Many people now move between short questions and longer pieces of work inside AI tools, yet the practical details still depend on which device is open and whether a conversation stays active.
Features that continue on their own have become more common, but setting them up or directing them often required returning to a desktop. OpenAI has been adjusting one of its newer options in that space.
Dots, the always-on agents OpenAI began rolling out in late September, each run on their own cloud computer and can keep progressing on tasks between conversations while connecting to apps through plugins.
The initial details around their launch, including access limited to certain paid plans in eligible markets and the earlier requirement to create them on desktop, are covered in an earlier piece on the rollout.
In an update on October 9, users who already have access can now create a dot directly inside the ChatGPT app on iOS or Android.
The process includes giving the agent a name, choosing its appearance, and connecting plugins from the phone. An option also exists to open the app straight into the dot’s conversation so previous work is ready to resume. Previously, creation was restricted to the desktop app or a desktop browser, while the mobile app only supported interaction with an existing agent.
The same update expands what a dot can do with coding work in Codex.
A dot can now start new work there and follow up on existing threads, drawing context from prior ChatGPT conversations, Codex threads, and automations. It is also better at deciding whether to continue an ongoing thread or begin a fresh one, which helps keep related activity together.
In addition, dots can review and edit scheduled tasks inside ChatGPT Work, including checking what is set to run and updating recurring items when plans change.
Alongside those changes, OpenAI listed several smaller adjustments that users had requested.
Cloud browsing and sidebar loading are faster, with smoother transcript scrolling and fewer pauses while waiting for replies. Extra phone notifications that appeared while the app was already open have been reduced.
Code blocks and attachments display more cleanly, character rendering issues in Safari and Firefox were addressed, and threads managed by a dot can rename themselves as the work shifts.
Outlook setup is simpler, plan guidance is clearer with easier retries for failed loads, and access for Enterprise users is more consistent on the web and when starting voice calls.
Availability remains gradual and limited to eligible Pro and Business Premium accounts in supported markets, with Enterprise workspaces able to enable a beta once an administrator turns it on.
Pro access continues to exclude the European Economic Area, Switzerland, and the United Kingdom at this stage, while Business Premium is listed across supported ChatGPT regions.
The mobile creation option applies to accounts that already have the feature, and further expansion of access or additional dots per user has been described as planned for later.
These adjustments add concrete options for starting and directing the agents from a phone while tightening a few existing workflows.
