I Tested OpenAI's New Personal Assistant Agent: DOTS

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: I Tested OpenAI’s New Personal Assistant Agent: DOTS · Channel: Futurepedia · 10:48 · Watch on YouTube

Topics: AI agents, Task and workflow automation, AI developer tools

Summary

Futurepedia’s host tests DOT, an always-on assistant that connects ChatGPT conversations, projects, tools, skills, and scheduled tasks in one place. The video demonstrates DOT creating project notes, coordinating work across multiple chats and skills, searching for products by voice, and setting a reminder for a future task. The host says DOT’s main benefit is continuity and delegating work across projects, while noting that some features are still rolling out. The test also encounters bugs with the cloud browser and scheduled-task interface.

Key points

DOT is described as an always-on assistant that can access ChatGPT memories, connected tools, skills, and scheduled tasks.
It can start new chats and work or Codex sessions, including on the user’s computer when access is enabled.
DOT has a persistent cloud computer, and the host says its chat can continue across phone and computer.
The host sees continuity across threads and projects as DOT’s main advantage over using ChatGPT, work, or Codex separately.
DOT created a scratch pad and project subpages for storing notes, checklists, and links.
The host asks DOT to coordinate three skills across different threads: drafting video intros, creating motion graphics, and starting a thumbnail.
In voice mode, DOT searches for magnetic wooden objects and shares links, but cannot open tabs in the host’s browser from the call.
DOT can create scheduled tasks, reminders, triggers, and automations.
The host reports bugs during testing, including problems with the cloud browser and opening a scheduled task.
At the time of the video, access is limited to Pro and Enterprise plans, with more capabilities expected to roll out over time.

Chapters

Start	Chapter	Summary
00:00	What DOT is and how to access it	The host introduces DOT as an always-on assistant that brings ChatGPT projects, tools, and tasks together, and explains its initial availability.
02:02	Setup, capabilities, and project notes	The host explores DOT’s settings and computers, asks about its capabilities, and uses its pages feature to organize notes for a video project.
05:04	Delegating work and testing voice mode	DOT coordinates skills across separate chats to create video materials while the host tests voice search and encounters a cloud-browser issue.
07:41	Reviewing results and scheduling tasks	The host reviews the generated intro, motion graphics, and thumbnail, then demonstrates a scheduled task and reflects on DOT’s benefits and bugs.

Quotes

I think of it as the bridge to all of these systems. — 00:30

The biggest reason to use me is continuity. — 03:02

So it’s orchestrating across all of this. — 05:36

It’s nice to have one central place to delegate from instead of jumping between each of them all the time — 10:17

Concepts and entities

DOTS, OpenAI, ChatGPT, Codex, Claude Code, Pages, skill, Computador na nuvem

Full transcript (original language)

00:00 Open AAI just released an always on proactive personal assistant agent called DOTS. I’ve had early access and have been testing it for a few days already, so I want to show you how it works and what it’s capable of. They gave me early access, but this video is not sponsored in any way. Now, the easiest way to understand what it is is to see it in action. But first, I’ll give a quick explanation. Dots is an agent that’s always on and has access to all parts of your chat GBT. the memories, your connected tools and

00:30 skills, your scheduled tasks. It can start new chats, new work sessions or codec sessions, including on your computer when it’s on. I think of it as the bridge to all of these systems. The more threads and projects across chat, work, and codecs you have going, the nicer it is to have one central place you can go to understand and direct it all. Dot can be the orchestrator of all those systems at once from one place. And it will even find and suggest things to you based on all it knows. Plus, it

01:00 also gets its own persistent cloud computer it can use to pull things up and store things over time. And you can talk to it on your phone or your computer and the chat persists between them. Now, like I said, I’ve been testing it already, but I actually just reset it so I can demo setting it all up from scratch. as it’s rolling out. It’s only available on pro and enterprise plans. I have no information on when it will be rolled out to everyone. If you have it, it will be over on the left panel near the top where it says your dot. So, it’s introduced itself, said,

01:31 “Message or call me anytime. I’ll keep things moving even when we’re not talking and check in with updates or questions.” Said, “What do you want to call me?” And it’s already starting to be proactive here. So, it said, “I’ll start looking for ways to help, anything top of mind.” And it also made us a scratch pad and started suggesting things. First off, it did ask what I want to call it. I’ll say I’ll just call you dot. I’m good with that. And dot it is. Now, when I click on the dot up top, it will open up a panel on the right. I can easily edit the name or I can change the look of it. Got the dot icon or

02:02 there’s all these cute little characters. Personally, I like the little dot, so I’m going to stick with that. Nice and clean. Then you’ll see the options for call, email, or Slack. So, call is basically just voice mode. Now, with email, they actually haven’t given them their own personal emails yet, but there is the option for Slack. So, you can add it to a workspace, and I think it’s actually really good to set it up in Slack if you have it. And below that, you’ll see the computers. You have your dots computer. It gets its own personal cloud computer where it can look things up, pull up multiple tabs,

02:32 and you can see what it’s doing and take over if you want to. We’ll get to that later. Then, there’s also my computer since I am on the desktop right now, which I am going to give it access to. then it will be able to control all of my codeex threads. Let’s toggle that back off and let’s ask it something. Since me and Saj from Skillap both had early access, we put together a free resource with a few use cases and sample prompts you can copy paste to get started. It also explains what it is, how it works, and some best practices. And we are going to keep updating this

03:02 resource over time as we build up our use cases. Download that for free using the link in the description. So, I’ll just ask it. What types of things can you help me with? List out everything that you’re capable of and why I would use you over chat work or codeex. And it responds once right away. The biggest reason to use me is continuity. And that has been my experience. The continuity in having one place to go for everything is the biggest benefit for me. And I like that it sends me a little bit while

03:32 it’s working on the rest. So then it’s got the longer response now talking about using it for ongoing projects that connect several things, videos, research, email, files, tools, and follow-ups. Still use chat for quick thinking and codecs directly when you want to work closely with it on a build. And then gives a bunch of example use cases. The strongest use case for me, it says, would be something like, “Help me keep this video moving through research script, demo assets, and release checks. Bring me decisions when you need me.” So I’ll try out an example like that in just one second. First, I want to go back to this scratch pad it created. So

04:03 I can open up that and this is another new feature they have called pages. So it’s kind of a notion like system for taking notes and storing knowledge. So you can build that up over time since your dot is just one really long chat thread. Sometimes previous context can get lost in the mix. So having this scratch pad or with those pages you can create brand new ones for separate projects. So it’s nice to have that to keep all of your ongoing notes and context. This is an awesome new feature that I’m sure I’ll go deeper on in a separate video. It’s already created this from what it knows about me and it has all the sections you can look

04:33 through and then it has these hyperlinks with more information. You can create subpages if you’ve used notion before. It is a lot like that. So I’ll say since you understand what dots can do, create a new subpage in the scratchpad for a video I’m making about the launch of dots. All right, it created the dots page. It has all these notes about it. Made some checklists. links out to some different chats we’ve had, created multiple different subpages, and you can see the pages it’s created up here. Now,

05:04 I’ve got a home for any notes about the video. And this has all been pretty basic so far. So, I’m going to ramp it up a lot with this next one. I mentioned it can access and use basically any part of your chat GPT. So, I’ll have it use a few things at once. I’m going to hide this sidebar, and then those tabs will still be there anytime I want to go back in. I’ll say, “Now, use my skills to get this video started. Use the intro skill to write three intros in work. Read through them. Then pick the best one. Next, take that intro and make some motion graphics that follow the script

05:36 using the codec skill. Then also get a thumbnail started using the thumbnail skill in a separate chat. It’s going to use three skills across multiple chat threads to accomplish this task. And one of them is actually running a skill for the intro using work. Then it’s going to take that response once it’s done and start a new thread in codeex and use a separate skill there. So it’s waiting for one to finish before moving on to the next step somewhere else while at the same time starting a whole separate chat to create the thumbnail. So it’s orchestrating across all of this. And

06:07 while that’s working, I’ll show the voice mode as well for something else. And I’ll do this one from my phone because that’s the only time I ever use voice mode. So I’ll click my dot up top. Then I’ll hit the call button. I actually think it’s kind of weird that they make it seem like a phone call and have it ring. >> Hey, welcome back. >> Hey, so recently I saw a really cool art piece that was like a magnetic wooden car. >> Ah, okay. >> Yeah, they were made of real wood then had magnets so they could stack up and

06:37 look balanced like a Kairen. Can you use your browser to find some online that are like that? Just open any links you find in your browser as separate tabs. so I can see them. >> Got it. Let me take a look. >> For me, I only really use voice mode while I’m >> Okay, I found a really strong match. Handshaped walnut stones with hidden magnets, plus a few similar and mixed wood sets. I sent the four closest links to our chat for you. The walnut one is especially close. I can’t open tabs in

07:09 your browser from this call, though. >> Now, currently, cloud computer is down. I don’t know what’s going on there. It was working before I reset my dot. Um, doesn’t seem like it can figure it out. So, it did pull these up in separate tabs using codeex’s inapp browser. It’s a good workaround. And that is exactly what I was looking for. I did a bunch of testing with this browser. It’s been working perfectly for the past few days. Maybe it’s because I reset my DOT. Also, this is a beta version of the chat GPT app. That could have something to do with it. I was using this function all

07:41 day, so it should be working at launch. But let’s go check in on the other project I gave it. Looks like it’s got some of this going. So, let’s see what it did. It’s got it actually created a thumbnail already. You can see here where it was messaging back and forth. It’s got the three intros. Then it’s like picked one and sent that into this chat here. Selected intro number one. Then it’s asking it to use the specific motion graphic skill I have. And that’s another benefit of having dots delegate to codecs or work. It will prompt itself with a lot of detail that you might not

08:12 put into your prompt yourself because it would take too long. And it looks like it’s got a thumbnail for me, but I actually already had an idea for this thumbnail. I’ll ask for a quick change to the thumbnail. And the motion graphics are done. And this is obviously supposed to go with the script where I’m talking about how it works and then it is going along with exactly what I’m saying just like I do in my normal videos. But I’ll skip ahead here. Yeah, that looks awesome. That was a nice transition there. And the intro skill and this remotion motion graphics skill are both ones that I was working on in

08:44 Claude Code. I actually just barely transferred them here so I could demo this. They’ll probably need a little finetuning to work smoother in here, but that was still really solid. Just about there on the thumbnail I will be using for this video. I’m actually currently working on refining my entire content system, so these skills are still brand new. But I’m trying to get everything to kind of match the midjourney background I have behind me and keep it cohesive across the thumbnail, the motion graphics, and my background. And been working on that for the past couple days. So, these skills are not all the way there yet. My last video was the first time I’ve used them. But that

09:14 example involved multiple tasks and threads over multiple responses with DOT acting as basically the project manager across all of it. And you can get it managing far more than that and managing multiple projects at once. Another thing it can do is set up recurring tasks, triggers, reminders, and various forms of automations. I’ll do something simple. Tomorrow is OpenAI’s dev day. Set a reminder for yourself. Review the main keynote. Soon as it’s finished, give me a summary everything they

09:45 launched. Then under scheduled tasks, there’s now a separate tab for your dot, which for some reason is not able to open right now, but the task is there. And these work like the existing scheduled tasks where it can trigger from something being added to a form or an email coming in or set it up on a schedule every day or every week or just a one-off task you want done in the future. All sorts of different ways to automate work just like before, but these ones go directly to your dot. So that’s been my experience so far with DOT. It’s worked best for me as a bridge to all my

10:17 threads and projects. It’s nice to have one central place to delegate from instead of jumping between each of them all the time and as a way to get work started while I’m out. It is brand new and they are rolling out some of the capabilities a little slowly. More will be getting added to it over time like multiple dots, their own emails, the ability to pay for things, things like that and probably a lot more. And as you can see, there were a couple hiccups along the way. There are some bugs still. I said that I tested its cloud browser extensively before I started filming, but then I guess after I reset

10:48 my dot, it just stopped working. Then for some reason, this scheduled task won’t open. And this is a separate beta version of the desktop app that could have something to do with it as well. I’m sure people will be flooding the feed with new use cases and experiments over the coming weeks.