OpenAI has developed a new type of assistant that functions continuously even when the user is idle. The company announced the launch of Dots, personal agentic assistants, during the Dev Day event on Tuesday. Dots is a personal agentic assistant driven by GPT-6 Astra. According to OpenAI, these are “always-on agents that can handle everything,” available in ChatGPT the same day.
TechCrunch reporter Lucas Ropek wrote an article about the launch of Dots. He argues that they combine features available through Codex and similar technologies that allow an agent to act independently. Moreover, Dots differ from previous tools in the design, as they are presented as small floating cartoons to make the software more approachable.
What Dots Are
OpenAI indicates that Dots are designed to work toward targets set by the user, persistently and with little human input needed. You don’t work in a conversation window, asking the dot questions and getting answers. You assign the dot a task, and it goes away and does it.
The setup begins with a single primary dot. OpenAI’s announcement indicates that you can personalize that dot, giving it a name and making it yours. This individualization is part of the appeal, as seen in the cartoon mascot covered below.
The company’s longer vision is considerably more ambitious. OpenAI indicates that in the future, it intends to have systems in which multiple Dots work together on behalf of the user. For now, the company is focusing on the launch of a single dot with the suggestion that a team is a goal for later.
How Dots Differ From Codex and ChatGPT
Unlike a particular piece of hardware or interface, dots are not specifically constrained to either Codex or ChatGPT. This is the key difference TechCrunch draws between the two similar but different technologies.
A dot is more human-like than a window. It is constantly being driven toward the goals it is given rather than waiting to be prompted by a person at each step in the process. However, Codex or ChatGPT can still be used to launch a dot, so the two technologies remain connected even though it is not entirely contained within either one.
Who Can Use Them Now
Dots are available in ChatGPT for Pro and Business Premium users in eligible markets, starting on Tuesday. A small first audience, limited both by plan and by region.
The report from TechCrunch only mentions eligible markets, not any plans to roll this out to free or cheaper plans. If you don’t have a Pro or Business Premium subscription, you’ll have to wait until OpenAI tells you anything else.
A small initial user base gives OpenAI something to learn about before they release to a bigger audience. The report does not say whether or not such a wider launch is planned at any point.
How People Talk to Dots
Users can communicate with Dots through Slack, Teams and other organizational platforms. Text message support is coming soon.
This choice matters: Most work place requests tend to already be made using chat tools, and a dot that responds there doesn’t require users to adapt to a new medium, but can slot into the conversation wherever it already happens. And the text message option, when it comes, would extend that to users who aren’t at a desk.
Two Examples From OpenAI
OpenAI’s announcement sketches a few ways Dots could be used. TechCrunch highlights two.
- A software developer deploys a dedicated dot to monitor customer feedback, then implement bug fixes and requested features as needed
- A scientist employs a dot to rerun analysis and investigate unexpected results as experimental data becomes available
Both examples have the same pattern: the action is ongoing, it depends on new incoming information, and it does not require a participant to press a button every round. This is the gap that OpenAI is targeting with its announcement. A chat assistant responds to a question; a dot is intended to react accordingly.
These are hypothetical scenarios, not results that OpenAI has reported, as the announcement, as TechCrunch has noted, does not provide figures for how many of Dots’ tasks have been undertaken.
Specialist Dots and Access Controls
OpenAI also proposed the concept of specialist Dots, each with a particular responsibility. Individual Dots may be allocated particular identities, credentials, and equipment in existing systems.
This is an easy detail to overlook, but it is this very detail that makes the concept so appealing at the corporate level. After all, a dot that fixes flaws or reruns analyses needs access to actual systems, and corporations are usually eager to ensure that only the right people and programs have access. By allocating individual identities and credentials, each dot can be incorporated into the same permission control framework that a corporation already employs.
OpenAI is already collaborating with Microsoft to incorporate Dots into the company’s Agent 365 security controls. TechCrunch does not go into detail regarding what the collaboration entails, but the mere fact of the announcement suggests that OpenAI is aware that corporations will want to have oversight capabilities over agents that, by design, should require limited supervision from individual users.
Teams of Dots Are Still an Idea
The team concept needs to be mentioned separately. According to OpenAI’s message, they plan for teams of Dots to work together, but it is presented as something that could come later. The introduction focuses on single dots, and how several of them would split up the work and not interfere with each other is not something the announcement covers.
Old Capabilities, New Package
Ropek makes a point worth noting. Dots are a clever product that bundles a lot of things together. They were available before in Codex and similar agentic harnesses, but the bundling changes the game. The value proposition is different: it focuses on agential actions with independent capabilities.
This is an important distinction. It means that the discontinuity caused by Dots is primarily a bundling discontinuity. It is not a discontinuity of capabilities: those that were available before are still available in Dots. But bundling changes the ease of use and creates a new continuity of experience. The continuity hypothesis suggests that it will change the way users interact with these tools, but it remains to be seen if it will have a significant impact on their capabilities.
For users that are already using Codex, the question is what does a dot add to the mix. The answer, based on the TechCrunch review, is that it provides a continuity of experience, focused on independent actions. It does not offer new capabilities.
Why the Cartoon Look
Dots look like dots. The branding makes them look like something straight out of a cartoon, making it reminiscent of Meta’s recently introduced Muse agent, says TechCrunch. These little floating cartoons are a change for OpenAI’s software.
Ropek says this is the latest example of an industry making an effort to make AI more relatable. There’s a good reason for it. An agent that works behind the scenes and takes action on your behalf is a bigger ask in terms of trust than a chatbot that merely answers questions. A friendly face is a good way to reduce that ask.
Moreover, this means that two of the biggest players in AI, OpenAI and Meta, have now released agents that have a whimsical personality. TechCrunch stops short of saying this trend suggests anything in particular. It’s still far too early to form any conclusions.
What’s Still Unclear
The announcement leaves several questions open, at least in the details TechCrunch reported.
- Pricing beyond the Pro and Business Premium plans isn’t mentioned
- The list of eligible markets isn’t given
- There’s no timeline for text message support beyond “soon”
- It’s not clear how a user reviews or limits what a dot does while it works with minimal oversight
The last point is essential. Minimal oversight is the sell, but it will be the thing that gets examined. The Microsoft integration suggests how OpenAI intends to deal with it for enterprises. How individual users might keep tabs on what their dot is up to is not addressed in the piece.
The Bottom Line
OpenAI’s Dots are its bet that people want assistants that run continuously, take on goals and work across the tools people already use. The first version is limited to Pro and Business Premium users in eligible markets, and much of the underlying ability already existed elsewhere in OpenAI’s lineup.
What’s new is the shape of the product. A named, personal agent with a cartoon face, reachable through Slack or Teams, working in the background on goals you set. If that shape catches on, teams of Dots may follow, as OpenAI says it hopes. If it doesn’t, Dots will look like a repackaging of tools that were already there.
FAQs
Q1: What are OpenAI Dots?
Dots represent the latest innovation in personal agentic assistant from OpenAI, and they are powered by GPT-6 Astra. They allow for continuous pursuit of user-specified goals with minimal supervision.
Q2: When did OpenAI announce Dots?
OpenAI unveiled Dots at its Dev Day event on Tuesday, and they became available within ChatGPT the same day.
Q3: Who can use Dots right now?
Dots are currently available in ChatGPT for Pro and Business Premium users in eligible markets, and they can be launched from either Codex or ChatGPT.
Q4: How are Dots different from Codex or ChatGPT?
Unlike the third-party models, there is no particular physical location associated with a dot, and they operate independently within their own background environment.
Q5: How do you talk to a dot?
Slack, Teams and other enterprise messaging services can be used to send messages to a dot, while OpenAI said that text message support was coming soon.
Q6: What are specialist Dots?
Specialist Dots take on specific responsibilities. Each can be given its own identity, credentials and tools through existing systems.
Q7: Does OpenAI work with Microsoft on Dots?
Yes. OpenAI says it’s already working with Microsoft to integrate Dots into the company’s Agent 365 security controls.
Q8: Is this a brand new capability?
Not exactly, as TechCrunch noted, much of the capability was already available through Codex and other agentic harnesses, but it was packaged into an independent agent focused on carrying out actions on its own.
