3-Line TL;DR
- Dots is the OpenAI announcement I am most eager to try, because it brings the persistent assistant idea that excited me about OpenClaw into a managed service
- My first experiments would be stock research, tracking work, and possibly running parts of this blog, with useful results that carry over to the next day
- OpenAI announced it on September 29, but access is rolling out by plan and region; this is my intended trial, not a hands-on review
OpenClaw made me want this kind of bot
- In my DevDay recap, I covered the whole event, but Dots deserves its own post because this is the direction I have been waiting for
- OpenClaw made the idea of an assistant that keeps working between messages especially appealing to me
- Its creator, Peter Steinberger, announced in February that he was joining OpenAI to bring agents to more people, while OpenClaw would continue as an independent open-source project (His announcement)
- That makes Dots feel like an important arrival in the same direction, without establishing that it is simply OpenClaw renamed
- OpenAI describes Dots as persistent agents powered by GPT-6 Astra, with their own cloud computers, connected apps, and access through ChatGPT, Slack or Teams (Dots announcement)
- The part that interests me is continuity: I want to give it a responsibility, then return to something that has progressed
Three jobs I would try first
- A dot can continue work on its cloud computer while my own computer is off, and carry context across conversations (Dots guide)
- That is a more interesting proposition than another chat window waiting patiently for me to remember what I was doing
- I would start with three recurring jobs where I can tell whether the output actually helps
| My first trial | What I would give it | What I would want back |
|---|---|---|
| Stock research | A watchlist, company filings, earnings dates and preferred sources | Changes since the last check, original links, and questions worth investigating |
| Work management | Tasks, deadlines and the projects I permit it to access | Blocked items, approaching deadlines and a short list of decisions I need to make |
| This blog | Topics, existing posts, writing rules and approved access | Sourced drafts, checks across both languages, and confirmation of what was actually published |
- For stocks, I am interested in reducing the time spent collecting filings and news before forming my own view
- A neat summary is useful only if I can trace a number to the filing and distinguish a confirmed event from someone else's prediction
- I would keep actual investment decisions and orders with me; the use case above is my proposed research workflow, not an announced brokerage integration
- For work, I want it to notice what is stuck and what needs my response, then help advance the tasks I assign
- If asking it to organize my work just creates another long summary to read, it is taking minutes rather than moving the work forward
This blog would be a revealing test
- Blog management is tempting because it contains the whole loop: find a subject, check sources, write, prepare images, publish and verify
- I would first test whether Dots can maintain the editorial context over several posts instead of rediscovering the same rules each time
- The useful output would be a draft with sources attached, matching facts in Korean and English, and a clear record of the publication state

My proposed trial, not an official Dots workflow or a completed test
- The distinction in the diagram matters: OpenAI says proactive background research is read-only and cannot itself send messages, change connected apps or control a browser (Privacy and safety FAQ)
- Actually carrying out an assigned task is a separate step, governed by connected permissions and action review
- I would give it the publishing responsibility explicitly after reviewing the draft, then expect it to check the resulting pages and images
- A blog assistant should be able to distinguish “draft ready” from “published”; otherwise it has automated the most annoying status update
- The API already gives this blog a way to publish, so the interesting experiment is whether the agent can reliably complete the surrounding editorial work
- Rules and approval controls help define the job, but they do not establish that the writing or execution will be correct
I would try it as soon as my account gets access
- The launch includes a primary dot for eligible Pro and Business Premium users, while specialist dots begin in an enterprise pilot (Launch details)
- The current setup guide lists Pro availability outside the EEA, UK and Switzerland, Business Premium in supported regions, and an Enterprise beta that admins must enable (Getting started)
- Rollout is gradual, so an announcement does not establish that the feature is already enabled on my account
- Initial creation happens on desktop web or the desktop app; mobile app use follows a supporting update
- Conversations with a dot and work it executes have different accounting, with Work or Codex tasks using the usual allowances (Usage guide)
- I would begin with one small recurring responsibility, such as a watchlist briefing, and see whether it stays useful across several days
- What I am excited about is the possibility of a bot that understands the ongoing job and keeps it moving, including the slightly self-referential job of helping maintain the blog writing about it
Community Reactions
- One reader thinks Dots is a more useful launch than Custom GPTs, which captures some of the excitement about moving beyond configured chatbots (Reddit comment)
- Another worries about privacy, a relevant concern when a persistent assistant accumulates context from connected services (Reddit comment)
- A DevDay viewer found the demonstration unconvincing; that is an individual impression of the demo, not a measured reliability result (Reddit comment)
Q&A (Field Notes)
- Q. Is Dots the official service version of OpenClaw?
- That is how the direction feels to me, but the available announcements do not establish a simple rebrand or identical implementation
- Q. Have I tried it yet?
- No, these are the jobs I want to test when access reaches my account
- Q. Can it run stock trades or publish this blog automatically?
- I have not established a trading integration or tested Dots with this blog's API, and I would start with research before explicitly assigning any publication task
- Q. What would make the trial successful?
- Accurate sources, remembered context, useful follow-ups and work that genuinely progresses, rather than a growing collection of summaries
