GPT 6 Astra AI Full COURSE 1 HOUR (Build & Automate Anything)

0:00Welcome to this one hour masterclass on

0:03GPT-6 and Astra. I'm just going to show

0:06you exactly how to automate basically

0:08anything using GPT-6 Astra, which has

0:11blown me away since it's been released.

0:13It's super powerful, fantastic to use,

0:16really fast, very very impressive.

0:19Um honestly, I never thought I'd see the

0:21day or I wouldn't didn't expect to see

0:23the day this year where, you know, I

0:24would switch from Claude to GPT, but it

0:28that day has come, right? GPT-6 Astra is

0:30very very impressive. So, today what

0:32we're going to do is we're going to run

0:33through a full step-by-step course on

0:36how to build and basically automate

0:38anything using GPT-6 Astra. In this

0:42course, in step-by-step we'll run

0:44through exactly what it is, why it's so

0:46powerful, how to use it, how to get

0:48access to it, and how to basically

0:49automate anything with it. If you want

0:51to skip through anything, I've got the

0:53list of what we're going to cover today

0:54inside this document. So,

0:56you know, if you want to skip through

0:58there's certain things that you prefer

1:00to check out first, you can skip through

1:03the timestamps. But, we're going to get

1:04straight into this, all right? So,

1:07let's have a look here and we'll have a

1:09look at some of the things that you can

1:10create. So, I mean, you can see here for

1:13example,

1:14like just from their examples, like it

1:15can fill in spreadsheets, it can for

1:18example do game development, it can use

1:213D MCPs and modeling like for example

1:24Blender, which I'll show you some

1:26examples of in a minute. It can do your

1:28basics like formatting documents and

1:31quality control and everything else. You

1:32can also build apps, tools, games,

1:35websites, etc. In fact, I've built out

1:38full-blown like dashboards like you can

1:40see right here. This is I I I literally

1:42cannot code, right? I cannot write a

1:44single piece of HTML, but I can create a

1:46beautiful tool like you can see right

1:49here. Um and the UI looks amazing. And

1:52the way that I've done this is by using

1:55GPT-6 Astra to help me improve the UI,

1:57to help me add the functionality and

1:58everything else around it. So, it's an

2:01incredibly powerful tool. And with that,

2:03let's get straight into exactly how you

2:05can use it, all right? So, first place

2:07you can go to is you can go over to

2:10ChatGPT.

2:12Now, this depends on what plan you are

2:14on, all right? So, you need to make sure

2:16you're on the right plan. If you're

2:17wondering which plan do you need to

2:19actually use GPT-6 Astra and get used to

2:22get access to it right now.

2:24So, you see the details here. GPT-6

2:26Astra is rolling out to limited set of

2:28organizations and will then be available

2:30to ChatGPT Plus, Pro, Business, and

2:33Enterprise users, as well as through the

2:35API, right? You can also get access on

2:38Microsoft Azure and AWS Bedrock.

2:42So, there's a few different ways you can

2:44use this, as you can see right here. Um

2:47and you know, feel free to use it

2:49however you want to. For me, the two

2:51places that I'm going to be using it are

2:53inside ChatGPT directly here, and also

2:55inside the ChatGPT work app here. All

2:59right? So, those are two of my favorite

3:01places to use ChatGPT. And I'll run you

3:04through the basics of how you can use

3:06each of these in a second, all right?

3:09So,

3:10now, you might be thinking, "Okay, where

3:12do you start with all of this? What can

3:14you do? What can't you do? What are the

3:17limitations, etc.?" So, if we have a

3:20look, for example, and we we go to

3:22ChatGPT and GPT-6 Astra, as you can see

3:25here,

3:27I have found literally no limitations.

3:30Now, I will say that some things are

3:31better than others when it comes to

3:33these designs and that sort of thing,

3:34and I'll cover what we can do and what

3:36we can't do. But, I mean, for example,

3:38if we have a look at this, this was a

3:39video that we generated with AI, fully

3:43edited, all fully generated, as you can

3:45see in this example. Looks beautiful.

3:47It's better than most editors could

3:49create. And this was fully done with a

3:51free tool, a free open source tool, and

3:55GPT-6. And the other thing that blows my

3:57mind with all of this is that you can do

3:59it with one single prompt.

4:01So, if you want to get access, two ways,

4:03right? Make sure you're on the right

4:04plan. Which plans can you get it on?

4:06Plus, Pro, Business, and Enterprise.

4:09Where can you get it? You can go to

4:10chat.openai.com. And then, the thing

4:12that I've noticed here inside ChatGPT is

4:15like

4:16it it's got 6 Pro, as you can see if you

4:19scale up the thinking and effort over

4:20here, or you can use higher or extra

4:22high. I think for default here on the

4:24models, if we check,

4:26is

4:27we

4:30Um but, you can also go to work here. If

4:32you go to work, you can actually select

4:35Astra inside the

4:38a scroll here. And my favorite

4:41place to use it is actually inside the

4:43ChatGPT app. You know, I think you can

4:45do a lot more with it. I think it's a

4:46lot more fun. Um and, you know, if

4:48you're watching this live, by the way,

4:49feel free to ask any questions as you go

4:51along. But, yeah, I I think, honestly,

4:53inside the Codex app is much better.

4:56So, we've covered how to get access.

4:59Let's talk about some example

5:00automations I've built recently and what

5:02we can do with this. So,

5:04we've covered

5:05video generation

5:07as an example here. What else can we do

5:09with this? So, if we go over to this

5:12section, we actually created a full game

5:15with this system, right? So, you can see

5:18this game right here.

5:20Right? This game.

5:21We created this game in pretty much one

5:24single prompt using the combination of

5:28Blender

5:30and GPT-6 Astra. All right. So,

5:34the the reason that the graphics look so

5:36nice on this is like you can basically

5:38you can either generate images and turn

5:39those into games, or you can use this

5:40image generation alongside Blender,

5:43which is like a 3D modeling software.

5:45Uh we And then, you can

5:47basically get to control that and model

5:49it all for you. But the quality of that

5:51game is beyond

5:53anything that I've seen, uh, anything

5:55else create really, especially in one

5:57single prompt, you know. You might say,

5:59"Okay, should I be using, for example,

6:01Claude?

6:02Or should I be using Gemini or that sort

6:04of thing?" I mean, it depends what you

6:06prefer day-to-day and you know, where

6:09your setup is. But even for me as

6:11someone who's been using Claude for for

6:14months now and really favored it over

6:16ChatGPT, I will say that ChatGPT's

6:20Astra

6:21is smoother, faster, better at computer

6:24use, does everything that Claude can,

6:27and can do it better. It's also much

6:30better at being direct. And what I mean

6:32by that, for example, is like if you ask

6:35ChatGPT Astra to start doing something,

6:39it just gets straight to it, and it's

6:40very direct. If you ask it a question,

6:42it'll answer your question. And you

6:44might think, "Well, that sounds basic.

6:46Like every AI can answer a question,

6:48surely." But actually, if you go over to

6:50something like Claude, Claude sometimes

6:52it refuses to do stuff, sometimes it it

6:55doesn't answer the question properly. It

6:58does seem to be very bloated and slow

7:00compared to anything I previously

7:02used with any other model, you know,

7:04with literally if I compare Opus 5 or

7:07Fable 5 or anything like that.

7:09It just gets distracted, right? Whereas

7:12I know that if I ask Astra to go and do

7:14something, it will just get it done as

7:17well. And the other thing that I like

7:19when it comes to

7:21agents particularly versus Claude, like

7:25Claude, you can't plug into Hermes

7:27agent. And Hermes agent is probably like

7:30the best open-source AI agent out there.

7:32Now, the problem with that is like if I

7:34can't use Claude with my subscription

7:38I'm already paying for inside Hermes

7:40agent, I can't plug it in, but I can

7:42with GPT-6,

7:44then I'm far less limited with GPT-6,

7:47right? I can use GPT-6 Astra frontier

7:49model inside Hermes agent, just using my

7:51assistant subscription. I don't need to

7:53pay for an API or anything like that.

7:54That is amazing compared to like not

7:57being able to use Claude. And if I want

7:59to use Claude as a brain of Hermes agent

8:01or any other agent that I use, well, I

8:03have to pay for the API, and that's kind

8:05of annoying, honestly. So, and also

8:08pretty expensive. So, I wouldn't

8:10recommend that, you know, like and and

8:11these are some of the biggest reasons

8:13that I'm switching over to Astra. It's

8:14just it's faster, it's more direct, it's

8:17smoother, it creates better quality

8:19outputs, and also

8:22we can plug it into our favorite AI

8:23agents. Having said that, some

8:24limitations that you need to be careful

8:26of. I've mentioned already that you

8:28don't get it on every single plan. And

8:30also, another big limitation is the

8:33weekly usage limit, right? So, I haven't

8:36used it like a massive amount. I

8:38wouldn't consider myself like an

8:40absolute power user of GPT-6. Bear in

8:42mind, I only got access about 2 days

8:45ago, right? 2 days ago, and that feels

8:48like a long time ago, actually. But,

8:50when I look at my usage here, you can

8:52see we've already used 44%

8:5544%

8:57of our weekly usage, and I'm only like 2

8:59days in. So, just be careful of that. Be

9:01mindful of that. It's like, yes,

9:03fantastic, but it can use up a lot of

9:06tokens. So, there are limitations to

9:07this.

9:08Um having said that, super powerful,

9:10easy to use, and everything else. So,

9:13just to recap here, we've talked about

9:14examples of automations with video

9:16generation, uh games. I've shown you the

9:19agent OS tool that we've built with

9:22Astra as well. So, you can build all

9:24sorts of cool, fun, wonderful stuff. If

9:26you actually go on X, X is a fantastic

9:29place to to really see what people have

9:31created. And honestly, it's just so many

9:34amazing things I've seen people create

9:36with this tool like it's unbelievable.

9:39Right? And

9:40you know, I I think that

9:42we're really I mean

9:43this it's a little bit hypey. You know,

9:46I like I don't want to hype anything up

9:48too much, but if you look at what I

9:50think it was Nvidia that came out and

9:51said yesterday maybe Jensen Huang that

9:53said GPT-6 is AGI. Now,

9:56for me personally, I mean, what even is

9:58AGI? Right? Like everyone has their own

10:00sort of definition of it. At the same

10:03time, would I say it's the most powerful

10:05AI I've ever seen? Yes. Is it the most

10:07impressive model leap I've ever seen

10:09between GPT-5.6 to 6? Absolutely. So,

10:12yeah, I I do

10:13I do think it's a fantastic model. All

10:15right. So, we've talked about video

10:17generation, games, tools, websites, etc.

10:20Let's talk about how to automate

10:22anything and just get straight into the

10:24action.

10:25So, let's say for example, you wanted to

10:29build with GPT-6 Astra. You may be

10:32asking yourself at this point, "Okay,

10:33what should I build?"

10:36I would start with where you spend your

10:37time the most. So, for me for example,

10:40if I'm spending a lot of time on like

10:41generating ads, then guess what I'm

10:44going to automate first with GPT-6

10:47Astra, an ad studio. Why? Because that's

10:50the thing I'm spending all my time on.

10:52So, you know, whatever you use with

10:55Astra, don't try and automate things

10:57where you uh

10:59automate for the sake of it. Right?

11:01Automate something that you actually

11:02spend time on day-to-day. And I know

11:03that sounds obvious, but you'll be

11:05surprised by the amount of like people

11:07who are running a business and just

11:08automating stuff that they don't even

11:09need. Like wasting their time with

11:11things that aren't even useful. Don't do

11:13that. Right? Just focus on the things

11:16you actually spend your time on, and

11:17that will change week to week as you add

11:19more automations in. So, as you build

11:21more automations, you're going to find

11:23that you're spending your time on new

11:24things, and therefore you need to create

11:27new automations for those new things

11:28you're working on. So, loops up all the

11:30time. Loops up all the time. We're

11:31actually going to go inside the AI

11:33Profit Boardroom.

11:35If you haven't checked it out already,

11:36link in the comments description or go

11:37to the AI Profit Boardroom.com. If you

11:39go to the classroom here, we've just

11:42added a new section inside the beginner

11:44to expert course that basically shows

11:46you how to identify your time, how to

11:49stay focused, and how to never feel like

11:50you're overwhelmed again. Because,

11:52you know, a lot of people like, "Well,

11:54where do I start with AI?" And most most

11:56people know the power of it, but they

11:57don't know where to get started. We show

11:59you exactly where to get started inside

12:01the community. Plus, you can ask me

12:02questions and I personally answer them

12:04with video tutorials just to help you as

12:06as much as I can. And then also, you can

12:07jump on weekly coaching calls and that

12:09sort of thing. So,

12:11let's say for example, I am spending a

12:13lot of time with ads. I would say,

12:15"Okay, can you add a section inside my

12:16tool?" Or, you know, "Can you create a

12:18new dashboard for me where you automate

12:20ads, right?" And then from there, it's

12:23going to create it and you have to test

12:25it out, right? So, you see how I talk to

12:27it in plain English and I just tell it

12:28exactly what I want to build. That's

12:30what you should do, too. So, with that,

12:33let's get straight into this and I'll

12:35show you an example. So, if we go to a

12:37new chat here,

12:39and then we're going to delete this

12:41section. We're going to open up a new

12:43folder. Now, you don't have to work in a

12:45project, so you can click on don't work

12:46in project if you want to. And it's a

12:49new chat to use this. By the way as

12:50well, you might say, "Okay, do you use

12:52ChatGPT or do you use Codex?" Because

12:54there's two different versions of this

12:56inside the app. And if you're wondering

12:58how to get access to the app,

13:00you can just go into Google here

13:03and you just get this one, right? So,

13:05you just download this. This is the app.

13:08So, once you've done that, you can get

13:09it on both mobile and you can get it on

13:11desktop as well. Once you've done that,

13:13you download it, you open it up, you log

13:14in, and it's going to look something

13:16like this.

13:17If you've already got access to ChatGPT,

13:19but you don't have access to GPT 5.6 and

13:21you're on the right plan, just go to the

13:23check for updates here,

13:25update it, relaunch it, and then you

13:27should get access. And then from here

13:28you can change the model. And now once

13:31you're inside

13:33Chat GPT like so, you can switch between

13:35Chat GPT over here or Codex. For me

13:38personally, I would stick to Codex

13:40because I think that if you are using

13:43Chat GPT directly,

13:45it's it it like you're for actually like

13:49for business stuff and for actual

13:50building stuff, like it's much better to

13:52use the Codex. I just feel like it's

13:54much more controlled and that sort of

13:55thing, right?

13:56So,

13:57this is for like creating, learning, and

13:59exploring. This is for like actually

14:01building stuff. And today we won't to

14:03build anything, right? We won't to build

14:04and automate anything, which is what

14:06we'll focus on. So, we're going to use

14:07the Codex section. Now, inside the

14:10models over here, you can see that we

14:12can switch between different models. So,

14:14if you tone this down, it will use less

14:16tokens, which means you will get more

14:18out of your subscription without facing

14:20limits. However, the lower you put it

14:24down, the less powerful the model's

14:26going to be. And if you want to use

14:27GPT-6, which is the newest version, you

14:30would switch up here. Now, you can

14:31choose between Astra light.

14:34You've also got medium and you have

14:36extra high. For me personally, I like to

14:38stick to medium. That's my favorite. So,

14:41we click on new chat, then we're going

14:43to select a new project. And let's say

14:45for example you just wanted to build an

14:47app, maybe for like I don't know,

14:48fitness tracking or something like that.

14:50So, we can put in a project name like

14:51so. We won't be spending a lot of time

14:53on this. But around like for for

14:55building stuff like this, it could be

14:57personal or it could be for work, right?

15:00So, for example, for me personally, I I

15:02train Muay Thai. I like to train Muay

15:04Thai. I could create an app for that,

15:06right?

15:06If I spent a lot of time playing guitar,

15:08which I used to, but not so much since

15:10AI came out, then I could create an app

15:13for learning and playing guitar.

15:14Whatever you prefer. My main point here

15:16is like

15:17it doesn't have to be work related. It

15:18doesn't have to be business related.

15:20Although you'll probably get the highest

15:22um outputs in terms of

15:25your time versus outputs with something

15:28business related. But, let's just uh

15:29create a random app here. So, we're

15:31going to click on fitness tracking,

15:33click create project, and then from here

15:35we can choose between local

15:38new local work tree, cloud, and you can

15:41also see here that we can check our

15:43usage.

15:44And then we've got this option here,

15:46which is branches.

15:48You don't need to worry about that too

15:49much. It's a little overcomplicated. The

15:51main thing here is like

15:53this is very, very useful to know at

15:55this point.

15:56So, if you are running

15:58uh a new project

16:00and you want to share it with people

16:02you're better off doing that in the

16:03cloud. And you should probably

16:04understand that before you start uh

16:07building anything, before you start

16:08prompting it here. So, for me, for

16:11example, I have automations running in

16:13the cloud for SEO.

16:16So

16:17these cloud-based automations, they

16:19create the my website. I don't need my

16:21computer open. I don't need ChatGPT open

16:24because it's running in the cloud. It's

16:26not on my local computer. Fantastic. Cuz

16:29that can just run 24/7 as a routine. And

16:31we'll come on to routines later. I'll

16:33actually add that on the list, too.

16:35My point here is that if you're going to

16:37shut down your laptop but you need

16:38something to run in the background

16:40choose cloud. If you prefer something

16:42running locally and it's just you who's

16:44going to be accessing this project, then

16:46you can go to local. Make sense, right?

16:48Make sense. So, from here we'll click on

16:50local, and then we're going to say,

16:52"Okay,

16:53create a beautiful fitness tracking

16:57that is

16:59designed for mobile devices.

17:03Um make it super useful and actionable."

17:06All right. So, we're going to start

17:07creating that app, as you can see here.

17:09Now, you'll notice as soon as I pressed

17:11enter there

17:12the chat just disappeared, right? Where

17:14did it go? I pressed enter. It's not

17:16there. It's not obviously here on the

17:19left-hand side. Where did it go?

17:21This is something that puzzled me the

17:23first time I used it. So, I want to

17:24explain that quickly here. So, what

17:26you'll see on the left-hand side is you

17:28have to scroll down. So, we have our

17:29pinned projects at the top here.

17:32Would you agree?

17:33And then as we scroll down here, we've

17:35got our other projects. So, pinned means

17:38that we keep it at the top. So, for

17:40example, if we're looking at that

17:41fitness project that we're working on

17:42right now, the fitness tracking one,

17:44we're going to right click that, and

17:45we're going to click on pin. Why do we

17:47do that? Because it keeps it at the top,

17:49and it keeps it easy to track.

17:50Otherwise, you don't want to be

17:51scrolling down here every single time.

17:53And if you have a lot of projects that

17:54you're working on,

17:56this is something that's going to save

17:58you time and keep everything much more

18:00organized. So, it's important to

18:01understand that.

18:02So, if we go to this section here,

18:04you can see now that we have the fitness

18:07tracking system being built. Fantastic.

18:11Now, that may take a while. Something to

18:13understand here is like sometimes you're

18:15going to build a project. I have noticed

18:17that Astro 6 is pretty fast, but it

18:20might take 5 10 minutes to build that

18:21out. So, what do we do in the meantime?

18:23We start working on the next thing. All

18:25right. So, you can build out multiple

18:26projects at the same time, which is what

18:28I'm going to show you right now. So,

18:29we've covered how to build an app.

18:32Again, you just prompt it in plain

18:34English. If you're coding something out,

18:36you don't need to be a coder to use

18:38this. If you don't understand how to

18:40prompt, doesn't matter at this point.

18:42You literally just describe what you

18:43want and explain it. The only job that

18:46we really have as humans now when it

18:47comes to using AI is being able to

18:50clearly define what we need and want,

18:53and understanding the biggest

18:55opportunity in terms of where we can get

18:57the most time back. Most people don't

18:59have the clarity of thinking to do that,

19:02but if you're watching this, it's a very

19:03powerful skill that I'd highly recommend

19:05you develop because it may be the most

19:08important skill as we go into the

19:10future. So, as we have a look at this,

19:13let me show you another example. So,

19:15here we have

19:18a website design.

19:19Right? So, if we open this up,

19:22we can see that we have a website being

19:24designed over here.

19:26So, we're just going to log in.

19:28Uh you'll notice that this is being

19:30hosted on a chat.gpt.site

19:34URL.

19:36And you can see the page has been built

19:37right there. So, if you want to build

19:39websites like this, again, you just type

19:41in what you want. So, actually said,

19:42"Can you redesign this website so it

19:44looks amazing?" And then it redesigned

19:46our whole existing website. So, if you

19:48want to build a website with AI, pretty

19:50simple. You just give it any information

19:53or examples of what you like, and then

19:55you go over to it and you say, "Hey,

19:57create this for me."

19:59If we go over to this video section

20:01here, and let me just take these off as

20:05we go along. So, we've talked about

20:07um a SaaS tool design.

20:09If um we've talked about apps

20:12as well.

20:13Let's talk about website design. This is

20:15pretty important to understand. So, from

20:18what I've seen so far,

20:19when it comes to designing a website

20:22with chat it is not particularly great

20:25straight out the box.

20:27So, if you go to chat.gpt,

20:29and let's say

20:31we're creating a new website like so.

20:33If you just say create a website for

20:35this or do this, etc., doesn't look

20:37particularly nice. What you want to do

20:38instead is either train it on a skill

20:41that helps you create websites, or you

20:44can actually just give it a screenshot

20:46of an example that you really like. So,

20:49let's say for example, and we'll take a

20:51screenshot here. Let's say for example,

20:53I really like this design

20:55for a website. I can take a screenshot

20:57of that.

20:58Then we can go over here, and we can

20:59say,

21:00"Can you create a beautiful website like

21:02this for my AI community?

21:06Make it clean.

21:07Follow same principles, make it super

21:10useful and interesting.

21:12And so, when you give it examples of

21:13what you like, it's much more likely to

21:15give you something that's closer to what

21:17you had in mind. If you do not give it

21:19examples of what you like, and you don't

21:21give it like a screenshot or anything

21:23like that of of how you want the design

21:25to look like, it will probably turn out

21:27very bland, and it won't be that

21:29interesting at all. So, for example, if

21:31we go back to this page here, to me

21:34personally, it looks okay, but it's not

21:36that interesting. Right? It's It's It

21:38kind of looks like a ChatGPT website

21:40straight out of the box. I've seen Codex

21:42follow that same design over and over

21:44again. The other thing that I'd

21:46recommend here is that you can go over

21:47to somewhere like GitHub, and you can

21:49get like 3D website design

21:52uh projects as well.

21:54And you can take a look at GitHub.

21:55GitHub is basically

21:57it's kind of like a website where you

21:58can get ideas

22:00for and you can get like these kind of

22:03add-ons that you can quickly install to

22:05your AI agents so that they give you

22:08better outputs. So,

22:10what that means essentially

22:13is that when we are doing website

22:15design, we can give it a skill from

22:17somewhere like GitHub,

22:19and train our agents on exactly how to

22:22design a beautiful website.

22:25So,

22:26if we have a look over here, we've got a

22:28bunch of website design ideas,

22:30design systems.

22:32Here's an example.

22:34Let's see what else we got here.

22:36Page designer as well.

22:39That looks like an interesting one. But

22:40basically, you can get these skills to

22:41teach your AI agents how to design.

22:44There's also another one

22:46called UI UX

22:48Pro Max,

22:50which is another website design skill,

22:52and you can just type this into Google,

22:53and then you'll find it on GitHub like

22:55so.

22:57And you can actually

22:59go to Codex

23:02like so.

23:03And you can

23:05say, "Hey, install this skill for

23:07website design."

23:09So, if we go to a new chat here,

23:11we'll type in a new project.

23:13I'll call this UI/UX Pro Max.

23:16Paste in the link of the GitHub with the

23:18skill and say, "Use this skill to create

23:21a beautiful website for the AI Profit

23:24Boot Camp."

23:25And then obviously, when you're creating

23:26a website or anything like this, you

23:27want to train it on information

23:30about you and your context.

23:33So, if we look at this example skill,

23:35we would inside the prompt normally give

23:37it information about us, what we do, how

23:39we working on things, what we what we

23:42working on, etc.

23:43There we go.

23:45So, that's website design. And there's

23:47two different options for that like I've

23:48mentioned. You can give it a skill

23:50like UI/UX Pro Max

23:53to make better beautiful websites or

23:55you can actually give it an example of a

23:57screenshot

23:58and say, "Hey, go off and create this."

24:01Now, what you're also going to notice

24:02here

24:03is that we now have three projects being

24:07done in the background. So, we can check

24:09on these. If you see this loading icon

24:12here inside Codex, you can see that we

24:14can check through these different

24:17systems here and we can see what's being

24:20built and where we're up to. So, for

24:22example, if we click on the UI/UX Pro

24:24Max project, we've got an update and it

24:27says it's creating the website. If we go

24:28over to the AI Community section here,

24:31you can see that it is now

24:33working on building out that other

24:35website. And if we go to the mobile

24:37fitness tracker here, boom shakalaka, we

24:39now have the website ready to preview as

24:43we can see right here. Now,

24:45what you're going to notice is that we

24:46have a preview on the right-hand side

24:49like so and we can flip through, we can

24:51test it and that sort of thing.

24:53And we can actually open that up in

24:55Chrome as well.

24:57Now, some people at this stage

24:59might be saying as well, "Okay, well,

25:00that's great that you got the mobile

25:01app, but how do you, for example, submit

25:04it to the App Store?" What you would do

25:07is you would say, "Okay,

25:09I want to submit this to the iOS App

25:12Store. Can you get it ready to export?"

25:16So, that would be number one.

25:18And then number two, if you wanted to,

25:19for example, create an app that you're

25:21going to issue publicly, you need to

25:23make sure that people can log in. So,

25:25for example, if we look at this whole

25:27app here, nobody can log in. The data

25:29isn't stored anywhere. Okay, so, like if

25:31if I'm using this and then someone else

25:33is using it, well, how do they log in

25:35and they store their data? So, what you

25:37want to use is something called

25:38Superbase. Superbase is another free

25:41plugin you can give to your AI agents

25:44like Codex, and that can then plug in

25:46Superbase as a database for people to

25:49log in to our apps. So, what we can do

25:52here is we can say, "Okay,

25:54make sure that people can log in using

25:56Superbase so that you can store data and

25:59you can have a database of data for each

26:01person who's using the app."

26:04And then also, the final thing that

26:05you'll notice here is like the colors

26:07look a bit bland and a bit boring. So,

26:10anytime you have feedback, if you test

26:12something, it doesn't work, or if, for

26:14example, you look at something you're

26:15like, "I'm not really into that design.

26:17That's not my style." You just give that

26:19feedback to the AI agent, and you

26:21iterate from there. Bear in mind like

26:23you can get the 80/20 very, very

26:25quickly. We've got the 80/20 of an app

26:27right here. Like that's good to go. The

26:29problem is that we need to customize it

26:32so that it doesn't look like everything

26:34else. It doesn't look mediocre. That

26:36it's actually useful, et cetera. And so,

26:38how do you do that? You just give

26:39feedback to your AI agent over here, and

26:41then it begins to code out the project

26:43on the right-hand side. So, for example,

26:45here if we say, "Okay,

26:47can you start

26:49improving the UI, making it a bit more

26:51colorful, making it much more

26:53interesting, entertaining to use, etc.

26:55Gamify it as well.

26:57And so you see how we give the feedback

26:59to the agent, we make sure it has a

27:01database for using the app. And then

27:03also, we're going to give it that the

27:05details for exporting it to iOS and

27:08Xcode and everything else.

27:10Pretty simple and easy.

27:12Uh one of the questions we just got here

27:14actually is, you know, who do you think

27:15is winning the AI race right now? Is it

27:18GPT-6 or Fable 5.1? I think in terms of

27:22actual intelligence, in terms of the

27:24best AI to use right now,

27:27undeniably, it is GPT-6 Astra. I tested

27:30out Fable 5.1, it's okay,

27:33but honestly, for me, I think Claude

27:34needs to step up a lot at this point.

27:36Like, GPT-6 Astra is way, way better.

27:39So, you can see here that it's beginning

27:40to now uh come up with some ideas, and

27:43it's going to give us options.

27:45So, for example, we're going to just

27:47give some options here, and then it can

27:48start using that to fill out the rest of

27:51the details.

27:52Now, what you see is the

27:55uh

27:56page for the community. So, you can see

27:58here, for example, we gave it details of

28:01the page we wanted to create, and it's

28:02created that over here.

28:04And so you see, like, when you when you

28:06give your AI agents examples

28:09of the UI or the design that you want to

28:11do, whether that's with a screenshot or

28:13whatever, it will look 10 times nicer.

28:16And it just won't look like your generic

28:18sort of um AI fluff. You know, see here,

28:20it's asking me to

28:22log in to Superbase, which we do in the

28:24background. But this is the page that

28:26it's created um for the website that we

28:28asked it to build.

28:30Pretty nice and easy. You know, and it's

28:32it's created that community page right

28:33here.

28:34It's pretty cool.

28:36So,

28:37we've built out a website.

28:39We've built out

28:42another website over here with the UI UX

28:44ProMax system.

28:46So, if we open this up, this is actually

28:48hosted locally. And you can see that

28:50basically ChatGPT has its own browser

28:53over here.

28:54But, what you can also do is you can

28:56load that up locally inside Chrome or

28:58whatever you prefer to use. So, if we go

29:00to Chrome here and we type the local

29:03address in, you can see that we have the

29:04page ready to go. So, it's found the

29:06details of um

29:09the page we want to build, what we want

29:11to add inside there. And that just looks

29:12a lot nicer, as well. So, you can easily

29:14build out designs and that sort of

29:15thing. First thing I'll say here is like

29:18it's still it's still not customized

29:20exactly how I want it, but again, that's

29:22where the feedback stage comes in. And

29:24you'll notice that for any sort of

29:26project that you build with ChatGPT,

29:27whether that's a video, a game, a

29:30website, a tool,

29:32there's basically three stages. You

29:34describe what you want,

29:36then you test it, and then you iterate

29:39with more feedback. And you just loop

29:40round. So, some people watch this and be

29:42like, "Oh, you know, that's not the best

29:44design in the world, you know what I

29:45mean?" And it's like, "Yeah, I know. I

29:46know. You need to do a lot of feedback.

29:48You need to give a lot of feedback to

29:50your AI agents. Just to test it, go

29:52through it, see what you think about it,

29:54customize it exactly how you want from

29:56it." So, if we go back to, for example,

29:59that tool that I built earlier, the

30:00Agent OS, which again is available

30:02inside the AI portfolio. This tool is

30:04amazing.

30:05But, it took me day, I mean, months at

30:07this point of like just more feedback

30:10every single day, right? Create

30:11something, build something, test it,

30:13give it more feedback, loop back around,

30:15build a new and test that, loop back

30:17around. And so, like if you see anything

30:20today that you're like, "Ah, I could do

30:21better than that." Everyone could do

30:22better than that. It's just going to

30:24take a bit more feedback and a bit more

30:25time. Just something to bear in mind

30:27there.

30:28So,

30:29we've talked about website design.

30:32What else we got here? Images.

30:34So, ChatGPT can generate images. Now,

30:38it's using something called ChatGPT

30:40image 2.0.

30:43So, when you're using Astra, Astra 6,

30:46you want to think of it as the brain.

30:48So, it's the brain, but image 2.0 is its

30:52image generation skill.

30:54And so, when ChatGPT generates an image,

30:57it's using ChatGPT image 2.0. So, Astra

31:00is not actually creating the image, but

31:02it can orchestrate the skill to build

31:04the image for you. So, if we go into a

31:06new chat here, we don't need a project

31:08for this. So, I can just say, "Okay,

31:11create an image of a cat eating

31:12pancakes."

31:14All right? And then, you can see here

31:17that it will start building out the

31:18image. Now, at this point, you might be

31:20thinking, "Okay, that's great,

31:22but why would I just want to generate

31:23one single image?"

31:25And I would say you've got a great point

31:27there. So, when you're generating an

31:28image,

31:29you don't just stick with the image. You

31:31can use it for social media. You could

31:33create an image for ads. You could

31:34create

31:35um thumbnails, for example. You could

31:36create

31:37an image and then use that on your

31:39website design. You could take an image

31:41and then get Astra to turn that into a

31:44game.

31:45So, my my point here is like, images are

31:48very powerful

31:50because

31:51you can animate those. You can take them

31:54and transform them into websites, apps,

31:56games, tools. It can make all of your

31:58pages look a lot nicer. So, if we have a

32:00look at this, for example, this image

32:01right here,

32:03we can now go back

32:05inside the chat

32:07and let's have a look where we are now.

32:09Again, you see how we have to scroll all

32:11the way down to this recent chat.

32:14We can comment on here. We can also

32:15remove the background. We can remove it

32:16or resize it, as you can see.

32:19So,

32:20what we can do here is we can say,

32:21"Okay, can you, using your skill for

32:24videos, remotion, could you turn this

32:26into animated video, just like a a quick

32:2910-second promo video?

32:32Can you also use this image to create

32:35like a a 3D racer game or something like

32:37that. So, I'm going to show you two

32:39different examples

32:41where basically

32:42you can

32:44get an image like this

32:46and then you can turn that into an

32:48actual project.

32:49So, you can see that's now beginning to

32:51work on both of these. Now, in the

32:53meantime

32:54we've talked about images. That's where

32:56we're going to come on to videos. So,

32:59video is pretty cool with

33:02Chat GPT Astra because Astra, number one

33:06is fast and number two is very, very

33:07good at coding. And the way that AI can

33:10create videos is with code. All right,

33:12it's it's using the skill that basically

33:15um

33:16creates HTML.

33:18And with HTML, it can generate a

33:20beautiful

33:21video. So, for example, here

33:24we have the fitness tracker being built.

33:27We also have the cat pancake images, you

33:30can see. But, if you go back to the

33:33video section here, you might be

33:35wondering, "Okay, how did you create

33:36this promo video? It looks pretty nice."

33:38It's pretty simple and easy. So, what

33:40you can do is you can use something

33:41called Remotion. Now, Remotion

33:44is a skill you can install to your AI

33:47agents to basically create videos. And

33:50it's a free open-source skill, which

33:52means that you can give it to any AI

33:54agent. You can create videos so

33:55genetically and you can create these

33:58awesome animations just in plain text.

34:01So, what you would do is you would grab

34:03the GitHub link of Remotion. Then,

34:06you're going to go over to Chat GPT or

34:08whatever agent you prefer to use.

34:10And then you can say, "Okay, use this

34:13and create an amazing promo video for

34:15whatever." Right? And it could be like,

34:17"Okay, create an amazing ad. Create an

34:18amazing promo video. Create an amazing

34:20presentation. Create an animation." You

34:22see my point? Like, anything that can be

34:24animated, you can do this

34:26with this system. And one thing that I

34:28actually really like to do, and this is

34:29something I built as a custom workflow,

34:31but I'll show you it for

34:33inspiration.

34:34So, inside our agentic operating system,

34:37we have a video agent, and it uses

34:40Remotion

34:42to create amazing videos like you can

34:44see right here.

34:45And

34:47if

34:49we go back to our videos

34:52over here, you can see some of the other

34:54videos that we've created.

34:56And these all have like audio in the

34:57background and that sort of thing. And

34:59it's pretty crazy like some of these

35:00videos

35:02they are

35:04quite long videos. So, let me show you

35:07an example here.

35:08So, this one

35:10is a 5-minute long video.

35:13What it does is it's used a combination

35:15of

35:16another tool for generating the B-roll,

35:18then it's created a voiceover and a

35:20script for the video,

35:22and it's put it all together and edited

35:23it. And literally all we do is we go

35:25inside here and we type in the type of

35:28video we want, and it can generate it

35:29and save it into our workspace.

35:32Here's another example. So, this video

35:33you can see an AI avatar down here,

35:36and you can see that he's talking on the

35:38screen and it's green, right? And so, my

35:41point here is like it doesn't even just

35:43have to be for a basic animated video.

35:45Like you can create something

35:47pretty unbelievable

35:49just just by combining Remotion and

35:52other APIs. So, for example, like for AI

35:55avatars, you might use API like HeyGen.

35:58You can grab the API key from HeyGen,

36:01then you plug that into the Astra

36:04conversation that you have for the

36:05video, and you say using a combination

36:07of this HeyGen API

36:08and your skill for videos, Remotion,

36:12create animated edited avatar video. And

36:15that's basically what we did with this

36:17video right here.

36:19It's pretty cool.

36:20Now, if you come back to our project

36:22with the app, I want to show you an

36:23example of logging into the app. So,

36:25this is a mobile app that we built.

36:27We just need to sign in to Superbase.

36:30And then it will allow us to to log in

36:33and create

36:34a login for that. All right.

36:36So, you can see here that

36:38we can now log in

36:40to our account

36:42on that habit tracking app.

36:45And it will save our data using

36:46Superbase. And I don't need to edit it.

36:49I don't need to add any of that in.

36:52Astra just logs in for me and once I've

36:54logged in, then with Superbase we can

36:57add the login page right here.

36:59So,

37:00if you come back to that

37:02example

37:03with the image, you see how it's created

37:06this

37:07image for the video and it's generated

37:10the full animated video right here. All

37:12right.

37:12And then it's it's created this ready

37:14set pancakes

37:16animated promo using the image

37:19inside the video itself. And that's just

37:21a combination of Remotion as a skill.

37:25And then the image generation skill,

37:27which is with ChatGPT. That's another

37:30big benefit to using something like

37:34GPT versus Claude. Claude can't generate

37:37images. You have to give it an API. So,

37:39you have to give it access to image

37:42generation API like for example ChatGPT

37:44image 2.0, which is the best one. So,

37:47with ChatGPT inside your assistant

37:49subscription, you get access to this as

37:51well, which is a massive benefit

37:53personally.

37:54So, you can see here it's now creating

37:55the game as well, which we can open up

37:58later.

37:59That's essentially it for videos. Now,

38:01let's talk about computer use. So,

38:05we're at the point now and honestly

38:07before Astra came out, this was not

38:09particularly great. But now we're at the

38:12point with AI where it can control our

38:15computer, you know, I can control

38:18pretty much anything on our computer and

38:21do anything that a human would do. Now,

38:23it's not perfect. It still struggles in

38:26parts, so I want to be 100% transparent

38:27with you, but it's way better than

38:30anything else I've seen, and it's pretty

38:32impressive so far. So, for example, if

38:34we go over here, we have ChatGPT and the

38:37Chrome extension set up. If you want to

38:40find the Chrome extension, I'm just

38:42going to explain this cuz some people

38:45struggle and they they put this in the

38:47comments sometimes. So, if you're

38:49looking for the Chrome extension for

38:50ChatGPT, you would just go to

38:53you would just go to Google and type in

38:54Chrome extension ChatGPT. So, anything

38:57that you're trying to find today,

38:57whether that's like Remotion, skills,

38:59whatever, you can just go to Google and

39:01ask it for it. So, for example, here you

39:03can see we've got ChatGPT. We just have

39:05to add that to Chrome. Once we've added

39:06it to Chrome, we log in, and now you can

39:08see we've got ChatGPT over here. Now,

39:11for me personally, the Chrome extension

39:13is not that useful. The only thing that

39:15I actually use it for is computer use.

39:17So, for example, if we go over to

39:20ChatGPT here, and by the way, it's

39:22it's actually created

39:24the game based on the image we generated

39:28earlier.

39:29Doesn't look that similar to the

39:30original image, but I mean, look at the

39:31game itself. Like, pretty nice. 3D game,

39:34nicely edited. It just created it in the

39:36background while I was talking to you as

39:37well. So, you get the idea. Like, it's

39:40it's really really good for generating

39:42and creating games and all sorts of cool

39:43stuff. Anyway,

39:45if you go to a new chat here,

39:48and then we're going to open up a new

39:49project. We'll say, "Okay." We'll call

39:51this

39:52school post, for example, like this.

39:55And then I'm going to

39:57go over to

40:00my

40:01school group. We'll copy that and paste

40:03that link in

40:05just as a reference point.

40:07We'll say to ChatGPT,

40:09"Hey, can you go to the page above using

40:14computer use? And then what I want you

40:16to do is make a post about chat GPT

40:19Astra and how we have loads of great new

40:21training inside the AI Profit Boardroom.

40:25On exactly how to use it. And also a

40:27bunch of free bonuses inside there, too.

40:30And then I'll just give it a link to the

40:32AI Profit Boardroom.

40:34And I'll say add urgency, make it

40:37interesting

40:39third-grade

40:40reading level

40:42Dan Kennedy style

40:44marketing post.

40:46And create a new post. So this is an

40:49example, right?

40:50Uh we're going to hit enter like so.

40:54So we can see

40:55that is now thinking and it's going to

40:57open up the community. So you can see

40:59here it's beginning to open up the

41:00school community. And

41:02so it's connected to our Chrome and we

41:04can see it's opening up Chrome right

41:06here.

41:07Now you saw how fast that is. Really,

41:09really quick when it we're using it.

41:11And so if you look, we know that chat

41:14GPT is using that tab.

41:17So it's having a look here and here.

41:19And it's controlling our browser. Now

41:20look how fast this navigates. So this is

41:24chat GPT right here controlling

41:27the screen and navigating across it.

41:31And now it's just putting in the title.

41:33Right? So it's added the details here

41:35and then it should put in the post.

41:38And the crazy thing about this is it can

41:40navigate the buttons really quickly. You

41:41can scroll around the page. This isn't

41:44AI controlling the computer and it's

41:46fast. But also if you got a virtual

41:48assistant to do this and you gave them a

41:50race, like how long would it take to

41:52create a page like this? Probably ages.

41:56So you can see it's now created the

41:57post.

41:58Right? So you can see it's created the

42:01marketing post for us. Uh it's added a

42:03couple of links.

42:05The other thing that I would say is it

42:07is actually using information from the

42:09page originally and then

42:11use that inside the post if that makes

42:13sense. So, it's not just like generic um

42:17AI fluff. [snorts] Like, it's actually

42:19personalized the post itself

42:21to the information it saw on the sales

42:23page for the Air Profit boarding, which

42:25is fantastic. Also, I like the way it's

42:27written. Like, it's easy to understand.

42:29It's very clear. It's beautifully

42:31formatted.

42:32It's added urgency.

42:35I mean, and it did it in first time

42:37round whilst I was talking to you.

42:39Now,

42:40I think at this point

42:42for like if if you got basic tasks for

42:44like like for example posting, and you

42:46need it to come up with the creative in

42:48terms of the writing and what's going to

42:50be in the headline,

42:51and also post it for you. We're at the

42:52point now where I think you can get

42:54Astra to do most of that for you, okay?

42:56And you can even have that running as a

42:58schedule in the background, which I'll

42:59come on to in a second.

43:01But, yeah, really powerful. And then it

43:02said published. You know, it only took

43:0490 seconds. Imagine how long it would If

43:07you've ever hired a a virtual assistant

43:08or a human to do stuff like this for

43:10you, they might take like 10, 20

43:12minutes, 30 minutes thinking about this.

43:14This has cut that down that time window

43:17down into 90 seconds.

43:21That's how powerful GPT-6 Astra is. And

43:23it wrote it better than most humans

43:25possibly could. It's personalized it.

43:28There's going to be no mistakes in terms

43:29of the spelling or the punctuation

43:31inside the post because obviously Astra

43:34is very good at understanding

43:35grammatical rules. And it's published it

43:37and it's ready to go.

43:38Okay? Just happened in the background.

43:40So, so fast and easy and smooth. And I

43:43can tell you now if you got for example

43:45like Fable 5 to do the same thing,

43:48it would struggle. It doesn't do it

43:51quite as well. It can It can do the same

43:52job, but it'll be much slower. And I

43:55would say the content itself in terms of

43:57the post, the headline, everything else,

43:58just wouldn't be as nicely designed as

44:00this. So, this is way nicer.

44:02All right? So,

44:05pretty awesome. And

44:07as you keep going along now,

44:09we've covered computer use, we've

44:10covered apps, SAS tool design, website

44:12design, images, video, computer use,

44:14right? We've We've covered a lot

44:15already, which is fantastic. I want to

44:17show you an example of the game design

44:18that we did earlier. So, I've I've shown

44:20you a couple of games that we built with

44:21them already. So, we created the 3D

44:23racer game with the cat.

44:25If you're not sure about that, it's over

44:27here. This is the game that we just

44:29created in the background.

44:30All right, so this is the game. And it's

44:32even got like controls, which is pretty

44:34fun as well.

44:35So, we can pause this and we can play

44:38around with this and whatever. But then

44:40also,

44:41if we go to the app, we've built the

44:43app, we created the video earlier today,

44:46as we mentioned.

44:48Now, this is where it gets quite

44:50interesting.

44:51So, we can also create cloud routines,

44:54schedule tasks that run in the

44:57background.

44:58So, what do I mean by that? Well,

45:01let's say for example you had a website.

45:04All right?

45:05So,

45:06if I go to the aiprofitboarding.com,

45:09this is my website that I own. And we do

45:12a lot of SEO for this website.

45:15Now, the great thing about AI,

45:17let me show you this is the results

45:19we've had with this website. The great

45:21thing about AI is like it can definitely

45:23get you results with SEO. So, for

45:25example, if you look at this website,

45:26we've grown it from like zero to 423

45:30clicks a day using AI SEO. And it's

45:33fully automated. I can't even log into

45:34this website, which is amazing. And by

45:37the way, if you want a free SEO strategy

45:38session, link in the comments and

45:39description, and we can give you a

45:40custom tailored game plan for this whole

45:42stuff. But my point here is that the you

45:45can get AI to automate your SEO for you.

45:48However,

45:49you still have to prompt the AI to do

45:52this daily. You still have to come up

45:53with the keywords. You still have to

45:55check that the work was good. And you

45:57still have to have your computer on and

45:59make sure that you remembered to create

46:01that content on a day-to-day basis.

46:04Which,

46:05you know, in the long run of things,

46:07that might take you 5 or 10 minutes a

46:09day, which is like

46:11you know, 150 to 300 minutes per month.

46:14That's like nearly 6 hours a month.

46:16So,

46:18well, 5 hours a month, sorry.

46:20But, you get my point. Like,

46:22why would you spend 5 hours a month

46:24doing something that you could automate

46:26that would save you more time? And

46:27that's where schedule tasks come into

46:30play. So, schedule tasks

46:32basically allow you to run routines with

46:35AI

46:36so that something can run in the

46:37background without you,

46:40which means that it's fully automated

46:41and you never need to think about it

46:42again.

46:44So, if we have a look at this, for

46:45example,

46:46I said to ChatGPT, "Can you set up a

46:49cloud routine that runs without having

46:51the Codex app open in the cloud so you

46:53can generate content for my websites?"

46:54And then it asked me like, "Which

46:56website should it cover?" It should

46:57cover my five websites I own on Netlify.

47:00It says, "Where to set up a computer

47:01save the Netlify website?" I said,

47:03"Netlify." Okay. And then it created a

47:06cloud routine, as you can see right

47:08here.

47:09So, this is a routine that happens in

47:11the cloud. Now, you could have a local

47:14routine as well, but I don't like to

47:15have local routines running simply

47:18because

47:19you have to have your computer on and

47:21you have to have the Codex app open.

47:23It's kind of a limitation because, you

47:25know, yeah, not everyone's going to have

47:27their computer open all the time. Or

47:28sometimes you accidentally close the

47:29app. No problem. What you can do with

47:31this is that you have a cloud running

47:34routine

47:35that just runs in the background. I

47:37totally forgot I set up. I didn't even

47:38know how to access it for the first

47:40couple of days.

47:41And so,

47:42this is amazing for AI research, for SEO

47:46projects, for, for example, something

47:49that's recurring on a weekly basis.

47:51It It It would be amazing, for example,

47:54for anything that can run in the cloud.

47:56Now, the the one thing that I would say

47:58here is that it's not good for computer

48:01use. So, if it's like, for example,

48:03posting to X or something like that,

48:06probably wouldn't recommend it. If it

48:08was something more API related, like,

48:12for example, creating SEO content and

48:14then publishing it via Netlify via the

48:15API or the connector,

48:17that would be great. That's what it's

48:19designed for. So,

48:21have a think about all the recurring

48:23tasks that you work on data day that you

48:25could easily automate with this system.

48:27Cuz

48:28when you think about it, there's so

48:29much. There's no limitation to that, I

48:30think.

48:31So, you can easily create cloud routines

48:33for you that runs with Codex closed.

48:37It's got my SEO skills embedded into

48:40creating it. It runs on a daily routine.

48:42So, you can see here it's scheduled in

48:43for 7:00 a.m. each day. And then it can

48:45create five articles per website.

48:48And it's automatically configured to

48:50publish and it's ready to go.

48:53Now, if anything breaks on that routine,

48:55you know, the first time you set it up,

48:57like creating the apps, like creating

48:58the website, like creating the videos,

49:01you have to give it feedback in the

49:02first few days. But, the more feedback

49:05you give it, the better it becomes over

49:06time. So, definitely something to pay

49:08attention to right now.

49:10So, that's something else that's really

49:11cool. And then also we have something

49:13called plugins and connected. So,

49:17if you go to the top left over here, we

49:20can make GPT-6 Astra infinitely more

49:22powerful because we can connect it to

49:25other apps

49:26that allow us to take the power of Astra

49:30and not have it confined to this tiny

49:32little app. So, just like the cloud

49:34routines connect to Netlify, we can

49:36connect Astra

49:38to some of the biggest apps in the world

49:40to do recurring work for us. So, for

49:42example, it can read and manage your

49:44Gmail. And if you want to set up any of

49:46these, you just go to plugins in the top

49:48left inside Codex, right?

49:50Uh so, you go to Codex plugins.

49:53And then you can see here that we can

49:54switch between personal and public. And

49:57then we could, for example, set up Gmail

49:58here.

49:59So we can actually

50:01uh, try Gmail, like you can see. And

50:03then we can tag it in this workflow. So

50:06if we type in app

50:08and then Gmail, we can tag that in and

50:10it's ready to go.

50:11And then you could set up a new routine,

50:13like we were talking about before, where

50:15it analyzes or answers your emails for

50:17you. Where it drafts all the responses

50:19to your emails for you. Whatever you

50:21prefer. But my point here

50:23is that you can connect this to apps.

50:25They can run on schedule tasks. And then

50:28you can

50:29basically automate a lot of the, you

50:32know, all a lot of the work that you're

50:33doing day-to-day.

50:35Save so much time. Definitely worth

50:37doing.

50:38The other thing that you'll notice here

50:39is that we've got this option for full

50:41access. Now, if you want to limit the

50:45amount of access that your AI has,

50:48especially if you're coding on a project

50:50that you've already developed a lot of

50:52time on, or for example, if you have

50:54certain files on your computer that you

50:56don't want ChatGPT to touch, then you

50:58can switch between three different

51:00levels of approval.

51:01And so, for example, you can ask for

51:03approval twice, like so. You can say

51:06approve for me, or you can give it full

51:08access. Full access means that it's got

51:11unrestricted access. You can edit files,

51:13you can delete stuff, etc. You can also

51:15have approve for me. So if it identifies

51:18anything, like for example, deleting a

51:20file, then you can switch to approve for

51:21me. Or you can switch to ask for

51:24approval. And ask for approval is kind

51:26of slow. It's kind of annoying cuz you

51:28have to like approve everything it does.

51:30And but at the same time, if you want to

51:31play it more safe, that's probably an

51:32option too.

51:34We can also check our schedule tasks

51:36over here. So we can create new schedule

51:38tasks. These are running normally

51:42locally. So you can just say, "Okay,

51:44let's set up a schedule task." Maybe you

51:45check the latest AI news headlines every

51:48day, something like that. And then,

51:49you'll see inside the explore section

51:52here, we have options to customize.

51:55So, for example, for me, I don't connect

51:56this to pull to to GitHub, so I'm

51:59probably going to uncheck the pull

52:01requests.

52:03And then we can click on explore

52:05and then sites, like so. All right, and

52:07we can see all of the websites we built.

52:09So, you know we were just building

52:10websites a second ago, these are stored

52:13so that you can come back to them here.

52:16Now, also, if you're building out an app

52:18like this,

52:20you can have multiple different chats

52:22inside the same project where you work

52:24on multiple different things. So, for

52:26example, we could have one chat for UI,

52:28we could have one chat for adding new

52:30plugins, we could have one chat for

52:32setting up so it's ready to go with

52:33Xcode.

52:35And so,

52:36you can have

52:38the same project parallel conversations

52:40going on, editing different angles of

52:43the project. It could be like part of it

52:44is back end, part of it is front end, so

52:46that your app is much more useful over

52:49time.

52:50So, yeah, there's just there's so many

52:52things you can do with this. Now, you

52:53can use it inside ChatGPT on work,

52:57like so. I just I just don't

53:00feel like it's as useful there. But,

53:03what I do like to do is I have this

53:05agentic operating system, and any sort

53:08of custom workflows that I'm going to

53:09come back to, I like to build into an

53:12agentic operating system, into an agent

53:14OS mission control. If you're wondering

53:16how to set up a a mission control

53:18dashboard like this for all of the AI

53:21agents, where for example, you have

53:22Claude, Hermes, Antigravity, Codex, all

53:24built into one dashboard that you can

53:26control, you would literally, same

53:28process as I've talked about earlier,

53:30you would go into your AI agent and you

53:32would say, "Hey, build out a mission

53:33control dashboard for me. Bu- Build out

53:36a agent operating system for me that I

53:38can control." And if you want to get my

53:40system that's ready to go with all of

53:42this built in, you can get it inside the

53:44air proper board room. Just go to the

53:46classroom over here and then go to new

53:48daily updates and you can find our agent

53:50OS system that we keep updated. It's got

53:52a full video tutorial and a zip file as

53:54well to install.

53:55So,

53:56the cool thing about this is that we can

53:59use GPT-6 Astra inside our agentic

54:02operating system and we don't have to

54:04pay for the API or anything like that

54:05because we've already got it included as

54:07part of our subscription. And then we

54:08can use it alongside Claude and Hermes

54:10and anti-gravity. And I think that's the

54:12most powerful way to use these AI models

54:14because, for example,

54:16I mean, this is a big problem that I

54:17think most people facing with, right?

54:19It's like imagine, for example, you have

54:21and probably some people watching this

54:23will be like, "Yeah, this is me."

54:24Uh imagine, for example, you have

54:25ChatGPT and a subscription, then you

54:28have, for example, Claude

54:30and then you have, for example, open

54:32Claude and then you have Hermes and then

54:34maybe you have, for example, custom

54:36workflows and dashboards in other

54:39uh tabs. Everything gets very messy.

54:42Everything is all like all over one

54:43place.

54:44And the the thing is like then you're

54:46switching between tabs, you have all

54:48these different apps open, they're

54:50slowing you down, they're confusing you,

54:52you kind of forget where stuff is and

54:53it's all a bit overwhelming. That's the

54:55problem that the agentic operating

54:57system solves. And what we've done is

54:59we've built in codex into it. So, you've

55:01got GPT-6 Astra here and even if you

55:03don't use codex, um I don't know why

55:06you'd watch uh a course like this about

55:09GPT-6 Astra if you didn't have codex.

55:11But even if you didn't have codex, well,

55:13you can switch to Omni root or HY3,

55:16which are both free

55:18um AI models you can plug into this

55:20system as well. And then over here we

55:22have Hermes. So, what you'll find here

55:24is that we have GPT-6 Astra built into

55:27Hermes and I absolutely love this

55:30because

55:31what this means is that you have an

55:33agentic harness, which is Hermes itself,

55:36then you have the brain that plugs into

55:38it, for example like Chief 6 Astra. And

55:40then you have uh all of your different

55:43AIs plugged into this well. So, for

55:45example, we have Gemma 4 over here. We

55:47have, for example, Quen 3.7 over here.

55:50We have uh a custom trained

55:54version of Deep Seek over here. And we

55:56have like local models plugged into

55:58Hermes as well. I can just click on the

56:00drop down switch between all of them.

56:01That is so useful. And then also what we

56:04have is we have different versions of

56:06this Sugentic operating system depending

56:09on the workflow I'm trying to use. So,

56:10if I want if I need to do research, you

56:12know, I've been mentioning about uh

56:13doing scheduled tasks with research,

56:16well, we can go into Hermes with Chief 6

56:19Astra, and that just runs on a 24-hour

56:21loop, and it gives us the latest news in

56:23our industry. So, you'll see here, for

56:25example,

56:27we've got Chief 6 Astra.

56:30We've got the latest news here. It's

56:31inside one beautiful workflow. It's

56:33using Chief 6 Astra as a brain, and then

56:36Hermes agent as harness. And this is

56:38like a custom beautiful workflow inside

56:41one dashboard, where again it's just

56:43running on a loop on a schedule on a

56:44routine, and I don't need to be there to

56:47to check it out. I can also hear the

56:48brief here, so I can listen back to the

56:49brief if I want to. We can consult the

56:51article if you want to refresh Hermes

56:53article. We can actually

56:55uh

56:56take this the ideas. So, for example, if

56:59Chief 6 Astra is trending, then we can

57:01actually click on draft this, and it'll

57:03draft a post for Twitter, or we can

57:05click on publish to WordPress, and that

57:07will actually publish an SEO article

57:09directly to WordPress on that particular

57:11topic. And we can see previous news

57:13headlines

57:14from the days before.

57:16So, it's really easy to stay up to date

57:18and see how the industry changed. Not

57:20only that, but

57:22something else that you'll find with AI

57:24models in general, and I see this all

57:25the time, is that people struggle with

57:28contexts, right? So, for example,

57:33if a new model like Astra comes out or

57:35Claude comes out or for example, let's

57:37say, I don't know, GLM 6 comes out

57:40tomorrow,

57:42these models are probably not ones

57:43you're using day-to-day, right? They're

57:45brand new models that have just dropped,

57:46and sometimes that means you need to

57:48switch over to that new model. Now, the

57:50problem with that is it doesn't have any

57:51training or personalization or

57:53understanding of what you've done

57:54previously. So, how do you fix that? The

57:56way that you fix that is you have a

57:58second brain that plugs into all of your

58:00AI agents, and not just that, but your

58:02AI agents actually update it for you

58:04automatically. And then it becomes a

58:06positive feedback loop, where basically

58:08your AI agents update your memories for

58:11you

58:12based on what you're doing. So, you can

58:13see, for example, here, we're building

58:15out that app with Codex, and it's

58:18plugged it into our memory system.

58:20Amazing. And then,

58:22we have our memory galaxy here

58:24visualized,

58:26and we can see all of our memories

58:29inside one place. Super useful.

58:31So, what this means essentially is that

58:34if we do something with G6 Astra,

58:38well, that is updated inside our memory,

58:41inside our second brain, and all of our

58:43AI agents see that, and then they

58:46understand what we've been working on

58:47today as well. That is amazing. Super

58:49useful. It saves so much time.

58:52Now, if you're wondering how to do that,

58:53how to plug in system like this, you

58:56want to get Obsidian. Obsidian is a free

58:59memory plugin. It's open source, and you

59:02can get your AI agents to to update it

59:04for you. So, this whole

59:07uh database of memories, all these

59:08notes, they're just simple markdown

59:10files. I don't update it myself. I've

59:13never created any of these files myself.

59:15But what it essentially does is it

59:17updates everything for me.

59:19So, all of the latest files inside here,

59:22if we go to, for example, let's have a

59:24look here.

59:25Uh yeah, so if you have a look at this,

59:26um,

59:27example,

59:28Codex has updated Obsidian with the

59:31details of the pancake-eating game that

59:35we just built a second ago. And it did

59:37automatically while I was still talking

59:38to you. So, it's running in the

59:39background. It's stating all my agents

59:41and what we worked on recently. And then

59:43our AI agents can pull that as context

59:46from this system.

59:48So, it's free. It's open source. The AI

59:50agents manage it for you, so it saves

59:51you a lot of time. And then if a new

59:53agent comes out tomorrow, let's say for

59:54example, everyone stops using Hermes and

59:56then something more useful comes out, no

59:58problem, mate. Because what's going to

1:00:00happen is we can just take this memory

1:00:02and we can plug it into whatever comes

1:00:05next. And then it's always improving.

1:00:07And it doesn't matter what model comes

1:00:08out next as well. You know,

1:00:09GPT 10 comes out, no problem. It's going

1:00:11to have a full understanding of me, my

1:00:13life, and everything I worked on from

1:00:15day one. And that is absolutely amazing.

1:00:18So, that's what I'd recommend. And I

1:00:19think there's levels to this. So, level

1:00:21number one is that you're just creating

1:00:23one basic project. Level number two is

1:00:26that you're automating a few different

1:00:27things and running schedules. And level

1:00:29number three,

1:00:30the best and most liberating model I

1:00:32think is is basically the agent

1:00:35operating system and having Codex and

1:00:38Hermes and everything else plugged into

1:00:40one place. We also have custom

1:00:41workflows, like for example, that video

1:00:43agent that we built before.

1:00:45So, we've talked about Hermes, we've

1:00:46talked about routines, talked about the

1:00:48agent operating system, and we've talked

1:00:49about where to start, which is basically

1:00:51you want to start where you're spending

1:00:52your time.

1:00:53And that's basically it. That's the

1:00:54whole system. We've talked about pretty

1:00:56much how to automate everything.

1:00:58You know, how to get access, examples,

1:01:00automations. We built a few things

1:01:01today, apps, SaaS tools, website design,

1:01:04images, videos, computer use. We've

1:01:06talked about game design with Blender as

1:01:08well, routines, Hermes, Agent OS, and

1:01:12also the plugin systems as well. So,

1:01:15that is basically it. That is the full

1:01:17GPT-6 Astra AI full course, 1 hour on

1:01:20how to build and automate anything. If

1:01:22you want more training on this sort of

1:01:23stuff or if you want to connect with me

1:01:24personally, you can do that inside the

1:01:27AI Profit Boardroom. Link in the

1:01:28comments and description or just go to

1:01:30the AI Profit Boardroom.com. Inside the

1:01:32community, you can ask questions, get

1:01:33help and support whenever you want to.

1:01:35So, you can see for example over here,

1:01:38we have a full

1:01:41community where you can ask questions

1:01:43and I personally answer these questions

1:01:44every day. Plus, if you have like a

1:01:47question that you're struggling with,

1:01:48for example like this, you can see that

1:01:50I will create a personalized video

1:01:52tutorial for you and give you my notes

1:01:55breaking down exactly how to handle that

1:01:56situation. So, the great thing about

1:01:58this is you get everything you need to

1:02:00win. Plus, we all help each other out.

1:02:02We will share wins. We will um help each

1:02:04other out and you know, give each other

1:02:05answers and

1:02:07try and give each other advice that's

1:02:09useful. And that's fantastic. Also, have

1:02:11a cool stuff like for example, we

1:02:13actually just dropped a new course

1:02:15inside the AI Profit Boardroom and I do

1:02:18a weekly update. So, you can see this uh

1:02:21video update here where basically I help

1:02:23you as much I can and I just break down,

1:02:24you know, what's useful this week and

1:02:26what's actually helping.

1:02:27Inside the classroom, you can get access

1:02:29to all of our best trainings. So, if

1:02:31you're a complete beginner, you can

1:02:32start with the beginner to expert course

1:02:34in 2026 and we break down step-by-step

1:02:37how to get more leads, how to get more

1:02:38customers, how to grow your business

1:02:40with AI.

1:02:41We also got a full section where you can

1:02:43start and learn anything you want.

1:02:45Inside the new daily update section, we

1:02:47have advanced tutorials. So, if you're

1:02:49more than a beginner, if you're if

1:02:51you're advanced and that sort of thing,

1:02:52you can check out our advanced

1:02:54trainings. And then if you have any

1:02:55questions or if you need personalized

1:02:57advice, you can post inside the

1:02:58community, too. We also have all of our

1:03:00newest and latest trainings as you can

1:03:02see right here. So, you get access to

1:03:03all of these new trainings and courses,

1:03:05which is pretty amazing. And we have

1:03:06four weekly coaching calls where you can

1:03:08get help and support in real time.

1:03:11So, the great thing about this is

1:03:14the you can jump on this course, you can

1:03:17ask questions, you can share your

1:03:18screen, you can meet other members, you

1:03:20can see what they're working on. It's

1:03:21just an amazing community where we all

1:03:23learn and we grow together. Inside the

1:03:25map, you can meet people in your local

1:03:27area who are building out with AI agents

1:03:29like you. And this is all available

1:03:31inside the AI Profit Boardroom. Link in

1:03:32the comments description or go to the AI

1:03:33Profit Boardroom.com. Thanks so much for

1:03:35watching and I'll see you in the next

1:03:36one. Cheers. Bye-bye.