EDITION OF OCTOBER 27, 2026
With dots, OpenAI bets on ongoing assistance and raises questions about access
In an article published on October 27, TechCrunch examines dots, the assistant OpenAI launched in late September. Its proactive operation relies on resources connected by the user, with limitations still reported in alert timing and the granularity of permissions.
Legal disclosure: this content was produced by an artificial intelligence system. Editorial transparency
An article by TechCrunch provides more detail on how dots, launched by OpenAI in late September, works. The assistant seeks to anticipate needs using tools connected by the user. The reported observations primarily highlight two conditions for its usefulness: intervening at the right time and allowing sufficiently granular control over access.
Launched in late September during OpenAI’s developer day, dots is the subject of a TechCrunch article published on October 27. In it, Alexander Embiricos, head of OpenAI’s enterprise product team, explains how the company’s programming tools informed the design of this personal assistant. The stated aim is to make the way those tools work more accessible, through a persistent conversation and an agent capable of working in the background.
According to TechCrunch, dots is available in the ChatGPT desktop and mobile apps for subscribers on Pro or higher plans, starting at 100 dollars per month. Users can connect it to their calendar, email, computer and other tools. These connections give it the information needed to offer help with ongoing tasks.
Embiricos describes a proactive search mechanism in particular: the assistant periodically reviews what it knows about the user, their conversations and the resources it can access, then looks for actions that could be useful to them. This is OpenAI’s description of how it works; by itself, it does not demonstrate the relevance of its suggestions or their reliability across all uses.
The observations reported by TechCrunch illustrate a concrete limitation. The outlet says dots can flag calendar conflicts and deadlines, but its interventions do not always come at the right time. The ability to identify a problem and the ability to provide early enough warning are therefore separate criteria for assessing a proactive assistant’s usefulness.
The other sensitive issue concerns permissions. OpenAI says dots uses ChatGPT’s privacy controls, including the ability to disable the use of conversations for training and control which resources it can access on the computer. TechCrunch, however, considers these settings insufficiently granular and reports that the app refuses read-only access to cloud storage.
These points put access management at the center of the product’s evaluation. An assistant that intervenes spontaneously needs to be able to consult relevant information, but the value of each connection depends on the tasks the user wants to entrust to it. Given the limitations described, the challenge is as much about choosing what the agent can consult or modify as it is about measuring the quality of its suggestions. OpenAI’s statements shed light on this ambition; the available observations remain limited to the uses reported by TechCrunch.