Put your dot on call

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Put your dot on call · Channel: OpenAI · 02:04 · Watch on YouTube

Topics: AI agents, Task and workflow automation, AI models

Summary

Ali, an OpenAI engineer, demonstrates how dots can act as persistent cloud agents with access to a local computer and identity across services such as Slack. After connecting a dot to Slack, Ali sets it to monitor a channel, triage incoming pages, and create investigation threads. During a 3 a.m. on-call incident, the dot investigates and creates a Codex thread that Ali can review and steer from a phone or laptop. Ali also suggests using dots to triage feedback or create pull requests through Slack.

Key points

Dots are persistent cloud agents with their own cloud computer and access to a local computer.
A dot can have an identity across services such as Slack.
Ali connects a dot to Slack through the Codex app.
Once connected, the dot can be tagged in Slack threads to pick up work.
Ali sets an automation to monitor a channel, triage incoming pages, and create investigation threads.
During a 3 a.m. incident, the dot investigates and creates a Codex thread before Ali opens the ChatGPT app.
Ali can review and steer the investigation from a phone, with the same Codex thread available on a laptop.
Other suggested uses include triaging channel feedback and creating pull requests through Slack.

Chapters

Start	Chapter	Summary
00:00	Introducing dots and connecting to Slack	Ali describes dots as persistent agents and shows how to connect one to Slack.
00:31	Setting up on-call triage	Ali explains how a dot can be tagged in Slack and sets it to triage incoming pages and create investigation threads.
01:02	Handling a 3 a.m. incident	Ali describes receiving a page and finding that the dot has investigated the issue and created a Codex thread.
01:32	Continuing the investigation and other uses	Ali can continue from a phone or laptop and suggests further Slack automations, including feedback triage and creating pull requests.

Quotes

We built dots to be a persistent cloud agent that has its own cloud computer — 00:00

monitor the channel, and any incoming page I get, please triage the page and create a thread — 00:31

my dot’s already done the investigation, figured out what’s going wrong — 01:02

at 3 a.m. I don’t need to go out of bed and grab my laptop — 01:32

I’m really excited for you to try it out. — 02:04

Concepts and entities

Dots, Codex, Slack, ChatGPT, OpenAI, Ali

Full transcript (original language)

00:00 Hey, I’m Ali. I’m an engineer here at OpenAI, and I’m excited to show you how I’ve been using dots in my engineering workflows. We built dots to be a persistent cloud agent that has its own cloud computer, but also has access to your local computer and has its own identity across all these different services that you already work, such as Slack. To show you what that unlocks, I want to go into my Codex app here, and if I go ahead and click on my dot, I can connect it to Slack. And when I go and add it to Slack here, this actually gives an identity for my dot

00:31 in Slack, so I can tag it in any sort of thread to get it to pick up work that I don’t want to do or anything that I can get it to pick up. And now that it’s connected, I can actually come into the composer and I can do something really cool. So, I’ll show you the workflow that I set up for my on-call shift. I’m going to say, monitor the channel, and any incoming page I get, please triage the page and create a thread

01:02 to investigate. So, I’ve sent this automation now, and I can tell you about how I actually use this to triage an on-call incident that I had at 3 a.m. So, the first thing that happened is I received a page on my phone, so I grabbed my phone, I acknowledged the page, and the next thing I did is I went into the ChatGPT app to see what my dot was working on. And you can see that when I went into the ChatGPT app, my dot’s already done the investigation, figured out what’s going wrong, and it’s also created a Codex thread for me that I can actually click into and start to steer

01:32 the thread and send messages to for me to continue the investigation. And these are all messages that are being sent to my laptop as well. So, it’s running this Codex thread on my laptop, and it’s helped me to figure out what’s going on, but also made it so that at 3 a.m. I don’t need to go out of bed and grab my laptop and start to investigate this incident, because it’s already done the investigation for me. And in this particular case, I can resolve it even from my phone as well, but say there’s an issue

02:04 that requires more investigation, I can put down my phone and I just come to my laptop and I see that same Codex thread in my laptop as well. So, this is one small thing you can do with your dot, but there’s a lot of cool uses you can set up by just sending a simple prompt to set up an automation that, for example, triages all the incoming feedback you get in one specific channel, or you can tag your dot on Slack to create PRs for you and iterate with it directly on Slack. And there’s many more use cases, and I’m really excited for you to try it out.