent action goes through the usual instructions and checks. This distinction matters when deciding whether to connect email, documents, calendars, or workplace systems.
OpenAI's two-minute launch video illustrates a dot coordinating a website update, a board-meeting deck, an app migration, and personal scheduling. The film is a product scenario, not a measured account of how reliably those tasks run.
Use cases that fit a continuing agent
A dot is most useful when the work has a continuing goal, several inputs, and a reviewable result. OpenAI gives examples across development, research, launches, sales, and content work. Each can be turned into a specific assignment with a clear stopping point.
Turn customer feedback into a tested fix. Give the dot access to a feedback channel and a software project. Ask it to collect repeated complaints, reproduce one small defect, make a branch, run the relevant tests, and bring back a pull request with a short demonstration. The useful output is a reviewable change and a record of what it tested. A developer still decides whether the diagnosis is right and whether the change should be merged.
Keep an analysis current. Point the dot to a defined dataset, an existing analysis, and the research question. Ask it to check for new data, rerun the same method when data changes, investigate surprising differences, and prepare revised figures with a change log. The result is more useful than a fresh summary because it shows what moved and why. A researcher checks the data, method, and interpretation before using the figures in a paper or decision.
Follow a launch through scope changes. Connect the approved product brief, release plan, and draft communications. Ask the dot to compare each new product decision with the current FAQ, launch email, and landing-page copy. It can prepare proposed edits and flag claims that no longer match the product. The team approves the final promise to customers and any publication or outbound message.
Process a recurring content intake. When a new interview transcript arrives, ask the dot to identify moments worth clipping, supply timestamps, draft show notes, and prepare a few posts in an established voice. The editor gets a package to review rather than starting every asset from a blank page. Check quotations and context against the recording, then approve any public post.
Catch an administrative loose end. An independent professional could ask a dot to compare completed assignments with an invoicing log, prepare a missing invoice from approved terms, and flag a due date. OpenAI describes an early tester whose dot prepared an overlooked invoice and sent it only after approval. That approval boundary is important: finding an omission and drafting the document are different from sending a financial request to a client.
Coordinate a personal plan when details change. A household could give a dot the approved schedule and a short list of vendors, then ask it to find a replacement if a booking falls through, compare available times, and propose a revised plan. OpenAI's launch film depicts a canceled wedding cake vendor and a replacement tasting. In a real assignment, the person should approve any new booking, payment, or message before the dot commits to it.
These are proposed workflows based on OpenAI's stated capabilities and examples. They do not establish a success rate for unattended execution. A one-off factual question, a short rewrite, or a task with no follow-up may be simpler in an ordinary ChatGPT conversation.
A small first assignment
Start with one project and a result that is easy to inspect. For example: “Watch the approved launch brief and product document until Friday. If a requirement changes, draft the corresponding FAQ and email edits, show me the differences, and ask before publishing or messaging anyone.” This defines the sources, time window, output, and approval point. After seeing the first result, correct the dot's interpretation before granting more app access or a broader goal.
How to start and what to check
OpenAI says eligible users can create a first dot from the ChatGPT desktop app or a desktop browser, name it, then connect the apps it needs. The Help Center's rollout details say Pro access is starting outside the European Economic Area, Switzerland, and the UK. Business Premium is available across supported ChatGPT regions, while an Enterprise workspace admin must enable the beta. Access is gradual, so eligibility does not guarantee that the control appears immediately. Creation is not yet available on mobile, although users can message an existing dot in the mobile app when mobile access is available.
OpenAI says the first dot is included with Pro or Business Premium and that conversation with it does not consume ordinary ChatGPT usage limits. Deeper work has a plan allowance, and tasks the dot starts or manages in Codex or ChatGPT Work use those products' normal limits. The company says launch allowances are extended for the first month, with later usage terms still to come. That makes the amount of sustained work a dot can perform a practical detail to check before relying on it for a large project.
Permissions deserve the same attention as the task. Choose the connected apps, inspect the activity, and use Custom Rules to specify actions the dot may take, must ask about, or must avoid. OpenAI says separate action review checks steps that could change accounts or share information. Its safety explanation also identifies prompt injection in webpages, email, and documents as a risk for agents that act on outside material. The safeguards can pause suspicious work, but OpenAI says dots can still make mistakes. Review a draft, file change, message recipient, or financial detail before treating it as complete.
The first test of Dots is therefore a modest continuing assignment with clear permissions and an output you can inspect. Its value is in carrying that assignment forward as information changes, while keeping decisions that matter visible to the person responsible for them.
