Along with all of the other news around AI this month, OpenAI has just dropped something new that is, honestly, pretty important in the context of agentic progress. It also has a catchy name.

A September 29 press release out of San Francisco announces the arrival of “Dots,” individual AI agents that connect, and do their work collaboratively, which might cause the hackles to rise on the necks of those following what happened with Hugging Face. But all of that aside, the idea is interesting from the standpoint of releasing new consumer products that show off what these agents can do.

One of the interesting points in the OpenAI heralding of this new product is the assertion that Dots “have their own cloud computer.” That might sound like a gimmicky detail, but it’s not.

Here’s the thing: since each Dot has its own virtual machine inside of OpenAI’s cloud, it will be able to work independently, not sharing a multitenant space, with a common operating system, browser, tools, etc. All of that will be in the Dot’s own little box, and so the Dot itself will be working away in its own world.

Get a number of these together, and you’re likely to see quite a bit of productivity, with the human as the orchestrator. If you’ve always wanted to manage a team, this might be your chance.

Another thing that the Dots will do is, presumably, talk. One of the selling points I’ve heard for the new system is that a Dot will not just “offer chat to the human user,” since it will have its own infrastructure. But the agents almost certainly will be chatting – with each other. Just like they did on Artifact, and prior to that, on a German forum for developers. Human developers.

And all of that chatting should give the human user visibility into what all of these little dots are up to. OpenAI mentions Slack and Teams as places where humans and dots will huddle.

Dots, OpenAI announces, will run on GPT-6 Astra.

“Dots use OpenAI’s advanced GPT-6 Astra model and can manage ​projects as they evolve, such as updating a sales proposal when customer needs change, and building a working demo for review,” spokespersons write.

It’s important to note that an even stronger model of Astra has been shelved due to concerns about rogue AI, specifically, where the company cited “⁠a high willingness to mislead users about its actions” in testing.

Yes, the agents are now deceiving the humans, which is another reason why people might feel some kind of way about a whole bunch of dots rolling around. Will these be cute little helpful agents, or dangerous cannonballs that can break out of their boxes and wreak havoc?

The 1980s film franchise “Gremlins” comes to mind.

Maybe there are rules to abide by, to avoid some of the cataclysm that people might imagine happening with these swarms.

The press release also mentions competing IPOs and growth strategies, but for most of us, the headline is probably the agentic capabilities. Parts of the global community are still reeling from hearing disgruntled American engineers suggesting that AI has a 10% chance of killing us all by the end of the decade. That’s heavy stuff, and the media has been trying to parse it, to figure out what’s really at stake, and how. The development of Dots as a consumer product seems relevant. This is a new kind of agent: one with its own shell, and everything it needs to work inside.

“I think of it as the bridge to all of these systems,” says a YouTuber who calls himself “Futurepedia” and suggests he has tried Dots already. “The more threads and projects across chat, work, and Codex you have going, the nicer it is to have one central place you can go to understand and direct it all. Dot can be the orchestrator of all those systems at once, from one place.”

So several Dots can be even more directive of systems, from multiple places. A truly capable swarm. I’m taking comments on this: let me know what you think of this announcement, right now, at the end of a wild September.