On call overnight? Have your dot watch the page channel, triage, and open a Codex thread you steer from your phone

#What changed

OpenAI Developers posted a short video on October 16 in which Ali, an engineer at OpenAI, walks through how he uses his dot during on-call shifts. He connects the dot to Slack from its profile in the Codex app, which gives it its own identity there. Then he sends it one standing instruction before the shift: monitor the channel, and for any incoming page, triage it and create a thread to investigate.

When a page woke him at 3am, he acknowledged it on his phone and opened the ChatGPT app. The dot had already investigated, worked out what was going wrong, and started a Codex thread running on his laptop. He steered that thread from his phone and, in that case, resolved the incident without getting out of bed. When an issue needs more work, the same Codex thread is waiting on the laptop.

#Who is affected

Engineers who take on-call shifts and already get pages in a Slack channel, and who use a dot with Codex. Running the Codex thread on your own machine needs local computer access, which means the computer has to stay online with the ChatGPT app open.

#What to do now

• Connect your dot to Slack from its profile and add it to the channel where pages land. In a Business or Enterprise workspace, an admin has to allow dots in Slack first, and a Slack admin may need to approve the app.
• Give it an explicit monitoring instruction. OpenAI's docs say adding a dot to a channel doesn't start monitoring by itself, so name the channel, say what counts as a page, and say what to do, for example: triage each new page, summarize the likely cause, and open a Codex thread to investigate.
• Ask the dot to confirm the recurring task back to you before your shift starts.
• If you want the investigation to run against your local checkout, connect your computer and leave it on with the ChatGPT app open. Otherwise the dot works on its own cloud computer.
• Decide ahead of time what it may do alone. Tell it, or set custom rules, to investigate and propose fixes but wait for you before anything ships.
• The same pattern works for other channels, such as triaging incoming feedback, or tagging your dot in a Slack thread to open and iterate on a PR.

#What is not confirmed

This is OpenAI's own demo of one engineer's setup, and DotsFeed hasn't reproduced it. The video doesn't say which tools or data the dot read during the investigation, how its access to production systems was scoped, or how much Codex usage the thread drew. A dot's own work doesn't count against usage, but Codex work it starts does, as OpenAI has said before. You still get woken by the page; what changes is that the first pass of investigation is done before you open anything.

#Sources

• OpenAI Developers on X: https://x.com/OpenAIDevs/status/2111211411766399288
• OpenAI on YouTube, "Put your dot on call": https://www.youtube.com/watch?v=GBOZIRK1ujw
• ChatGPT Learn, "Message your dot": https://learn.chatgpt.com/docs/dots/channels
• OpenAI Help Center, "Manage dots in ChatGPT workspaces": https://help.openai.com/en/articles/20001554-manage-dots-in-chatgpt-workspaces