At DevDay 2026 on 29 September, OpenAI launched Dots — persistent agents built on GPT-6 Astra that keep working after the chat window closes — alongside ChatGPT Space, a shared workspace for humans and agents. The first dot costs nothing extra on Pro and Business Premium plans. It moves ChatGPT from question-and-answer into task delegation.
Key takeaways
- GPT-6 Astra powers Dots, the flagship model announced at DevDay 2026
- The first dot is included at no extra charge on Pro and Business Premium
- ChatGPT Space replaces Library for Pro, Business and Enterprise plans
- Enterprise, Edu and Healthcare get the beta once an administrator enables it
- OpenAI disclosed no intensive-work allowance and no price for additional dots
An agent that outlives the session
Dots differ from earlier agent features on one parameter: lifetime. The agent does not disappear when the tab closes. It watches a project, uses software, reacts to changing data and comes back with finished work for approval.
OpenAI frames this as delegating responsibility rather than asking questions. Each dot?dot: A single persistent agent inside ChatGPT — it has its own name and avatar and runs in the background between sessions. carries its own name and avatar. Access covers the mobile app and the web, with Slack, Microsoft Teams and phone audio next. Specialist dots remain in closed pilots with enterprises.
The diagram shows what OpenAI changed: the agent does not end with the session, and work only returns to the human once finished. The approval layer stays — the company itself notes dots still make mistakes.
ChatGPT Space replaces Library
Space replaces the existing Library for Pro, Business and Enterprise plans. Its core is Pages — documents a human and an agent edit at the same time, in real time. Space also stores files and presentations and connects to Google Drive and Slack. Spreadsheets and slides arrive later.
Simon Willison, live-blogging the conference, described the interface as close to Notion, with a slash-triggered menu.
| Plan | First dot | ChatGPT Space |
|---|---|---|
| Pro | no extra charge | replaces Library |
| Business Premium | no extra charge | replaces Library |
| Enterprise | beta once an admin enables it | replaces Library |
| Edu and Healthcare | beta once an admin enables it | not stated |
What OpenAI did not disclose
The numbers are missing. The company gave no intensive-work allowance, no cost for dots beyond the bundled one, no price for higher speed and no rate card for specialist dots. For a product that runs unattended in the background, those limits decide real usefulness. OpenAI also states dots still make mistakes and advises reviewing consequential work.
Why it matters
Dots shift the centre of gravity from single-answer quality to trust in an unsupervised process. That is a harder product problem than a benchmark — what counts is not average accuracy but the cost of an error found after the fact. With limits and prices undisclosed, companies cannot yet cost such an agent across a team. OpenAI's own warning shows the human approval layer stays in place.
What next?
- Slack, Microsoft Teams and phone audio announced as the next rollout stage for dots
- Spreadsheets and slides are due in ChatGPT Space, with no date given
- Specialist dots stay in closed pilots until a rate card is published
Sources
- VentureBeat — OpenAI launches Dots, always-on AI agent coworkers, and ChatGPT Space where they can collaborate with human teams
- Simon Willison's Weblog — OpenAI DevDay 2026 live blog
