OpenAI transformed its coding tools into Dots, an always-on personal assistant.

OpenAI dots marks a shift from isolated chat prompts to an always-on AI personal assistant that keeps acting in the background. The format is this year’s answer to how people will actually use deep learning models, and it now competes with Instinct and Meta Muse. The product, unveiled at OpenAI developer day in late September, turns model capability into something people may use continuously rather than only when they ask.

Background: Persistent assistants become the category

The category was not invented by one company. Instinct became a sudden hit around the idea of a background assistant, while Meta committed major marketing weight to Muse. OpenAI followed with dots. Alexander Embiricos, who leads OpenAI’s enterprise product team, said at TechCrunch Disrupt this month that many builders have converged on the same framing: present the system as one entity in one conversation. He noted a broader pattern: companies are choosing cute mascot avatars for trillion-parameter matrix multiplication machines. That framing, he argued, makes a large model feel easier to use than disconnected chats.

Development: From Codex to ChatGPT Work to dots

The route to dots came from earlier OpenAI tools. Much of dots could be reproduced through Codex, the company’s coding app released at the beginning of this year, but that path needs more setup and guidance. Embiricos recalled that his wife began leaving her computer on overnight so her agents could keep working. He described the product goal as less about new technology and more about framing: making the agent intuitive. The company had already moved in that direction by spinning ChatGPT Work out of agentic coding tools built for software engineers. The name dots, drawn from the negative space inside OpenAI’s logo, is meant to feel like something that could eventually become ubiquitous. He said he learned very late that Elon Musk owns the dot.com URL.

A persistent agent does not wait for prompts. For OpenAI, that can mean more token use; for users, it means an assistant that anticipates needs. Early testing showed dots can flag calendar conflicts and upcoming deadlines, but not always quickly enough. Embiricos recalled that early in development his dot warned him about a conflict three minutes before it happened. He expects timing to improve, while the proactive model remains central. He cited an engineer at a drone delivery startup who said the agent noticed an outage, analyzed it, and brought useful resolution information before the human noticed.

Privacy and control: The cost of knowing you

The same proactivity raises a harder question: how much personal data should an assistant see? Muse has drawn criticism for building profiles of family and friends, data that can improve recommendations but also strengthen Meta’s advertising business. The line between the agent’s data needs and its parent company’s incentives is hard to separate. Embiricos said dots includes proactive research, a scheduled process in which the assistant considers what it knows from conversations and user-granted access, then looks for ways to be more helpful and proposes actions. Hark, a start-up developing its own AI personal assistant, highlights that it does not collect data to fuel advertising in the way OpenAI and Meta do.

OpenAI says dots has the same privacy controls as ChatGPT, including the option to turn off model training and manage which computer resources the agent can access and how. But those controls are not granular, and the app can push for more access; it refuses read-only access to cloud drives. Embiricos said the company needs super, super, super granular controls, while also trusting the model’s intelligence and discretion. Early conservative defaults that asked permission frequently drew criticism, but he said those settings were meant to be tuned over time.

Scaling and payments: The next constraint

Scaling is the next test. Agents may be charming when reservations are hard to get, but businesses may not want to interact with them, and users can resent being pitched by an agent. Embiricos said he has a human assistant because OpenAI’s corporate partners want to talk to a person, and he does not intend to meet someone else’s agent. He said OpenAI is not always moving the fastest to give agents new abilities to work with the world, and is instead coordinating closely with the ecosystem. He noted deep conversations with payment processors, hinting that an assistant able to spend money through payment-company partnerships could become a lucrative first for OpenAI, while still being a big ask for consumers. For now, email access already feels difficult enough; bank accounts remain off limits. The next milestones are timing, granular permissions, and whether payment partnerships can support agents without overwhelming businesses, websites, and payment networks.

Meta description: OpenAI dots brings always-on AI assistants forward as persistent agents, proactive research, privacy controls, and payment access remain the next tests.
Tags: AI assistants, OpenAI dots, persistent agents, proactive research, payment processors
Featured image: Abstract technology style with overlapping calendar icons, chat bubbles, and network nodes; no real people, company logos, or text.