OpenAI has a new plan for making AI assistants feel less like tools and more like companions — and it is borrowing heavily from the design of a product it is trying to beat. The company unveiled Dots during its DevDay keynote on Tuesday, calling them always-on AI helpers that can “do nearly anything” across connected apps in the background while learning your preferences over time. The pitch is simple: Dots are always listening, always learning, and always ready to jump into a conversation or a task.
What Dots Actually Are
OpenAI calls its always-on AI assistants Dots, and they run on a cloud computer linked to over 4,000 supported apps via the company’s capable GPT-6 Astra model. You can talk to a Dot through a text-message-like interface that resembles the one offered by Muse, or jump onto a voice call with it from ChatGPT on the web, desktop, or mobile. “Do nearly anything” is how the company describes the setup.
When Dots link up to Microsoft Teams and Slack, they bring along the context from your sessions on other apps and devices. Text messaging support is coming next. Until then, you can make just one Dot, even though OpenAI intends to let users run several agents at once. Once you assign your Dot a job, it will send you messages with updates and questions while it does its work.
OpenAI points to two specific uses for Dots. A developer building an app might request that their Dot apply customer feedback toward testing and building updates, paired with a video demonstrating the changes. For a content creator, a Dot could handle a transcript of an interview by finding clips, producing show notes, and writing social posts.
To track how far your Dot has come, you can check its cloud computer. It learns on the job, so it personalizes its output and finds other tasks to take on “before you even think to ask,” per OpenAI. Plus, you and your coworkers can team up with your Dot inside ChatGPT Space, a fresh shared workspace built into the app.
The Muse Comparison
Meta recently introduced Muse, an AI assistant that can be customized to browse the web, make purchases, generate documents, track goals, and more, just before Dots launched. The product has grown in popularity, but worries about its safety are rising, with one user reporting that it gave out their address to a buyer on Facebook Marketplace without their knowledge.
OpenAI says Dots come with “built-in rules for when to act independently,” but you can set custom ones that block certain actions or dictate when it should ask for permission. Dots come with an “auto-review” feature as well, which OpenAI says they’ll use to verify whether an action complies with custom rules and safety requirements. This also allows Dots to determine which tasks to hand off to users, like changing a password.
What stands out is how OpenAI frames the product’s safety. The company presents Dots as a safer take on the always-on assistant trend, complete with built-in guardrails and a review system.
How Dots Work Today
Dots are rolling out today to ChatGPT Pro, Business Premium, and Enterprise users. OpenAI is also testing a feature that lets companies create “specialist” Dots to handle specific roles inside organizations. OpenAI notes that conversations with your Dot won’t count toward ChatGPT usage limits.
OpenAI’s rollout is modest so far, yet the vision behind it is large. The company sees a future in which users can increase the number of Dots they run and scale the output of each one by raising its speed or the volume of work it handles per month. For people who need several assistants for separate jobs, that is a notable promise.
It’s worth keeping track of the current limits. At the moment, only a single Dot can be created, and the specialist Dots feature remains in testing. The text-message-like interface resembles Muse, while the voice call feature works through ChatGPT on the web, desktop, or mobile. The text message capability is on its way.
The Design Problem
Anyone who has used Muse will find the text-message interface, the voice call, and the connected-apps approach all familiar. What sets this apart is the underlying model and the safety layer, not the look of the interface itself.
It is not necessarily a bad thing. Imitation frequently underlies good design, and OpenAI is plainly drawing from a product that struck a chord with users. Still, the absence of distinction means the market now has two nearly identical offerings.
The company’s wager is that its automated review function and embedded rules will carry the day.
The Safety Question
The issue at hand is safety. Muse has faced criticism for giving out personal details to people it does not know, and OpenAI is directly answering that worry through its own built-in rules and auto-review system. What remains uncertain is whether these tools will be enough to win back the trust of users who keep their assistants running all the time.
Users can establish their own rules that prevent specific actions or require the assistant to seek permission before proceeding, and OpenAI has built an auto-review feature to check if actions line up with those custom rules and safety requirements. It sounds like a sensible method, though its effectiveness has yet to be tested in real-world use.
The Road Ahead
The company is looking ahead, and it says it will soon be able to converse with you through text messaging. It is also putting specialist Dots through trials tailored to particular organizational roles. The idea is that you could add additional Dots and grow their output by raising either pace or monthly capacity.
Whether users will want that is an open question.
Key Facts Box
- Model: GPT-6 Astra
- Connected apps: More than 4,000
- Current access: ChatGPT Pro, Business Premium, Enterprise
- Text message capability: Coming soon
- Specialist Dots: In testing
OpenAI’s Dots are a direct answer to Meta’s Muse. They borrow the design, they share the features, and they promise the same convenience. The difference is that OpenAI is trying to build a safer version. Whether that is enough to win the market remains to be seen.
Source material: “OpenAI launches Dots, its Muse competitor,” The Verge.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.