At its DevDay keynote, OpenAI introduced Dots, a class of always-on AI assistants. The company says these assistants learn a user's preferences over time and, working in the background across connected apps, can "do nearly anything." The Verge frames the announcement as a direct answer to Meta's Muse, which the report says rolled out only weeks earlier.
Each Dot is powered by OpenAI's GPT-6 Astra model and runs on its own cloud computer, according to the source. That setup gives a Dot access to a web browser and more than 4,000 supported apps. Users interact through a text-message-like interface that the report compares to Muse's, and can also start a voice call with a Dot from ChatGPT on web, desktop or mobile.
The assistants reach into workplace tools as well. Dots can connect to Microsoft Teams and Slack, where the source says they carry over context from a user's sessions in other apps and devices. Text-message interaction is described as coming soon rather than available now.
Availability is narrow at launch. Starting the day of the announcement, OpenAI is rolling Dots out to ChatGPT Pro, Business Premium and Enterprise subscribers. Users can create only one Dot for now, though the company says it plans to let people deploy multiple agents at once.
What is being described here amounts to delegation of tasks, not a back-and-forth conversation. After a task has been handed over, updates and questions reach the user by message from a Dot as the work proceeds. Among OpenAI's examples, one involves a developer who asks a Dot to take customer feedback and use it to build and test app updates, then present a video showing the changes; another involves a content creator who has a Dot go through an interview transcript, locate moments worth clipping, produce show notes, and compose social posts.
Progress can be followed by opening the Dot's cloud computer. The source says Dots learn while they work, personalizing output and proposing additional tasks "before you even think to ask." Colleagues can also work with a Dot inside ChatGPT Space, described as a new collaborative workspace in the app.
A feature is also being tested by OpenAI that would allow companies to build "specialist" Dots for particular roles within an organization. ChatGPT usage limits, according to the company, will not be affected by conversations held with a Dot, and future plans reportedly include adding further Dots while scaling the output of each one, either by raising its speed or by enlarging how much work it can handle in a month.
On autonomy, OpenAI says Dots ship with "built-in rules for when to act independently," which users can supplement with custom rules that block certain actions or dictate when the assistant should ask for permission. An "auto-review" feature is said to verify whether an action complies with those custom rules and safety requirements, and to determine which tasks should be handed back to the user, such as changing a password.
The competitive framing matters. The Verge reports that Meta's Muse can browse the web, make purchases, generate documents and track goals, and that safety concerns are growing around it — including one user's account that Muse disclosed their address to a buyer on Facebook Marketplace without their knowledge. That is a single user's claim as reported, not an independently verified finding, and OpenAI's safety features are company descriptions rather than tested results.
For freelancers, designers and developers, the practical question is whether delegated background work replaces billable hours or adds a coordination layer. The documented examples — building and testing app updates from customer feedback, or turning an interview transcript into clips, show notes and social posts — map onto tasks many independent workers already sell. Whether a Dot's output is usable without heavy review is not established by the evidence.
Cost is a real unknown. The source gives no pricing for Dots, no per-task or per-minute rates, and no detail on what the promised speed or monthly work-capacity scaling would cost. Access is gated behind the higher-tier ChatGPT plans, so freelancers on Plus or free tiers cannot evaluate it yet.
There are also open technical questions the announcement does not answer: how the 4,000-plus app integrations are authenticated, what happens to a Dot's cloud computer state between sessions, how custom rules interact with auto-review in edge cases, and how much of a Dot's learning persists or transfers. None of these are addressed in the supplied material.
The broader pattern is agentic assistants moving from single-turn chat to persistent, tool-using workers with their own compute. OpenAI's framing — background operation, preference learning, proactive task suggestions — is a claim about behavior over time, and the evidence here is an announcement, not a longitudinal test.
A sensible posture for this audience is to treat Dots as a capability to watch rather than a workflow to depend on. If you already pay for a qualifying ChatGPT tier, the low-risk move is to point a Dot at a bounded, verifiable task and judge the output against your own quality bar; if you do not, there is no documented reason to upgrade on the strength of this announcement alone.