OpenAI is pushing AI assistants beyond the familiar chatbot model with dots, a new class of always-on AI assistants powered by GPT-6 Astra.

The idea is straightforward but potentially significant: instead of opening an AI app whenever you need something, dots are designed to stay around and handle ongoing work for you. OpenAI is positioning them as assistants that can understand context, take action, and keep working across tasks rather than simply answering individual prompts.

That puts dots in the same broader wave of AI products moving toward persistent, agentic assistants. Meta recently introduced Muse for small businesses, while OpenAI has been expanding ChatGPT Work and its own scheduled and background capabilities. The difference with dots is that OpenAI is presenting the assistants themselves as something that can remain continuously useful rather than as a conventional chatbot interface.

At the center of dots is GPT-6 Astra, OpenAI’s model built for complex, multi-step work. OpenAI introduced Astra earlier this month with improvements across coding, research, computer use and professional workflows. The company says Astra can work through complicated tasks, adapt when requirements change, and create documents, spreadsheets and presentations while following templates and instructions.

From answering questions to getting things done

The important part of dots isn’t simply that they use a newer model. It’s the way OpenAI wants the assistants to operate.

Traditional AI assistants are largely reactive. You ask a question, give them an instruction, and wait for the response. An always-on assistant is supposed to have a longer-running understanding of what you’re doing and take care of work that doesn’t necessarily fit into a single prompt.

That could mean keeping track of an ongoing project, handling recurring tasks, preparing information, or working across multiple steps without requiring the user to spell out every part of the process.

OpenAI has already been moving ChatGPT in this direction. ChatGPT Work, introduced earlier this year, can research and analyze information, work with connected apps and files, create finished documents and presentations, and continue work through scheduled tasks. Those tasks can run once, repeat on a schedule, or monitor for changes.

Dots appears to take that philosophy and package it around persistent assistants.

That distinction matters because the real promise of AI agents isn’t necessarily getting a better answer to a question. It’s reducing the number of times you have to ask in the first place.

GPT-6 Astra is doing the heavy lifting

Astra is particularly relevant here because OpenAI built it around tasks that require more than generating text.

OpenAI describes GPT-6 Astra as an upgrade for coding, research, computer use and complex multi-step work. It can also adapt when users introduce new requirements or change direction midway through a task.

That makes it a natural foundation for an always-on assistant.

Instead of treating every interaction as an isolated conversation, an assistant can use Astra’s ability to reason through a workflow and operate across different steps. The goal is less “ask AI something” and more “give AI something to take care of.”

That’s also where computer use becomes important. OpenAI says Astra is designed to work with computers and professional software, expanding the kinds of tasks an agent can perform beyond generating content inside a chat window.

The company’s recent ChatGPT updates show the same direction. Astra can now be used in ChatGPT Voice when harder questions require additional search or reasoning, while GPT-6 models have also been added to ChatGPT Work and Codex.

The always-on AI race is heating up

Dots arrives as AI companies increasingly compete over what comes after the chatbot.

Meta recently expanded its Muse AI agent to automate everyday tasks for small businesses. OpenAI’s own ChatGPT Work is built around longer-running jobs, connected apps and scheduled work. Other companies are also exploring assistants that maintain context and operate more proactively.

The appeal is obvious. Most useful work isn’t a single question with a single answer. It’s a sequence of small decisions, follow-ups, searches, edits and actions.

An assistant that can remember where a task stands and continue from there could therefore be considerably more useful than one that requires a fresh prompt every time.

But the “always-on” part also raises a different set of questions.

An assistant that is continuously involved in your work potentially has access to much more context than a conventional chatbot. That makes permissions, privacy, transparency and user control particularly important. OpenAI has already added additional safety monitoring to GPT-6 Astra to detect situations where an agent may not have correctly interpreted an instruction, with conversations potentially being paused or stopped for user review.

Those safeguards become increasingly important as assistants move from suggesting actions to actually carrying them out.

Dots could be the next step for AI assistants

The interesting thing about dots is that OpenAI isn’t really selling the idea of another chatbot.

It’s selling the idea of an AI presence that can remain useful between prompts.

That is a subtle but important change in how we interact with AI. The long-term goal isn’t necessarily to make people better at writing prompts. It is to make prompts less necessary for routine work.

GPT-6 Astra gives OpenAI a model designed for the complicated reasoning and computer interaction that this approach requires. Dots, meanwhile, provide a product concept built around keeping that intelligence available as an ongoing assistant.

Whether people actually want AI to be this persistent will depend on how well these systems balance usefulness with control. But the direction is becoming increasingly clear: the next generation of AI assistants won’t just wait for you to ask what to do. They’ll be built to keep track of what needs doing in the first place.