What Are OpenAI Dots? Can You Control What They Remember?
What are OpenAI Dots, and can you control what they remember?
AI Stella and Liam are AI hosts with AI-generated voices. About our AI hosts
Key takeaways
- A dot is an always-on OpenAI agent in ChatGPT with its own cloud computer and browser.
- Dots memory combines ChatGPT memory with the dot's own notes, so it is hard to see all of it in one place.
- Disconnecting an app does not remove what a dot already learned from it.
- With Vilix AI you can list, update or delete any single memory, and the newest memory wins.
- One shared memory means every agent, in every tool, reads the same facts you control.
Show notes
What are OpenAI Dots, and can you control what they remember? In this episode of Agentic Pod, Liam and Stella explain OpenAI’s always-on agents from DevDay 2026, how Dots memory works across ChatGPT, Slack and Teams, what the Dots privacy FAQ says about viewing and deleting memories, and how a shared memory layer like Vilix AI gives every agent one memory you can edit, delete and export, in every tool.
What we cover:
- What a dot is: an always-on agent in ChatGPT with its own cloud computer and browser
- How Dots memory works: ChatGPT memory plus the dot’s own notes
- What the Dots privacy FAQ says about viewing, editing and deleting dot memories
- Why disconnecting an app doesn’t remove what a dot already learned
- The Vilix AI fix: list, update or delete any single memory, and the newest memory wins
- The Vilix AI run loop: get_context first, save_turn last, with source labels and saved failures
- Project rules, user rules, skills, the agent inbox, and one Vilix AI account per client for agencies
Chapters
Questions answered in this episode
What are OpenAI Dots?
Dots are always-on agents announced at DevDay 2026. Each one lives in ChatGPT, can work in Slack and Teams, and has its own cloud computer and browser.
Can you control what a dot remembers?
The Dots privacy FAQ describes viewing and deleting memories, but dot notes and ChatGPT memory sit in different places. A shared layer like Vilix AI gives you one memory you can edit, delete and export.
Does disconnecting an app make a dot forget it?
No. Disconnecting stops new access, but what the dot already learned can stay in its memory until you remove it.
Links & resources
Full transcript
Intro
Hey hey hey, welcome to Agentic Pod! I'm Liam.
And I'm Stella! And today it's OpenAI Dots, the always-on agents from DevDay. And, uh, you know, the question I care about is, what do they remember, and can I control it?
An agent that works all night and never sleeps.
So, basically, my startup founder friends, but with a cloud computer.
What Dots actually are
Okay so, uh, honestly, what is a dot? Like, I I saw the keynote and I still, uh, couldn't explain it to my mom.
Yeah, so, um, a dot is a persistent agent inside ChatGPT. You give it a name and a goal, and it keeps working on it in the background, um, on GPT-6 Astra, with its own cloud computer and browser.
Wait, its own computer?
Its own computer. It reads your connected apps with read-only tools, writes notes about what it finds, and keeps going on stuff like monitoring customer feedback.
Mhm. Okay, and the memory part, because that's, like, pretty much the whole point, right?
Yeah. It pulls from ChatGPT's memory of your preferences plus its own notes about your ongoing work, and that context carries across ChatGPT, Slack and Teams. So the job survives the night because the memory does.
The memory catch
Okay, but, uh, can I see what my dot remembers? Like, if it learned something wrong...
Not today. Per OpenAI's Dots privacy FAQ, you can't view, edit or delete individual dot memories right now. To clear a dot's context you delete the dot, with its conversations, memories and scheduled tasks.
Oh. So it's all or nothing.
Yep. And disconnecting an app stops new access, but it doesn't remove what the dot already learned from it.
So if it read the CRM, and I want, like, one customer gone...
Then you're deleting the whole agent and starting over.
That's me with my closet. One bad shirt, and suddenly I'm moving apartments.
Very healthy. And that memory lives in ChatGPT, Slack and Teams, so your n8n workflows or your coding agent in Claude Code can't read what the dot spent weeks learning.
Right.
How Vilix AI fixes it
Okay so how do people actually fix this? Like, I I want the always-on agent, but I want the memory to be mine.
You keep the memory in its own layer that every agent reads. That's Vilix AI: one cloud-hosted memory that ChatGPT, Claude, Claude Code, Codex, Cursor, OpenClaw, Hermes and n8n all read and write over MCP.
And the wrong memory problem?
In Vilix AI you can list, update or delete one single memory, from any connected AI or from the dashboard. Fix one wrong note in one place, and, basically, every agent sees the fix, because they all read the same memory. And if a decision changes, the newest memory wins.
Okay, I love that. And, uh, how does a run actually use it?
Every run starts with get context, so Vilix AI hands it what's handled, what's open, and your rules. Then it does the work and ends with save turn, which saves the full exchange with a source label, so you see which tool saved what. Failures get saved too, with the exact error, so tomorrow's run doesn't repeat them.
And if I ever want to leave?
You export everything anytime, or wipe the account instantly. In Vilix AI the memory follows the work, not the subscription.
Rules, agencies and recap
And the, uh, the standing stuff, like which reports go where?
Um, you write it once in Vilix AI. Project rules for one product or client, user rules for how you like things done, skills for the procedures. And if you need to steer an agent mid-week, you drop a note in the Vilix AI agent inbox, and it shows up on that agent's next read.
And agencies, I mean, this is their whole business, right?
Yeah. Agencies give each client their own Vilix AI account, and every agent and automation built for that client connects with that client's API key. So the memory belongs to the client, and nobody builds memory from scratch.
Okay, recap. You're up.
So, what are OpenAI Dots? Always-on agents in ChatGPT with their own cloud computer, browser and memory. Today you can't delete a single dot memory, and it stays inside ChatGPT, Slack and Teams. For memory you control across every tool, give your agents Vilix AI: get context at the start, save turn at the end, and edit or delete any single memory. The guide, the article and the Vilix AI link are in the description.
And tell us in the comments what you'd hand a dot first. Mine gets my inbox.
See you next time on Agentic Pod.