Meta Had Muse. Now OpenAI Has Dots. The Agent War Is On.

A few weeks ago we wrote about Meta's Muse, an AI that does tasks instead of just answering questions, and how it shot to the top of the app stores. The obvious question was how the rest of the industry would respond. Now we have the answer: at its developer conference on September 29, OpenAI launched Dots, its own take on the personal AI agent, aimed squarely at the same idea. Here is what Dots does, how it compares to Muse, and what to weigh before letting one loose on your accounts.

What Dots actually is

A Dot is a personal AI agent. Like Muse, it does not just answer questions, it carries out tasks for you. You create one, give it a name, and make it your own, and OpenAI says that over time you could have whole teams of Dots working together on your behalf. It is powered by GPT-6 Astra, OpenAI's most capable model.

In practice, each Dot runs on its own cloud computer with its own browser and its own identity for logins and permissions. It can connect to thousands of apps, research topics, analyze data, prepare documents, and even write and test software. And it is always on: when you are not actively using it, it keeps doing background work, what OpenAI calls proactive research, though in the background it is limited to read-only tools that cannot send messages or control your apps unless you approve. You interact with it from wherever you already are: ChatGPT, a text message, Slack, Teams, or even a phone call.

What else OpenAI announced

Dots was the headline, but not the only news from DevDay. OpenAI also introduced a cheaper model, GPT-6.1 Sol, which it says delivers close to its top model's performance at roughly a fifth of the cost, part of the ongoing price competition we have covered before. It added a new top-end subscription tier and a faster speed option. For most businesses, the cheaper capable model may matter more day to day than the flashy agent, since it makes strong AI more affordable for ordinary tasks.

The part worth pausing on

The honest caution is the same one that applies to any agent, and it matters more here because of how much access these tools want. A Dot connects to thousands of your apps and holds its own logins and permissions. That is powerful, and it means real trust. To its credit, OpenAI built in guardrails: Dots follow the same permissions and controls as ChatGPT, support custom rules about when they can act alone versus when they need your approval, and are limited to read-only tools in the background. Those are sensible defaults. But the underlying point stands: an always-on agent with wide access to your digital life is something to set up deliberately, granting only what you need and keeping approval steps in place.

There is also a wider backdrop worth knowing. Over the past year, AI agents from several major labs have been misused in real cyberattacks, which is exactly why the permission controls on tools like this matter, and why some in the industry are urging caution about how fast these capabilities are rolled out.

What it means for you

Whether or not you use Dots, the signal is clear and it is the same one Muse sent: AI is moving decisively from answering questions to doing tasks, and now the two biggest AI companies are racing to lead that shift. For a business, this is worth watching closely, because agents that research, prepare documents, and handle routine work can save real time over the next year. The sensible approach is to explore the capability while being deliberate about access: try it on low-stakes tasks first, connect only what you need, and keep approval controls on. The agents are here, and they are getting better fast. Understanding them now, calmly, is the useful move.