Your ChatGPT dot can now edit the automations you scheduled, and start your Codex work
A ChatGPT dot can now read and edit the Work automations you have set up, and it can start fresh coding work in Codex or pick up a thread it left running. OpenAI published the change in a release note dated October 5 to 9, 2026 on its ChatGPT documentation site, read at 21:34 UTC on October 10.
OpenAI's own card for its dots documentation, which the October release note links to. Image: OpenAI
The same note moves setup off the desktop. "You can now create your dot in the ChatGPT mobile app," OpenAI says. "Give it a name, customize its appearance, and connect plugins, all from your phone." You can also set ChatGPT to open straight into the dot's conversation rather than a blank chat.
The dot now works inside your Codex threads
OpenAI says a dot "can start work in Codex and follow up on existing threads", search across your ChatGPT conversations, and draw on context from your existing Codex threads and automations. It is "better at deciding when to continue an existing Codex thread and when to start a new one", OpenAI says, "helping keep related work together".
"Your dot can read and edit ChatGPT Work automations, too," OpenAI says, suggesting you ask it to review what is scheduled or change a recurring task. A background agent that can rewrite your scheduled tasks is a different proposition from one that answers questions, and OpenAI gives no list of what a dot may not touch.
Threads a dot manages can now rename themselves as the work changes, with what OpenAI calls "fewer redundant 'created task' messages".
Dots launched in September and the plan rules have not moved
OpenAI introduced dots on September 29 as always-on agents that keep making progress between conversations, powered by GPT-6 Astra and given their own cloud computer. That announcement limited the rollout to eligible Pro and Business Premium users aged 18 and over in supported markets, said Pro access excluded the European Economic Area, Switzerland and the UK at launch, and made Enterprise access an off-by-default beta.
None of that is restated in the October note, which names no plan, no region and no age limit for the mobile build. OpenAI also said in September that dots usage would not count toward Pro, Business and Enterprise plan allowances for a month, with terms for each plan to follow.
The rest of the note is a long list of repairs
Cloud browsing and sidebar loading are faster, and OpenAI says it "fixed an issue that could cause excessive network usage". It stopped sending extra phone notifications to people who were using ChatGPT at the time, fixed character rendering in Safari and Firefox, and fixed Outlook setup getting stuck, adding autofill for email verification codes.
Two of the fixes point at the approval flow that governs what an agent is allowed to do. Safety banners are "more stable on the web", and approval buttons "no longer get clipped or hidden by oversized Gmail previews". An approval button a reader cannot see is an approval a reader cannot give, and a dot acts through the plugins that now claim space in ChatGPT's sidebar. Enterprise users also get fixes for access on the web and when starting a voice call.
OpenAI's evidence for both changes is two screen recordings, one of mobile onboarding and one of a dot running Codex tasks from a phone. There is no accuracy figure, no benchmark and no indication of how often a dot gets a scheduled task wrong.
