OpenAI Dot Just Entered the Agent Race
How Dot stacks up, what it costs, and the work people are already giving it.
Hey smarties 👋
The most appealing thing about Dot is the idea of coming back to a project and finding that someone kept it moving while you were away. That’s the promise OpenAI is making. This week, we’re looking at what Dot can do, how it compares with Grok Bot and Muse, what it costs, and how people are using it so far.
Sponsored
Stop betting on one AI answer
Sponsored
According to Artificial Analysis' AA-Omniscience benchmark, the best AI model is right only 67% of the time. And it sounds confident even when it's wrong.
Cuey is a free Chrome extension that gives you a second opinion, so you don't get burned by the first one.
Cuey lives inside ChatGPT, Claude, Gemini, Grok, and more.
One prompt, three answers, side by side. No tab-switching, no copy-paste.
See where the models disagree before a hallucination costs you.
Carry your context across 30+ models, so you never start from scratch.
01 / The big picture
What Is OpenAI Dot?
Your Dot can still be working after you close your laptop.
It is OpenAI’s personal agent inside ChatGPT, powered by GPT-6 Astra and equipped with its own cloud computer and browser. You can give it an assignment, carry on with your day, and return to the work without explaining everything again. Meet Dot
What interests me is the follow-through. A project rarely ends with one document. Someone changes a requirement, new information arrives, and the work needs updating. Dot can keep track of those changes, continue the assignment, and bring you decisions that need your attention.
It can also delegate tasks to Work or Codex while you continue talking to it. If you already use those tools, having one place to coordinate the work is a useful addition.
02 / THE COMPARISON
Dot vs Grok Bot vs Meta Muse
All three promise help with work that takes more than one sitting. The useful difference is how you work with them, and how well they fit the tools you already use.
What you’re comparing | OpenAI Dot | Grok Bot | Meta Muse |
|---|---|---|---|
Interface | Your agent inside ChatGPT, with messages and voice calls | A roster of Bots with different responsibilities | A dedicated app with task management and personalization |
Models | GPT-6 Astra | Automatic model selection managed by Cursor, without a user-facing model picker | Meta’s Muse Spark |
Working style | One main contact that can delegate work | Several Bots working separately or together in a shared thread | A personal agent working across connected services |
What stands out | Work, Codex, Slack and Teams | Signed-in websites, apps and reusable routines | WhatsApp, Facebook and Instagram business tools, plus other connectors |
In his hands-on comparison, Nate Herk found Muse easier to navigate, but preferred Dot for his own Codex-based workflow. That is a useful distinction: a friendlier interface and a better fit for your existing work are not always the same thing.
Grok Bot offers another approach if you would rather give separate responsibilities to several agents instead of managing everything through one.
03 / Access & pricing
What You Get With Each Plan
Muse is the easiest to try without paying. Dot has the highest individual starting price, although you may already have access through an eligible ChatGPT subscription.
Agent | Individual pricing | What is included |
|---|---|---|
OpenAI Dot | ChatGPT Pro at $100, $200 or $500/month | Dot access, with usage allowances depending on your plan |
Grok Bot | Cursor Pro from $20/month or SuperGrok from $30/month | Bot access; Cursor Pro includes a weekly Bot allowance |
Meta Muse | Free, Power at $20/month, or Maximum at $100/month | Limited free usage; Power includes 500 million Muse tokens weekly and Maximum includes 3 billion |
Note: Dot is rolling out gradually to Pro users aged 18+ outside the EEA, UK and Switzerland, and to Business Premium and Enterprise worldwide. Work and Codex tasks still use their respective allowances. Muse is available only in selected countries, so check availability in your region before subscribing.
04 / GETTING STARTED
Set Up and Customize Your Dot
You do not need to build an automation from scratch. The useful setup work is deciding what your Dot should know, what it can access, and when it should ask you first.
Create your Dot. Open ChatGPT in the desktop app or a desktop browser. Select Dots when it appears on your account and follow the introduction.
Give it a name and a look. Choose its shape, color, eyes and accessories. You can change these later.
Connect the essentials. Add the email, calendar or file services it needs for your first task. You can skip connections and return to them later. Setup guide
Explain how you work. Share your role, current priorities and preferences, including how much detail you want in updates. Correct outdated context rather than assuming it already knows everything. Memory and context
Set approval boundaries. Under Settings → Personalization → Custom rules, specify actions that need your approval. For a first task, I would keep sending messages, publishing and spending behind a review. Permission controls
Connecting your personal computer is optional. Its local files and apps are available only while the computer is online with ChatGPT open; Dot’s own cloud computer works separately. Computer access
05 / DOT IN ACTION
What People Are Creating With Claude
Example 01
Running a Marketing Team From Slack
Grace Leung messages her Dot through Slack to coordinate a marketing campaign. It delegates work to specialist agents, gathers the posts and visuals, and adds the content to her Notion calendar.
Example 02
Testing an AI-Powered Creative Agency
Ishan Sharma connects Dot to Artlist to create promotional videos for cafés, then tests outreach to potential clients. The video shows the creative and sales workflow, though paying customers haven’t been established.
Example 03
Preparing a Video’s Script, Graphics, and Thumbnail
Ryan Sael asked Opus 5.5 to explain camera focus by building an interactive lens lab. Turn the focus ring, and you can see the glass elements move and watch which part of the scene becomes sharp. It turns a technical explanation into something you can explore yourself. Ryan reports that the build took 1 hour and 26 minutes in a single run, costing $25.66 through the API.
Example 04
Catching Missed Invoices and Billing Gaps
Peter Yang asks Dot to review his sponsorship commitments. It flags a potentially uninvoiced deal and a $10,000 billing discrepancy for him to investigate.
06 / BEFORE YOU RELY ON IT
What Dot Still Needs From You
The videos make delegation look easy, but the setup behind them matters. Grace’s marketing team, for example, already had brand information, reusable skills, and defined roles to work with.
Give it context. Share the relevant files, preferences, and a clear description of the result you want.
Check the finished work. A progress update doesn’t mean the task succeeded. Open the deliverable and check that it meets your request.
Keep important decisions yours. Require approval before publishing, contacting clients, or spending money.
Start with one responsibility you know well. You’ll be able to spot mistakes, judge whether Dot actually saves you effort, and decide what to hand over next.
07 / READER POLL
Last Edition’s Poll Results
Thank you to everyone who voted. Here’s how you rated the last MonDive.
Amazing, very clear
66.7%
Good, but a few parts were tricky
16.7%
Not impressive, it still felt overwhelming
16.7%
A NOTE FROM A READER
Selected: Amazing, very clear
“Love it”
Our reply: Thank you! That’s lovely to hear. We appreciate you reading MonDive and taking a moment to share your feedback.
We’d love to hear from you!
Know someone who may be interested?
Thanks for reading. See you next Monday.
