In partnership with
Hey folks,
OpenAI launched Dots at DevDay this week. I had one running the next day.
Setup was quick. Pick a character (I went with the pixel guy whose head is on fire) and give it a name. Mine is dot-face. Then I asked it the obvious question: what can you do for me?
I expected the usual generic list. It named my projects and business ventures and how it could provide value which felt very personalized. It offered to research our competitors, review proposals, write scripts and build decks. It knew it could organise files on my Mac. It had been reading my past ChatGPT and Codex sessions, and it started from there.
Then I asked it to get me ready for tomorrow. That's where it got useful and a little creepy.
Let's dive in..
What’s a dot?
A dot is an agent that stays on. You give it a job and it keeps working between conversations, even when you close the laptop.
It runs on GPT-6 Astra, OpenAI's best model. It has its own computer in the cloud. It can also use yours if you let it.
It plugs into more than 4,000 apps. Mine has my Gmail, Google Calendar and Drive.
Your first dot comes with the Pro and Business Premium plans at no extra cost. Right now it's rolling out everywhere.
Everything happens in one thread.
The first task
I asked my dot to get me ready for the next day.
Dot-face went through my calendar. Then my email: the ones where I'd said I would do something, and the ones I'd forgotten to reply to at all. Then Drive, where it found files I'd been tagged in that needed action. It pulled all of it into one list of what I had to do tomorrow.
That's the lightbulb moment. It had full context across all three, and it condensed the lot into 1 very good list of action points.
It even caught the emails I'd forgotten to answer.
Then it turned that into a daily summary. It wrote its own scheduled task, 8am daily. Here's the brief it wrote for itself, word for word:
"Check for newer replies and resolutions before calling something outstanding, distinguish confirmed facts from uncertainties, include useful direct source links, and end with the most useful next action... Only read and summarize; do not send messages to others or change my calendar."
I didn't ask for any of that. It checks its own work before telling me something's overdue. It separates what it knows from what it's guessing. And it set its own limit: read only, hands off my calendar, no messages to anyone.
That's the brief I'd give a good executive assistant on day 1.
It runs its own computer
I opened dot-face's computer. It has its own desktop, with a browser, a terminal and free apps like Blender and GIMP already installed.
It all lives in OpenAI's cloud, so it keeps going after I close my laptop. And you can open it any time to see what it's doing.
That Take over button is the guardrail I trust most. A rule a dot writes for itself is only an instruction. Watching it is how you check.
Passwords stay away from the AI. The dot pauses, you type them into a secure login box, and it carries on.
OpenAI also double-checks anything that touches your accounts, and can pause the dot if something looks unsafe.
Access to my Mac is a separate switch which is off by default. I turned it on, so dot-face can also use my laptop, including my own browser when a site blocks its cloud one.
What it costs
The dot is included in Pro, however OpenAI halved what the $200 Pro plan gets you. Codex and Work drop from 20x on the Plus allowance to 10x. GPT-6 Pro messages drop from 200 a week to 100.
Above it there's a new $500 plan with 25x the Plus allowance and a faster speed tier.
The dot also comes with extended limits for the first month after launch. So both generous versions run out around the end of October. What you're trying right now is the best it will be on $200. Judge it in November.
Your task this week
You don't need a dot to try this. Any assistant connected to your email and calendar can run the same job. Paste in the brief my dot wrote for itself:
Send me a concise daily briefing based on my connected email and calendar.
Cover my top priorities, important outstanding decisions and commitments, meaningful updates, and today's schedule with a short look ahead to tomorrow.
Check for newer replies and resolutions before calling something outstanding, distinguish confirmed facts from uncertainties, include useful direct source links, and end with the most useful next action.
Only read and summarize; do not send messages to others or change my calendar.Turn off model training in your assistant's data settings, if it has one.
List what you've connected. If you wouldn't hand an assistant that folder on their first day, disconnect it.
Sound familiar?
Over 4 million people have had the same lightbulb moment.
Morning Brew is a free daily newsletter that breaks down what's happening in business, finance, and tech — clearly, quickly, and with enough personality to make it the best email in your inbox.
No yelling. No filler. Just the news, finally making sense.
Cheers, Tim
