OpenAI Dots vs. Grok Bot: Which Agent Should You Give Real Work To?
OpenAI’s most interesting announcement today is a small character with a very large job description.
Dots are always-on agents powered by GPT-6 Astra. Give one an ongoing responsibility, and the pitch is that it keeps the work moving between conversations.
Grok Bot is already in this category. It also works in the background, uses software, and keeps context. So the useful question is: which one would you trust with a real job?
OpenAI’s Dots announcement on X
What Else OpenAI Announced
The September 29 DevDay releases fit together around making agents useful outside a chat window:
- GPT-6.1 Sol: OpenAI describes near-Astra performance at lower cost for coding, computer use, and professional work. That is a vendor claim to test against your own tasks. Model documentation
- Ultrafast: A premium speed tier. OpenAI advertises up to 8x faster token generation in Codex and up to 6x in the API. Faster generation does not guarantee a task finishes that much faster. Announcement
- ChatGPT Space: Pages, files, and shared work in one place, with people and agents helping create and revise the material. Space documentation
- Plugins and Sign in with ChatGPT: Plugins package instructions and connected tools. Sign-in lets supported partners use your ChatGPT identity; additional app permissions are separate. Plugins · Sign-in
My read: OpenAI wants to own more of the workflow around the model—where work lives, which tools it uses, and who follows up.
Dots are the most personal version of that bet.
What a Dot Actually Does
A dot has its own cloud computer and browser. Cloud work can continue while your laptop is off. Supported plugins provide access to connected tools; local work requires your connected computer to remain online with the ChatGPT app open.
OpenAI documents delegation to ChatGPT Work or Codex. That makes a dot interesting as a coordinator: something you can keep talking to while other tasks run. Getting started
It draws on relevant ChatGPT memory and keeps its own notes. Those notes are not a perfect transcript. Recurring work needs a saved schedule, and event monitoring depends on what the connected service supports. Tasks and memory
The useful starting instruction is concrete:
Keep our launch brief current. Check the connected feedback channel, flag changes that affect the copy, and prepare revisions for review. Bring me unresolved decisions. Ask before publishing or contacting customers. Confirm what monitoring you can set up.
That gives the agent a responsibility, sources, and a point where it should hand judgment back to you.
Dots vs. Grok Bot
Here, Grok Bot means the persistent agent product, rather than the @grok account people mention for replies on X.
| What matters | OpenAI Dots | Grok Bot | | --- | --- | --- | | Work environment | Cloud computer, plus optional connected local computer | Persistent cloud computer shared by your Bots | | Coordination | Ongoing responsibilities with delegation to Work or Codex | Multiple Bots collaborating and passing work between themselves | | Context | Relevant ChatGPT memory plus the dot’s saved notes | Persistent context and learned routines | | Practical reason to try it | You already work in ChatGPT and Codex | You want multiple workers or X-focused workflows |
Grok Bot can learn a routine from a demonstrated workflow. Its Bots share files, browser sessions, and logins on one persistent machine; isolation is per user, not per Bot. Separate job titles do not create separate access boundaries. Grok Bot product page and FAQ
For X research, Grok has a concrete integration: search posts, read timelines, and check mentions through the X connector. That makes it a sensible candidate for launch feedback and competitor monitoring. X integration announcement
There is no controlled head-to-head result here establishing which product finishes your work more reliably. Model preference is a reason to experiment, not a verdict.
What X Is Saying
The launch-day posts I found point to three questions worth keeping:
Does the model change the experience? Gergely Orosz welcomed Dots as an answer to Grok Bot and expressed a preference for Astra. He said he planned to try it; the post is anticipation, not a completed comparison. His reaction
Is there an ordinary use case behind the demo? Rudrank Riyam says he has been using a dot named Ping to plan his schedule. It is a modest firsthand example, and probably closer to how adoption starts than the promise of automating everything. His post
Can people access it, and will they connect it? Gurbaksh Chahal asked whether anyone had received Dots, despite being in California. Luis questioned the event from the perspective of a European subscriber. Compound248 expressed unwillingness to give Dots broad personal access. These are individual reactions, not a survey—but access and trust are real adoption questions. Access · European reaction · Trust
Access and Cost Matter
Dots are rolling out gradually to adults on Pro 100, Pro 200, and Pro 500 outside the EEA, UK, and Switzerland. Business Premium and Enterprise are rolling out worldwide; Enterprise requires an administrator to enable access.
Conversation with a dot does not count toward ChatGPT usage limits, but delegated Work and Codex tasks count toward those products’ limits. Included deeper-work allowances have extended limits for the first month. Always-on does not mean unlimited work. Dots availability and usage
Grok’s current pricing lists Bot access with $30/month SuperGrok, and its Bot FAQ also lists eligible Cursor plans. Included usage and extra token charges matter more than the entry price alone. Grok pricing · Bot FAQ
Which Would I Try First?
Dots, if your work already runs through ChatGPT and Codex. The appeal is continuity: keep one coordinator informed while research, documents, and code tasks develop.
Grok Bot, if X is an important input or you want a roster of workers. Its X connector and multiple-Bot workflow give you specific things to test today.
For either, start with one bounded responsibility. Run the same job for a week and measure correct outputs, missed changes, unnecessary interruptions, and total usage. Open the sources. Inspect the work.
The agent that saves you the most supervision is the one worth keeping.
Research snapshot: September 29, 2026. Product details come from vendor documentation; X posts are attributed reactions and individual experiences. This is a launch analysis, not a hands-on comparative review.
— Max