OpenAI launches Dots, always-on agents with their own cloud computer

OpenAI is shifting ChatGPT from answering to acting, and it is taking Meta's Muse head-on to do it.

AINEWsLoot editorial teamSeptember 30, 20262 min readRadarRadar 89

OpenAI's Dots run on their own cloud computer, reach into 4,000 apps and keep working in the background when nobody is asking them anything. Several outlets reported the launch on September 29, 2026. The pitch is simple: the agent acts before you ask.

Was bekannt ist

Dots are built to stay on. The Decoder calls them always-on agents. They do not wait for a prompt. They keep going.

Each agent gets its own machine in the cloud. From there it can use 4,000 apps. That lets a Dot finish tasks across several services while the user's own device is off.

TechCrunch describes Dots as an agentic avatar. OpenAI gives its agents a visible character instead of just a text box.

The Verge and The Decoder both frame Dots as a rival to Meta's Muse. The two companies are chasing the same idea: an assistant that runs alongside you all day.

A side story followed the launch. According to TechCrunch, the internet is convinced that Elon Musk's xAI trolled the Dots release.

ChatGPT as a work platform

Dots arrived alongside a broader rebuild. The Decoder reports that OpenAI is turning ChatGPT from a chatbot into a work platform that looks more like an operating system. Engadget covers a dedicated space for work inside ChatGPT. TechCrunch says the result feels a lot like an office suite and puts OpenAI in direct competition with Microsoft. The Verge has a roundup of the DevDay 2026 announcements.

Einordnung

Chatbots have followed one pattern so far. You ask, the model answers, then it stops. Dots flip that. The agent moves on its own, with its own compute and access to thousands of services.

That is where the launch collides with the safety warnings about the model behind these agents. A system that only answers can be checked one reply at a time. An agent that can act across 4,000 apps needs hard limits on what it may do without asking first. More access means a bigger blast radius when something goes wrong.

For users, the trade is clear. Hand a task to an agent and you save time, but you also give up control. The race with Meta shows that the largest AI companies are making this move at the same time. And by turning ChatGPT into a work platform, OpenAI is going after office work, long Microsoft's home turf.

Was offen ist

The reports do not spell out pricing, regional availability or which model powers Dots. It is also unclear how users can restrict an agent's permissions and how OpenAI is responding to the safety warnings. Whether xAI deliberately mocked the launch is not confirmed; TechCrunch describes what people online believe.

Sources

More articles

+new~confirmed?RadarRadar 0–100 · arrow: change since yesterday
18:57?

Anthropic's IPO prospectus warns its AI could pose existential risks

A company asking investors for money states in a legal filing that its core product could threaten humanity, while reporting a multibillion-dollar loss.

Anthropic's IPO prospectus warns that its AI could pose existential risks to humanity, while the company reports an 8 billion dollar loss.
ArticleENRadar 81Radar
07:31?

AMD to buy Fei-Fei Li's World Labs for $8.2 billion

The chipmaker is paying for AI that builds spatial worlds, not another chatbot, and that choice says where AMD wants to compete.

AMD will acquire World Labs, the AI startup founded by Fei-Fei Li, for $8.2 billion.
ArticleENRadar 68Radar

Related

All reels