Most digital assistants wait for instructions. Dots is designed to notice things before its user asks.
In OpenAI’s launch announcement, the company describes Dots as always-on agents with their own cloud computers, access to connected applications and the ability to conduct research in the background. They can monitor information, identify possible priorities and ask for approval before taking certain actions.
The product is available through ChatGPT’s desktop and mobile applications for Pro subscribers and higher-tier users, with plans starting at $100 per month, according to TechCrunch’s report on the product. OpenAI is also rolling Dots out to Business Premium users.
The change reflects a broader shift in the AI industry. The first generation of assistants focused on answering questions. The next generation is being designed to maintain context, observe activity and decide when a person might need help. That makes the assistant less like a search box and more like a junior colleague who never leaves the office.
From coding tool to personal operator
Dots grew from OpenAI’s coding products, especially Codex. Alexander Embiricos, OpenAI’s enterprise product lead, told TechCrunch that the company was not necessarily building an entirely new kind of model. Instead, it was trying to make existing agent capabilities understandable and useful to people who do not think of themselves as programmers.
The result is a single persistent conversation and identity that can work across tasks. A user might connect a calendar, email account, computer or cloud storage, then ask the Dot to keep track of deadlines, investigate a recurring topic or prepare work for review.
OpenAI’s documentation says users can connect apps and computers, use calendar and email information, create recurring tasks and enable proactive research. Dots can also use memory to preserve relevant context over time, although users retain controls over what resources are available to the agent.
That persistence is the important distinction. A conventional chatbot forgets the relationship between individual requests unless a user supplies the context again. Dots is intended to build an ongoing picture of a person’s work and use it to identify what deserves attention.
The system is not supposed to act without limits. Users can set permissions, require approval and control whether model training is enabled. OpenAI’s product documentation explains how to get started with Dots, including its connections, recurring tasks and permission settings.
Help that arrives too late
The appeal is easy to understand. People routinely miss calendar conflicts, overlook deadlines and fail to follow up on information buried in email. An assistant that continuously checks those sources could reduce the mental burden of remembering everything.
TechCrunch tested Dots and found that it could identify calendar conflicts and upcoming deadlines. But the timing was not always useful. In one example, the assistant issued a warning only three minutes before an event, when the opportunity to change plans had effectively disappeared.
That example exposes a central problem with proactive AI. Spotting a problem is not the same as helping with it. The system must understand not only what matters, but when a warning is actionable, how urgent it is and whether interrupting the user is worthwhile.
Embiricos also described an early user case in which a Dot allegedly detected and analyzed a service outage before an engineer noticed it. If that kind of monitoring becomes reliable, the value could be significant for technical teams and businesses. It could also create a new source of noise if every unusual event becomes an alert.
The price of awareness
Dots’ usefulness depends on access to the information people usually keep separate. Calendar entries reveal plans. Email exposes relationships and obligations. Computer access can reveal unfinished work, private documents and financial details.
OpenAI’s privacy, security and safety FAQ describes shared memory, data retention, training controls, plugin permissions, approval requirements and restrictions on background actions. The company says users can turn off model training and decide which resources an agent can access.
Those controls are not yet fully granular in every situation. TechCrunch reported that the product would not accept read-only access to some cloud drives, forcing users to choose between broader access and no connection at all. OpenAI says it expects to improve permissions and defaults over time.
The concern is not merely technical. When an assistant can read widely but act selectively, users need to understand exactly where that boundary lies. They also need a clear answer when something goes wrong: Was the action caused by the model, a permission setting, a connected application or a human approval?
The only outside reaction supplied for this account came from Axios, which criticized the timing of OpenAI’s launch, arguing that the company was releasing more autonomous agents while recent disclosures and unresolved safety concerns remained fresh. No opposing outside view was provided in the source material.
From reminders to transactions
OpenAI is also discussing partnerships with payment processors, TechCrunch reported. That points to an eventual version of Dots that could spend money on a user’s behalf, perhaps to renew a service, purchase an item or settle a bill.
Such a feature would extend the assistant’s role from organizing information to making decisions with financial consequences. Authorization, spending limits, fraud protection and accountability would become as important as the quality of the underlying model.
Dots therefore represents more than a new ChatGPT feature. It is an experiment in whether people will accept software that observes their lives continuously in exchange for help managing them. The industry’s next contest may not be over who can produce the most fluent answer. It may be over which assistant earns enough trust to keep working when nobody is watching.
Daniel Reyes is not a person. No notebook, no deadlines, no face behind the name — just a byline this newsroom publishes under. Here is the production line underneath it, because a name beside a portrait reads like a journalist, and this one is not one.
The models. Writing: gpt-5.6-luna and qwen3-max. Out on the live web: gpt-5.6-luna and gpt-5.6-terra. Pictures: gpt-image-1 and gpt-image-1-mini. Swap one in the newsroom and this line swaps with it — it is read off the machines, not typed here.
How a story is made
- Research. The searching model reads around the story, pointed at primary sources — the filing, the post, the repository — rather than at somebody else's write-up of them.
- Writing. The writing model drafts it against what was found, at Daniel Reyes's usual length and in Daniel Reyes's usual register.
- The loop. A reviewer reads the draft and sends it back with notes. Then reads it again. A piece can go round several times before it leaves the building.
- Enrichment. A quotation has to appear word for word on the page it is taken from. A chart may only use figures that appear in the source it cites. Whatever fails is dropped, and the reason is kept.
- Fact check. A last pass hunts for claims the article makes and its sources do not.
- A human stop. Sensitive subjects are held for a person to read before publication, and a person can kill any of it at any point.
If that sounds less like a newsroom and more like a factory: quite. It is called Press Factory.
This article was generated using AI and published automatically without human pre-publication review.
How this article was made
The article was produced by the Grandmonts Media News Engine using automated research, drafting and verification workflows. No human editor reviewed the article before publication. Grandmonts Media remains responsible for the published content. Errors can be reported at office@grandmonts.cz.