OpenAI just launched its answer to Meta: 'always-on' AI agents called Dots

• OpenAI announced its new "always-on" AI agent product, Dots, at its DevDay conference on Tuesday.
• Dots will compete with Meta's Muse, part of a rush in the industry toward personal agents.
• A spate of other announcements included a new model, a $500 plan, and a collaboration feature.

OpenAI is throwing its hat in tech's shiniest new ring.

On Tuesday, the company unveiled a new product called Dots, its answer to the rise of "always-on" personal agents like Instinct and Meta's Muse. OpenAI's version, announced at the company's annual DevDay conference in San Francisco, will run on its frontier model GPT-6 Astra.

The releases of Muse and Instinct became a viral moment in Silicon Valley. OpenAI, without its own product, initially seemed to be slipping behind. Now, Dots will almost certainly be one of the industry's most capable agents.

The model it's built on is OpenAI's best yet at computer use, or moving around apps and websites like a human. The agent runs on a cloud computer and can use some tools even when a user isn't actively delegating a task.

In a blog post, OpenAI said Dots — dot, if you just use one — are like having "intelligence is always working by your side." It cited the example of a Dot investigating a bug as soon as it was flagged.

Dots require a trade-off. Like Muse and Instinct, the more access you give it to your computer, apps, and messaging platforms, the more it will be capable of handling.

For many people, Meta may still be the easier choice. Muse is free, while Dots will roll out first only to Enterprise and Business Premium users and to ChatGPT Pro subscribers. OpenAI plans to expand the offering.

If OpenAI can afford the compute, it can roll out the new product to a huge user base. The AI lab gave a little toot of the horn on Tuesday, saying more than 1.2 billion people now use ChatGPT each week.

Before this summer, the top name in the personal AI agent race was OpenClaw, which launched in late 2025. OpenAI hired its creator, Peter Steinberger, in February.

A slew of other releases make OpenAI feel like Apple

The company made a spate of announcements on Tuesday. It unveiled ChatGPT Space, a new page on the app where coworkers can collaborate. It launched a new $500-a-month Pro plan, which grants access to the company's fastest modes. It also announced a new "OpenAI Marketplace" where the company lets customers direct their committed OpenAI spend toward partner companies.

And then there's the company's bread and butter: a new AI model. A day after announcing that it had shelved one of its new models, OpenAI revealed GPT-6.1 Sol, less capable than its Astra model but far cheaper.

When giant tech companies decide to build certain products, they can easily wipe away an entire startup's potential. While OpenAI isn't at the scale of Apple or Google, several Tuesday releases will set off alarms.

The new AI model Jev caught Silicon Valley's eye this month for its ability to make lightning-fast decisions. Meanwhile, OpenAI announced a new "Decisions API" to serve the same market.

A whole slate of startups peddle AI for slide decks; OpenAI also announced a collaborative version that allows exports to PowerPoint and Google Sheets.

The same goes for AI-powered notetaking. OpenAI now has its own version: a meeting plugin that takes notes, makes summaries, and then deletes the audio.

All told, OpenAI touts "more than 20 major announcements across ChatGPT, Codex, our models, and entirely new forms of working with AI."

Have a tip? Contact this reporter via email at [email protected], or over text, Signal, Telegram, or WhatsApp at 415-757-8198. Use a personal email address, a nonwork WiFi network, and a nonwork device; here's our guide to sharing information securely.