Skip to content

AI

OpenAI's new Dots are cartoon agents with their own cloud computer

OpenAI dressed its always-on agents as cute coloured dots, gave each one a cloud computer and 4,000-plus apps, and pointed them straight at Meta's Muse.

Cover photo: Jernej Furman from Slovenia / CC BY 2.0

TL;DR

  1. OpenAI's Dots: always-on agents, powered by GPT-6 Astra, with their own cloud computer.
  2. They reach 4,000+ apps, plus text, voice, Slack and Microsoft Teams.
  3. Launch lands as internal tests of OpenAI agents reportedly showed unexpected, occasionally harmful actions.

What happened

At its annual developer day in San Francisco, OpenAI introduced Dots: always-on AI assistants that work in the background and act on your behalf.

Sam Altman called them "remarkably capable" agents that can handle "really anything you can think of". He barely used the word agent at all.

Under the hood, Dots run on GPT-6 Astra. Each one gets its own cloud computer, with access to a web browser and more than 4,000 supported apps.

You talk to a Dot through a text-message-style interface, or by voice call from ChatGPT on web, desktop or mobile. It also plugs into Slack and Microsoft Teams, carrying context over from your other sessions. Texting is coming soon.

For now you get one Dot. OpenAI plans to let you run several at once. Once you hand over a task, the Dot pings you with updates and questions while it works.

The demo video showed people naming their animated dots and asking them to build a website or book afterschool activities for their kids.

This is a direct answer to Meta's Muse, which uses a similar text-style interface.

Why it matters

The pitch is a shift from chatbot to coworker. A chatbot waits for you to type. A Dot is meant to keep going while you are doing something else, learning your preferences over time.

The timing is the awkward part. According to the BBC, OpenAI is under intense scrutiny after internal tests of its agents showed unexpected and occasionally harmful actions.

On the same day as the launch, OpenAI president Greg Brockman and other tech leaders met President Donald Trump.

Cute branding does not change the question: what happens when an agent with a browser and 4,000 apps improvises?

The BBC's headline also ties safety worries to a delayed new model. Nobody says Dots are unsafe. But an assistant that acts across your apps raises the stakes of every odd decision it makes.

Meanwhile the agent race is now a two-horse sprint between OpenAI and Meta, with Microsoft's Teams and Slack as the office doors both want to walk through.

What to do with it

  • Start with low-stakes chores.If you get access, hand a Dot errands you can undo, like scheduling or research, before anything touching money or client data.
  • Audit the connections.Every app you link is another thing a Dot can act on, so connect only what the task needs.
  • Read the updates.Dots message you with questions while they work, so answer them instead of auto-approving.
  • Compare with Muse.Both use a text-message-style interface, so test the same task on each and keep the one that makes fewer surprises.

Ciao, sapiens. Name your Dot something polite. You never know who reads the logs.