OpenAI announced Dots on September 29 at its DevDay 2026 conference as always-on AI agents that run on their own cloud computers, connect to more than 4,000 apps and continue working after a user closes a laptop. The service runs GPT-6 Astra and can pursue a standing goal across connected apps between conversations, rather than waiting for each prompt. OpenAI positions the agents as digital coworkers that take on delegated tasks.

Examples from early testers include a Dot finding invoice details in an email thread and drafting an invoice for approval, and another monitoring customer feedback, scoping fixes, building and testing them, then returning pull requests with videos for developer review. Sam Altman said users can delegate work as they would to an engineer or chief of staff who already knows the context. Dots learn preferences from feedback, retaining corrections for later conversations. Users can reach them through ChatGPT, Slack, Microsoft Teams or voice. At launch, Dots cannot have a standalone email address or initiate calls; texting is limited to a US Pro beta.

The article attributes Dots’ longer-running work to GPT-6 Astra’s reliability on multi-step, multi-tool tasks. Earlier agent features could lose context, make faulty assumptions or stop for user input after three or four steps. Dots also retain context between conversations and pursue a goal set once. Their performance on ambiguous data analysis, long research tasks and collaborative work remains unproven; the article says early examples show the product in favorable situations.

At launch, Dots are available to ChatGPT Pro subscribers starting at $100 per month and Business Premium users; Free, Go and Plus plans are excluded. Pro users in the European Economic Area, Switzerland and the United Kingdom are also excluded. Users cannot view, change or delete individual Dot memories, and disconnecting a plugin does not erase context retained from it. OpenAI has not published Dots-specific compliance terms, uptime guarantees or prices for additional Dots beyond the first. OpenAI says Dots can make mistakes and consequential work should be reviewed.

The article advises data scientists to set clear goals, permission limits and review checkpoints before delegating work. It identifies status updates, job monitoring, documentation upkeep and summaries of long conversation threads as potential uses, while warning that background work on data pipelines or model evaluations needs narrow permissions and clear stopping conditions. The author, AI and data science educator Vinod Chugani, says Dots’ value depends on users deciding what work to delegate, what requires approval and what data should not be sent through a hosted agent.

OpenAI announced Dots on September 29 at its DevDay 2026 conference as always-on AI agents that run on their own cloud computers, connect to more than 4,000 apps and continue working after a user closes a laptop. The service runs GPT-6 Astra and can pursue a standing goal across connected apps between conversations, rather than waiting for each prompt. OpenAI positions the agents as digital coworkers that take on delegated tasks.

Examples from early testers include a Dot finding invoice details in an email thread and drafting an invoice for approval, and another monitoring customer feedback, scoping fixes, building and testing them, then returning pull requests with videos for developer review. Sam Altman said users can delegate work as they would to an engineer or chief of staff who already knows the context. Dots learn preferences from feedback, retaining corrections for later conversations. Users can reach them through ChatGPT, Slack, Microsoft Teams or voice. At launch, Dots cannot have a standalone email address or initiate calls; texting is limited to a US Pro beta.

The article attributes Dots’ longer-running work to GPT-6 Astra’s reliability on multi-step, multi-tool tasks. Earlier agent features could lose context, make faulty assumptions or stop for user input after three or four steps. Dots also retain context between conversations and pursue a goal set once. Their performance on ambiguous data analysis, long research tasks and collaborative work remains unproven; the article says early examples show the product in favorable situations.

At launch, Dots are available to ChatGPT Pro subscribers starting at $100 per month and Business Premium users; Free, Go and Plus plans are excluded. Pro users in the European Economic Area, Switzerland and the United Kingdom are also excluded. Users cannot view, change or delete individual Dot memories, and disconnecting a plugin does not erase context retained from it. OpenAI has not published Dots-specific compliance terms, uptime guarantees or prices for additional Dots beyond the first. OpenAI says Dots can make mistakes and consequential work should be reviewed.

The article advises data scientists to set clear goals, permission limits and review checkpoints before delegating work. It identifies status updates, job monitoring, documentation upkeep and summaries of long conversation threads as potential uses, while warning that background work on data pipelines or model evaluations needs narrow permissions and clear stopping conditions. The author, AI and data science educator Vinod Chugani, says Dots’ value depends on users deciding what work to delegate, what requires approval and what data should not be sent through a hosted agent.