OpenAI’s dots keep working when you leave the chat

An invoice that almost went unsent

An early tester forgot to invoice a publication. His dot noticed, prepared the invoice and sent it after approval, OpenAI says. The example captures the promise of its new assistants: keeping a task moving even when the person responsible has stopped thinking about it.[1]

Launched on September 29, dots use Astra and have their own cloud computer, browser and connected apps. They can work on several projects, retain context and reach people in ChatGPT, Slack or Teams.[1]

A conventional chat waits for the next question. This arrangement gives an assistant a place to keep working and a route back to its owner when a decision is needed.

The work can continue across apps

OpenAI describes dots investigating bugs, preparing tested fixes and revising launch materials as plans change. Users can inspect the dot’s computer and redirect its work.[1]

The useful result is a project that has advanced by the time someone returns to it. But a prepared invoice and a sent invoice have different consequences. The assistant needs to recognize that boundary even when both actions belong to the same task.

Background proactive research uses read-only tools. Actions that affect accounts or share information go through review against instructions and permission rules; users can add rules for what to allow, block or approve.[2]

Who can try it

Dots are rolling out to Pro and Business Premium users in eligible markets. Enterprise, Edu and Healthcare workspaces can enable the beta through an administrator. The first dot is included, with an allowance for deeper work; tasks it starts in Codex or ChatGPT Work still use those products’ usual limits.[1]

That allowance matters because being available all day and doing substantial work all day are different promises. The launch lets users try a persistent assistant while still putting limits on how much work it can carry out.