OpenAI has a new product it hopes will make ChatGPT feel less like a tool you open and more like a colleague who never clocks out. Unveiled at its DevDay keynote on September 29, the feature is called dots: always-on personal AI agents that get their own cloud computer, learn how you like things done, and keep working toward your goals around the clock.
Each dot is powered by GPT-6 Astra, the company’s flagship model, and connects to more than 4,000 apps through OpenAI’s plugin ecosystem. You give it a name, tell it what matters to you, and it starts handling work in the background, sometimes before you think to ask. At DevDay, OpenAI gave the example of an early tester whose dot noticed an invoice that had never been sent to a publication, prepared it, and fired it off once the tester approved.
Dots arrived alongside OpenAI’s GPT-6.1 Sol launch, which we previewed ahead of DevDay, and landed the same week the company shelved another model over safety concerns. Taken together, DevDay 2026 was OpenAI’s most aggressive product push yet, and dots is its clearest bet on the future: not better chat, but agents that do the work.
A cloud computer that works while you sleep
The core idea is simple to describe and wild to think about. Your dot lives on its own cloud computer, separate from yours, with its own browser and access to the apps you connect. It can run several projects at once, and you can hand it new tasks without juggling separate chat threads or micromanaging every step.
It meets you where you already work. Dots are reachable through ChatGPT on desktop, web, and mobile, plus Slack and Teams, with texting promised soon. You can message it, hop on a voice call to talk something through, or let it ping you with progress updates and decisions that need your sign-off. Context carries across every channel, so a project started in ChatGPT can be finished in Slack without repeating yourself.
OpenAI also says you can inspect your dot’s cloud computer at any time to see exactly what it is doing, and optionally let it connect to your own laptop so it can work directly alongside you.
Proactive research: the always-on part
The feature most likely to raise eyebrows is what OpenAI calls “proactive research”. When you are not actively working with it, your dot scans connected apps for ways to help, a design choice that turns an assistant you summon into an agent that volunteers.
OpenAI is clearly aware this is the scary part, so it front-loaded the safeguards. Proactive research uses read-only tools, meaning the dot cannot send messages, change app content, or control your browser or computer while snooping around. A monitoring system can pause or stop the dot if it detects a safety concern, and an auto-review step checks any action that could affect your accounts or share information against your instructions and the built-in safety rules.
There are also hard lines. Certain sensitive tasks, like changing a password, are never delegated. Dots can use saved passwords to sign into websites on your behalf, but OpenAI says those credentials are never exposed to the model itself.
What OpenAI says it will not do with your data
For business and enterprise customers, the pitch is straightforward: content from ChatGPT Business, Enterprise, and Edu workspaces is not used to improve OpenAI’s models by default. On personal plans, you can control whether your dot’s conversations and work feed into model training.
Notably, OpenAI says it does not train directly on proactive research or on your dot’s private notes to itself. Information from those may still flow into an eligible conversation or task depending on your settings, which is worth keeping in mind before connecting your email or calendar.
As always with AI products, the honest advice applies: dots can still make mistakes, so review anything consequential before it leaves the building.
Dots vs Muse: the agent race is officially on
Dots do not arrive into a vacuum. Meta’s Muse agent is the obvious rival, and both companies are racing toward the same endpoint: a personal agent that knows you well enough to act without being asked. OpenAI’s differentiator is the infrastructure around it. The 4,000-app plugin ecosystem gives dots a running start on actually doing things in other software, and the planned integration with Microsoft’s Agent 365 governance tools is a deliberate play for enterprise buyers who need IT oversight.
The rollout starts narrow. Dots are launching for ChatGPT Pro, Business Premium, and Enterprise users in eligible markets, with the first dot included at no extra cost and conversations with it not counting against ChatGPT usage limits. OpenAI says more users, more dots per person, and the ability to scale each dot’s speed or workload are coming later.
What dots say about where OpenAI is headed
Every major AI company is converging on agents, but dots make OpenAI’s strategy unusually concrete. The chatbot era was about answering better. The agent era is about being present: a dot that learns your preferences, watches for things it can handle, and quietly finishes work in the background. If the company gets the trust equation right, this is the product that moves AI from something you use to something that works for you. If it gets it wrong, an always-on agent with opinions about your inbox is a much harder thing to forgive than a chatbot that hallucinated.
