I Tested ChatGPT dots Early, and It's Huge Update

Author: Skill Leap AI · Watch on YouTube

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 ChatGPT dots Early, and It’s Huge Update · Channel: Skill Leap AI · 11:18 · Watch on YouTube

Topics: AI agents, Task and workflow automation, AI models

Summary

The video introduces ChatGPT Dots as an always-on AI agent that can take responsibility for goals, use connected apps and information, and message the user proactively. The presenter describes setting up a Dot, choosing its access and rules, and testing it on a product launch and content-planning tasks. In the demos, it identifies document inconsistencies, drafts updates, and finds relevant material, though the presenter notes that it can send too many messages and offers limited visibility into its activity. The feature is described as gradually rolling out to certain paid, business, and enterprise plans.

Key points

The presenter describes a Dot as an always-on agent that works toward a responsibility rather than simply answering a prompt.
Dots can use connected apps and information, including email, calendar, Google Drive, and ChatGPT memory, depending on the access granted.
Users can name a Dot, communicate by text or voice, and optionally give it access to their computer.
The presenter says Dots can work in the background without the user’s computer running and can be paused or reset.
In a product-launch demo, the Dot found a pricing mismatch and asked a follow-up question.
The presenter set rules for the demo: draft content but do not send or post it, do not change prices, and ask before actions that could move the launch date.
The Dot later rewrote social post drafts and an email, and the presenter says it continued sending updates.
The presenter found that the Dot could send too many messages and says users can ask for a daily recap except in urgent cases.
The video says rollout is gradual and initially aimed at certain higher-tier paid plans, business premium seats, and eligible enterprise accounts.

Chapters

Start	Chapter	Summary
00:00	Introducing ChatGPT Dots	The presenter explains how Dots differ from chat-based answers and task-focused tools, and describes their early access and rollout.
03:08	Setting up a Dot and its access	The presenter walks through the app, explains connected context and permissions, and shows how to start text or voice conversations.
06:14	Giving a Dot a responsibility	The presenter compares Dots with other ChatGPT tools and gives one a product-launch goal, documents, and custom rules.
08:46	Proactive updates and practical limits	The Dot reports work on launch materials and content planning, while the presenter discusses message frequency and the feature’s potential.

Quotes

A dot is an always-on AI agent inside of Chat GPT, and it has its own computer. — 00:31

it messages you. You don’t message it. — 01:03

My main value is keeping track of things I’m following through without you having to restart a task each time. — 06:14

Draft anything, but never send or post anything. — 08:15

This is the promise of AI. — 10:48

Concepts and entities

Dots, ChatGPT, ChatGPT 6 Astra, OpenAI, Codex, Google Drive, Slack

Full transcript (original language)

00:00 Chat GPT just released dots, and I think this might be the biggest update since the release of Chat GPT. Basically, when we first got Chat GPT, that was to get answers, right? You talk to AI, Chat GPT answered you in text format. Couple years later, we got Chat GPT Codex, we got Chat GPT work, and we started getting real work done with AI. So, you give it a task, and it came back with finished reports, spreadsheets, and complete apps. Well, now there is dots. A dot is an always-on AI agent inside of

00:31 Chat GPT, and it has its own computer. It understands what matters to you, and it takes initiative. And it keeps working on your behalf even when you’re not in chat. So, it’s not about giving it a prompt and getting a response. What you do is you could give it a responsibility, and it figures out what needs to do next to get that done. I was part of a group that had early access, so I’ve been testing it, and I think this is the biggest change since the release of Chat GPT. So, in this video I want to show you how it works. I want to show you kind of the three-step process that I

01:03 went through when I was testing it. I want to show you how to set up your own dots and give it a real goal. And I want to show you how it takes initiative, meaning it messages you. You don’t message it. So, it’s a whole different type of AI than what you’re used to so far. And since I had early access, my team and I actually put together a free guide just for this video. And this guide has different examples with your dots so you could actually start using it right now. And it has starter prompts that set off that responsibility for it. And this is a completely free guide you could access. I’ll put a link in the

01:34 description, and I’ll put a QR code on the screen as well. And by the way, dots are rolling out today, but gradually. They’re going to go to the pro plans first, and they’re going to start with the higher tier, so you have to pay the $200 a month plan. And they also have a new $500 plan, too. And it’s also available in the business premium seats and enterprise if your enterprise has it turned on. And couple other things they mentioned on their website, this is powered by the new ChatGPT 6 Astra, so it’s going to have the most powerful AI model available to it right now. And if

02:06 you haven’t used plugins before, basically ChatGPT lets you connect to 4,000 different apps, and they learn from your feedback over time. So, this is an agent that improves over time, and they will work towards a goal 24/7. They basically have their own computer, right? So, you don’t have to have your computer running for them to be able to work. Now, right now you have one primary dot in your ChatGPT account, but the way they envision it is they want dots to work together on your behalf. So, having different dots kind of going off there as different agents with

02:36 different task and different responsibilities and getting work done for you. And I’ll also link this below, so if you want to read more about the safeguards that’s set up for this, access and permissions they set up for this, and the review and approval process, I’ll link it below. But, I want to really dive in to my ChatGPT account so I could show you this in action. Okay, so inside of the new ChatGPT app, you’ll see a new section under new chat is called your dot. And then when you name your dot, it actually changes the name over here. Now, blurring a lot of this out just because this is actually

03:08 using the context from my own emails and Slack channels and Google Drive. It has access to a ton of different things. But, it’s really up to you how much access you want to give it. I’ll explain that in a second. And I’m going to reset mine here. The other thing I wanted to show you is couple of new things you’ll notice inside of the new ChatGPT app. For example, there is a whole section called spaces here. I’ll make a dedicated video, but it’s a different way you could organize your files, almost like your Google Drive, but inside of your ChatGPT account. And in

03:38 their demo right now, today was Dev Day. OpenAI showed ton of different examples of things that are coming to the app. But, I want to make this video more focused on dots. So, what I’m going to do is I’m just going to go over here, and I’m going to go go and reset this so I could show you just from scratch. Now, the first thing typically it’s going to ask you to do is you need to give it access to stuff to make it useful. So, the more context it has, as much as you’re comfortable with, the more it’s going to be able to pull into chat, the more retroactive it’s going to be able to message you about different things it finds. So, so far when I was using it,

04:10 it read my emails. It actually read attachments in my emails like contracts. It found issues in those contracts and it told me things that I need to point out. It actually found things on my calendar. It was able to find me flights. It was able to pull in things from my Google Drive that I needed an update. So, all those were already connected. So, if you have plugins connected in your ChatGPT account already, it will be able to pull that in. So, you could see it looks like a text message thread here. And the first thing it’s going to ask you is

04:40 it’s going to ask for a name and you could give it a name. Now, a couple other things you’ll notice here is you could actually start a message at any time with this little icon. Now, this follows you around on different parts of the web, too. So, it’s not just in the ChatGPT account. So, if I go on a website, for example, you’ll see it over here and then you’ll see this other icon. This basically starts a voice conversation, so you could send it a voice command here and you could do this on the phone, too. Now, while I was testing it out, it does a couple of different things. One is if I don’t say

05:12 anything to it, it will say something to me. So, it will go through different things that it has access to and it will find different things it could help me with. Now, let me show you how you do that. So, over here, you could give it access to your computer. So, you could allow access, but by default, it does have its own computer. But, if you want it to take actions on apps on your computer, you could give it access there, too. So, I’m actually going to allow it access. That way, you could look inside of folders and things like that. And if you go to the plugins tab right over here, these are all the

05:43 different plugins that ChatGPT has access to and therefore Dot could have access to. So, I give it access to a ton of different things here on mine, but again, this is based on what you’re comfortable with. Now, let me ask it a question here. I’m going to ask it what you could help me with that chat GPT work or codex can’t do. So, right now, this may get a little bit confusing, but hopefully I could clarify this. You could chat with chat GPT. That’s your regular chat GPT where you get an answer, right? That’s good for a lot of different things and then you jump into chat GPT work where you want it to

06:14 create things for you like a presentation or a spreadsheet. And then codex is for typically building different things. So, it could be video animation or entire websites or platforms or apps. Dots is different than all of that. So, let me see what kind of answer he gives me here. And this is what he told me. There’s a lot of overlap. My main value is keeping track of things I’m following through without you having to restart a task each time. And then it’s also pulling some personal things here from memory. So, he does have access to your chat GPT memory too. So, he’s pulling some things

06:44 from that. And he says, “I wouldn’t claim those things are impossible. The difference is having one ongoing assistant across work and everyday life with access and approval control.” He also decided to create a personal scratchpad on his own to keep track of some things that I have going on. And he says, “This is a place for project and decision you want to move forward with.” So, it could keep notes to itself this way. And this is something by the way that’s going to show up in spaces more and more. So, these are interesting. But, let me actually show you a demo

07:15 here. I want to give it a goal or a responsibility and then have it go to work so I could show you some of the things that it does in the background here. And I want to actually do this with kind of a fake company. Otherwise, it pulls in too much of my personal information. This was actually really hard to demo. So, right now, I’m going to give it a goal to do a launch for a product. So, this is kind of a fake product I made for this demo. And I’m going to give it a folder worth of documents here so it has access to everything. Now, you could have this in your Google Drive and it has the

07:45 connector or the plug-in that it goes and grabs those things. You can’t always just grab and drop it like this as you would inside of ChatGPT or ChatGPT works, too. So, kind of works the same way there. Now, one thing that hopefully they improve on is right now it’s actually working. It did message me, but it’s not showing anything like thinking or working or anything like that. So, sometimes I was a little confused if it was doing something, but it is actually doing something and it’s working all the time. It’s actually on 24/7. If you don’t want it to be on, they do give you

08:15 an option here. You could pause it at any time here and you could also reset it so it becomes a new dot. Now, if I go back to my prompt, by the way, I asked for three custom rules. Draft anything, but never send or post anything. Never change the price and anything that could move the launch date comes to me first. And it went through and it set up those rules. It asked for my okay here and it already found some issues here like a pricing mismatch between a couple of the different documents. It asked me a follow-up question. And now, I’m going

08:46 to actually not say anything and let it go to work because I want to show you the best part of this so far from my testing, which is when it takes initiative and it lets you know what needs to be done. Okay, so about 10 minutes passed and it actually found something here that it could help me with. It says, “I’m also rewriting the three social post draft that it found some issues with those and I’m also reviewing and rewriting your email reflecting this new update.” And here is the word doc that it rewrote here with

09:17 those changes. Now, while it’s working on this task, which it will keep working on and it will keep messaging you over the weekend, for example, it messaged me a handful of times every single day and at the end of it it asked if it wants me to get less messages and it could prioritize only the most critical and then give me a summary of the rest. I did reset it for this demo though. So, you can give it multiple responsibility as long as right now we have this one dot that could do multiple different things. So, let me give it one other one. This is the prompt I’m actually

09:47 going to use this one. So, this will always have my next piece of content ready before I finish the one I’m working on. And this is from that dots resource by the way. So, if you want these prompts or these use cases, they’re linked below. And I’m going to ask it to go check my Google Drive here for those scripts. And you could also combine these type of prompts with ChatGPT skills. So, plugin skills and dots, that’s how you really get most of these type of workflows. Now, it found my five scripts here that I just mentioned. It also found something about

10:17 the video I’m recording right now. So, it did give me a note and a link for that. And it looks like it’s working still. So, it did find some news here, some meta announcement about Meta Muse, which is very similar to this dot option that I’m recording right now, but it’s from Meta. And this will keep working. But, the one thing I did notice is I thought, over the weekend especially, it was sending me too many messages. It was finding too many things to do. So, you can actually guide it and tell it, “Hey, just give me a recap once

10:48 a day unless it’s urgent.” Now, I’m hoping soon they’ll release this to more plans so more people could use it that don’t have the pro plan or higher. But, this is really the promise of AI. This is when you have a real assistant that is proactive, that is doing things for you, that takes a responsibility and runs with it, and then just tells you what it’s doing throughout the day and throughout the week. And I’m working on more in-depth videos here. On the day of the release, I wanted to introduce you to it to show you a little bit of hands-on on what it’s capable of. But, again, make sure you check out that resource that I mentioned that has more

11:18 about how this is all set up, the different plans it comes with, and some use cases. And we’ll be adding to that as well. Thanks so much for watching this one. Let me know what you think of this update, and I’ll see you on the next video.