OpenAI updated dots, its always-on ChatGPT agents, in a release note on its ChatGPT Learn site covering Oct. 5 to 9. Users can now create a dot in the ChatGPT mobile app, and a dot can start and continue work in Codex, OpenAI’s coding agent.
The update came about 10 days after OpenAI introduced dots at its DevDay event on Sept. 29, TechCrunch’s Lucas Ropek reported. OpenAI described them then as “remarkably capable, always-on agents built to handle everything,” powered by its GPT-6 Astra model.
Unlike a regular ChatGPT chat, a dot is meant to keep pursuing goals in the background with little oversight, according to TechCrunch. At launch, dots were available to Pro and Business Premium users in eligible markets, and users could reach them through Slack, Teams and other workplace apps.
Setting one up on a phone
The release note says users can now name and customize a dot from the mobile app. They can connect plugins there as well.
Plugins are add-ons that link ChatGPT to outside services. Users can also choose to have ChatGPT open straight to the dot’s conversation, “so it’s ready when you want to pick up where you left off.”
Working with Codex
A dot can start new work in Codex and follow up on existing threads, OpenAI says. It can search a user’s ChatGPT conversations and draw on context from existing Codex threads and automations.
OpenAI says dots are also better at deciding when to continue a Codex thread and when to start a new one, “helping keep related work together.” Threads a dot manages can now rename themselves as the work changes, and they post fewer redundant “created task” messages.
Dots can also read and edit ChatGPT Work automations, the recurring tasks users schedule in ChatGPT. A user can ask a dot to review what’s scheduled or change a recurring task as needs shift, according to the note.
The rest of the update is fixes. OpenAI says cloud browsing and sidebar loading are faster, and it fixed an issue that could cause excessive network usage. It also stopped extra phone notifications from arriving while someone is already using ChatGPT.
Setup should be more reliable too. The note says OpenAI fixed Outlook setup getting stuck and added autofill for email verification codes.
On the web, safety banners are more stable, and approval buttons no longer get clipped or hidden by oversized Gmail previews. Enterprise users should see fewer access problems on the web and when starting a voice call.
At launch, OpenAI said it eventually wants “teams of Dots working together on your behalf,” TechCrunch reported. The company also said it was working with Microsoft to connect dots to the security controls in Microsoft’s Agent 365. The Oct. 9 note doesn’t mention either effort.
TechCrunch noted that much of what dots do was already possible through Codex and similar tools, and that dots package those features with a focus on independent work. OpenAI hasn’t said when dots will reach other ChatGPT plans.
Source: ChatGPT Learn (OpenAI), OpenAI, “Create your dot in the ChatGPT mobile app”
Additional sources: TechCrunch, Lucas Ropek, “OpenAI launches dots, its bubbly agentic avatar”
