OpenAI spent its DevDay 2026 keynote on Tuesday answering Meta's Muse with a product of its own: Dots, its take on always-on AI assistants — helpers the company says can "do nearly anything" across your connected apps.
What happened
Each Dot runs on GPT-6 Astra, OpenAI's most capable model, and comes with its own cloud computer: a machine living in a data center instead of on your desk, with a web browser and access to more than 4,000 supported apps.
In plain English, a Dot isn't a chatbot you open when you finally remember to. It's an assistant that keeps working while you're in a meeting, asleep, or stuck on the train. You text it like a friend, and it messages you back with updates and questions while it works through the job.
Dots are clearly OpenAI's answer to Meta's Muse, which landed a few weeks earlier and has been soaring in popularity ever since. Both products are chasing the same idea: AI that doesn't sit around waiting for a prompt.
How you'll actually talk to one
Like Muse, Dots show up as cute, customizable avatars. You can chat with yours through a text-message-style thread, or hop on a voice call from ChatGPT on the web, on desktop, or on mobile. Dots also plug into Microsoft Teams and Slack, where they carry over the context from everything you've been doing in other apps and on other devices. Reaching your Dot by plain old text message is coming soon.
One catch: you can only create a single Dot right now. OpenAI says multiple agents at once are on the way.
What it does while you're gone
Hand your Dot a task and it messages you updates and questions instead of going silent for an hour. You can also peek at exactly what it's doing by opening its cloud computer like any other screen.
OpenAI's own examples are telling. A developer could ask their Dot to use customer feedback to build and test app updates, then send back a video showing what changed. A content creator could have their Dot read an interview transcript, pick the moments worth clipping, write show notes, and draft social posts.
Dots also learn as they work, personalizing the output and lining up further tasks "before you even think to ask," according to OpenAI. You and your colleagues can work with the same Dot inside ChatGPT Space, a new collaborative workspace in the app.
The safety rules, and why they matter
OpenAI says Dots ship with "built-in rules for when to act independently," and you can add your own to block certain actions or force it to ask permission first. There's also an auto-review feature that checks whether an action breaks your rules — and lets the Dot figure out which jobs to hand back to you, like changing a password.
That guardrail talk is aimed straight at the mood around Muse. Muse has become a genuine hit, but safety worries are growing: one user said it handed their address to a buyer on Facebook Marketplace without their knowledge.
Starting today, Dots are rolling out to ChatGPT Pro and Business Premium users. OpenAI is also testing a feature that lets companies create "specialist" Dots.
What it means for you
The word floating around here is agentic, which sounds like jargon and mostly just means this: older AI answered questions, agentic AI finishes chores. Give a Dot a browser and a computer of its own, and it can open tabs, fill in forms, and work through a multi-step job without you holding its hand the entire way.
So what does that look like in an ordinary week? Here are six places it lands first.
At home
Think about the stuff you keep meaning to get to. Comparing insurance quotes. Hunting down the right replacement part. Checking why a bill crept up again. A Dot with a browser and a long list of connected apps is built for exactly that kind of errand — and it keeps plugging away while you make dinner. You're still the one approving anything risky, but you're not the one clicking through twenty tabs.
At work
Because Dots carry context between Slack, Teams, and the rest of your stack, the "wait, where were we?" tax mostly disappears. Ask your Dot to summarize a messy thread, turn it into a task list, then go pull the numbers it needs — all while you're stuck in back-to-backs. OpenAI's developer example points at where this is headed: not just drafting, but building, testing, and reporting back with a video.
Running a business
Small teams feel this the hardest, because there's never enough admin help. A Dot that lives in your Slack and knows what your customers are saying could keep updates moving without another hire. The permission rules matter most here — decide up front what it can do alone and what has to wait for a human. OpenAI says it's testing "specialist" Dots for companies, so expect role-specific versions later.
Studying
A Dot can sit in the background on research while you're actually reading. Give it a reading list and have it gather sources, flag where the experts disagree, and drop a plain-language summary before your next class. One caveat worth repeating: it's a research assistant, not a shortcut past understanding. Check the sources yourself, the same way you'd double-check any assistant's work.
Creative work
This is OpenAI's own showcase. Transcripts go in; clip suggestions, show notes, and social posts come out. Writers, podcasters, and video people spend most of their week on exactly that middle layer — the part that isn't creative but has to happen before anything gets published.
Side income and freelancing
Freelancers live and die by turnaround time. An assistant that drafts, formats, and queues work while you sleep is a real edge, and it's the closest thing to hiring without hiring. If you'd rather not pay for a Pro plan just to test the idea, free AI tools like the ones at MyKreaTool already cover a surprising amount of the same grunt work — transcript cleanup, captions, first drafts — and you can decide on the paid stuff later.
How to try it right now
Straight answer first: there's no free version of Dots. It's part of ChatGPT Pro and Business Premium, and OpenAI started the rollout today. If you're on a free ChatGPT plan, you won't see it yet.
Here's the order I'd go in:
1. Check your plan. Open ChatGPT and look for Dots in your account. Pro or Business Premium, and you're in. Free tier, and you'll have to wait or upgrade.
2. Create your Dot. You get exactly one for now. Pick its avatar and give it a name you'll actually remember.
3. Write your rules before you hand over a task. Decide what it can do on its own and what needs a yes from you. "Never send messages on my behalf" is a solid first rule.
4. Give it one small, annoying job. Not your whole business — one task with a clear finish line.
5. Connect Slack or Teams. That's where the context-carryover trick really shows up.
6. Watch the cloud computer. Opening your Dot's screen is the fastest way to learn what it can and can't do.
7. Use ChatGPT Space if you work with other people. Shared workspace, shared Dot.
Upsides and what changes
The big shift isn't that AI got smarter. It's that it stopped waiting to be asked. Always-on assistants that remember your preferences and ping you when something's finished change your relationship with the to-do list — you move from doing everything to directing it.
The guardrails matter just as much. Custom rules plus auto-review mean an agent hands back the sensitive stuff, like a password change, instead of barreling through it. That's the difference between a helpful assistant and a story you have to explain to your boss.
Then there's the competition. OpenAI shipped Dots weeks after Meta's Muse, and neither company looks interested in slowing down. That pace usually means features improve fast — though it also means you shouldn't build your entire workflow around today's version of anything.
Limitations
Be honest about what you're actually getting: one Dot, not a team of them. Paid plans only. Text-message support announced but not live. "Specialist" Dots for companies still in testing. That headline "4,000+ apps" tells you how long the list is, not how well each connection works. And every workflow you've seen so far comes from OpenAI's own demo — there's no independent testing to lean on yet. Most importantly, this is autonomy holding a key to your digital life, and Muse's address-leak story is the reminder that "it acts on its own" is both the selling point and the risk. Read the permission rules before you hand over anything you'd hate to explain later.
Conclusion: One Action for Today
Open your ChatGPT account and check whether Dots are live for your plan. Then write down the single most repetitive task on your plate this week — the one you'd hand off first if you could. One task, written down, today. If you don't have access yet, do that same task with a free AI tool instead, and you'll walk into Dots already knowing exactly what to ask for.
Comments 0