Transcription
Today we are going to be doing a 1-hour course on how to build and automate anything with GPT 5.6. So let's just get straight into it. If you're not familiar with chat GPT or GPT 5.6 itself, this is a new model that just dropped and you can see an example of it right here in our dropdown. So you can go to chatgpt.com to get access and then if you click on this dropdown over here, select model. You can switch between GPT 5.6 and you can also switch between different models. So for example, you got GPT 5.6 sole. Uh, you have pro, extra high, high, medium, instant, etc. Instant is actually an old version as you can see. So that is GPT 5.5 and that's exactly how to get access to GPT 5.6. We will come on to stuff like codecs in a second, but that's essentially how to prompt it.
Now, when you're using chat GBT like this, this is basically just like a a back and forth very basic chat. Now, chat GBT itself has all sorts of powerful stuff that it can do. Uh, let me show you an example. So, let's say for example, we wanted to create content for our X account, right? Maybe that's something that you want to automate with Chat GBT. Well, what we could do for example is we could take all of my best performing content like so from X we can paste that into chat GBT. So if we copy and then paste that over here and then from here what we can actually do is we can say okay create an X account or create an X post about and then you just insert the topic that you're going to create content around. So let's say for example it's about the new GPT 5.6 soul release. So what we can do here is automate our social media by creating an X post. And the way that we can do it is train the AI on content that we've created previously. And then from there we can start creating the content itself. So we say create an X post about the new GPT6 so release. Use a similar style to below. You don't need to format anything when you're pasting in here. Like the AI is automatically going to understand everything. So then what we need to do is just make sure that we switch to the right model. So I'm going to use GPT 5.6 Soul for this. I'm just going to use medium, right? We're going to use medium with GPT 5.6 Soul. All right. Um, and then from here we can just click command and enter if you're on a Mac. And then it will actually send the post.
Now you can see it also has access to the web. So the great thing about this is if we're posting about something new and interesting when we're doing this essentially we are teaching the AI about content that we've previously created and this could be for research it could be for content it could be for emails whatever you want to automate and then from here we have the post itself generated like you can see and then we have the previous research that's done so what it's actually done is search a web based on the latest research has just come out out about this model. And so we have the content automated over here. And then we have the social media post over here. Pretty powerful. And so now we can actually automate content in my exact style just by training it on previous data that we've had. From here we can either edit the post. So we can edit it and you can actually ask for changes. So you could say, okay, make this more punchy, make this more direct, make this more interesting. And you can copy this and paste it over to Twitter if you want or wherever you're going to post social media content. You can go full screen on the editor if you want to as well. And then you can X and close that off as well.
Now, from here, you'll see how it adds hashtags, which is kind of annoying. It shouldn't have done that really, but we can just delete that and edit it before we post it. Then, if we wanted to create a social media post, we can just go over to X like so. And let's say, for example, I wanted to create a post about this one. I can just click over here and we can click on quote. Then I can paste in the content that we have and that is pretty much ready to go. It's in my exact same style. There's some things I would probably be aware of. Like for example, you can see this hyphen that I would never personally use, but it's written in my style and is completely automated in a couple of clicks using this system. So you can see how easy it is to automate uh content itself using GT5.6 and that's exactly how to start using it straight away.
You can also share conversations. So if you have a team for example, then you can share everything that you've created with other people. So let's say for example we had a conversation like this. Maybe it's something funny you just want to share with people. You can just click the share in the top right. You can also move that to a project which we'll come to later. You can pin the chat, archive it, delete it, and you've got the whole setup right here. So, pretty easy and simple to use. Easy to set up, works really well.
And again, GPT 5.6 just dropped. If you're not familiar with GPT 5.6, let me talk you through the benchmarks and everything else. So, as of the time of recording this, this is 24 hours after it's dropped. So, it dropped on July the 9th, and GPT 5.6 is the best model, the most intelligent model that's come out from Chat GBT recently. Now, when you're looking at announcements like this, you will get a bunch of benchmarks and you'll get a lot of like marketing hype as well. So, I always like to test this stuff out myself. I wouldn't really listen to uh anyone else when it comes to the benchmarks. And I would advise the same for you because, for example, you'll see quite often there's a huge difference between what people say and what actually is out there and and what's in reality. So, always test this stuff yourself. But so far from my own tests, I've actually seen the GPT 5.6, particularly Soul, is really, really good. There's another company called Anthropic and Claude if you're not familiar with them and they have a rival AI and that is called Fable 5. Now Fable 5 is considered the best available model out there for AI. Um, with that model you can basically get the best outputs possible out of all the models I've tested. However, I would say the GT 5.6 Soul is right up there. And if you're wondering like how does it compare? You can see the comparisons right here.
So if we have a look at this GT5.6 sooul is performing pretty well. There's also different effort levels with soul and so for example you could have it on pro extra high medium high instant etc. Now between these models when you're using something like chat GBT or even codeex which again we'll come on to later when you're using these models the difference is that you can run out of tokens very quickly if you're using for example the pro and also the responses will get slower. So it's really a balance between what are you doing for that task and you want to basically tailor the model that you select based on the task that you're creating content around. So for example, if you are doing something that is just like a back and forth in the chat, you could even use GPT 5.5 for that on instant and that's not a big deal, right? Whereas for example, if you're coding out like some crazy SAS tool, for example, we have a agent operating system for all our AI agents inside the air profit boardroom. When you're creating something like that, you want the most powerful AI possible, right? You want it, you know, at the the highest level it can possibly be. Pro I usually don't use, however, because it's just really slow when I've tested in the past. I might have changed recently, but I think sometimes that used to take like, you know, 10 to 15 to 20 minutes to apply at times. So, extra high is probably where you want in terms of the the best output. So, depending on what you're doing, adjust the level of effort that you are using.
Now when you look at charts like this as well you'll see something called API cost. API cost is a different way of using chat GPT. So if you are using chat GPT directly here instead of coding out with it and getting the API for it then you don't pay per usage. You just have it all inside the plan that you have. So we can keep going from there.
Now, one thing that I also say about chat GBT as we go along is that to get the full power out of a system like this, you wouldn't really use chat GBT directly. However, I want to start with the beginning stuff and the fundamentals first, the simple stuff and then we can move on to the more advanced stuff later. So, if you're watching this and you're like, okay, uh, you know, I want to do more advanced stuff with it, we'll come on to that later. If you're a complete beginner, then we're at a good starting point. So this is some of the most powerful AI uh some of the most powerful intelligence out there right now. What you also see is like a lot of interesting benchmarks. For example, uh long horizon is mentioned a lot. That's basically where you can give an agent a task and it can work for hours if not days on the same task and just carry on to completion until it's finally at the standard you want which again we'll come on to later. You can do that inside codeex. So yeah, these are all the benchmarks you got for example like terminal bench, you have browse comp, you have sec bench pro, etc. And yeah, it's a it's a really powerful system. You can also create beautiful websites with this. You can create like nice spyroraphs, interactive wave interference. These are all interesting ways of using AI as well. Let's have a look what else we got down here.
Quite often as well when you're running through all of the technical stuff here, it's really interesting to see the actual report and to see what sort of tests they did and how it works. And often you'll see when you're comparing models uh comparison in a table that looks like this. And so these are basically relative models are being compared. So you can see here that soul is performing the best out of all of these. And when we compare these side by side, for example, you'll also see the Terra doesn't quite perform as well as Soul. And then Luna is the fast model that is cheaper and faster, but doesn't perform as well on benchmarks. But you can see that even Luna, which is the lowest spec model from GT 5.6, is still being uh still outperforming GPT 5.5. So pretty powerful stuff. So these are professional benchmarks in terms of how it performs for example for like management consulting agents last exam artificial intelligence index then you have coding evaluations then you have the science and health and then you have computer use agents as well. Now computer use is basically where the agent can control your computer which we'll come on to later. Then you have for example like security benchmarks interesting stuff on self-improvement multimodal academic so like research related topics tool use and long context as well so that's all the benchmarks that's how it performs again I would put it right up there in terms of being on par or just below Fable 5 from Claude and then for example you have something like Gemini which I would say is way behind uh Gemini is from Google that is another model so we've covered how to use chatb 5.6 how to choose the effort levels what the models look like what the three different models are if you want to see more information on that I actually have this website where we've broken it down step by step so the way that I look at this is it's a 5.6 Six is the solar system. There's three different models inside there. So you got soul, the sun, the flagship. That's the biggest, strongest, most expansive brain. And you want to use that for like the hardest, trickiest tasks, the stuff that requires the best levels of thinking. Then you have Terra, the Earth, the workhorse. This is balanced in the middle. So a lot of serious workflows can be done through this. And then you have Luna, which is small, fast, cheap, great for high volume stuff. uh a lot smaller as a model, a lot cheaper as a model, but great for getting like quick instant answers. So, that's basically how it works.
Now, you might be wondering, okay, what sort of stuff can you build with this in general? Let me show you some of the things we actually built with Saw. So, for example, this was like an open world dragon game, as you can see. We also, for example, have this Doom game as well. So, you can build out really nice games with it. I like to test it with games because when you're testing out a new model, it's really good to see, okay, how does it perform side by side versus other models? And we test and compare all this stuff on Goldieb, but then also how does it look visually? What's the vibe like? What's the ambience like in the stuff that it codes out and I think that's really useful to see because it gives you a quick sort of way that your brain can look at something and go, okay, right, that's great, that's terrible, sort of thing. So what we can also do inside GPT 5.6 is we can also create images.
So for example here this is a prompt that I often use for for generating images. And then for example we can say okay create an image of a cat eating cheese. And if you want to like make it more interesting I use this prompt here to make it more interesting. So we have a basic prompt for creating a uh image and then we have some ways that we can add for example exciting, fun, crazy words like these to make the image stand out more. And then also if you want to generate multiple versions of the image that are very similar but slightly different for example slightly different colors or more contrasting or different titles etc. then I use this prompt as you can see right here. Now once we're happy with that we can click enter and it will start thinking and generating the image. Now the cool thing about generating images as well it kind of runs in a sandbox which means like it runs in its own environment inside the cloud which means like when you're doing this basically when you're generating images you can close the tab and then come back later to it. So if we add an image over here I'm just going to put an example like this. add that in. Plug this in and generate the image. Now you might be thinking okay does is it GPT 5.6 that's creating the image. So it's a combination where it does the thinking and reasoning with GPT 5.6 or whatever you selected as a model but then it's using something called chat GPT image 2.0 to generate the image itself. So the image itself is generated with that image model but you can use 5.6 6 to prompt it and get better outputs. And I've noticed that there is quite a big difference when you use for example like GPT 5.6 versus GPT 5.5 for generating the images itself because it's reasoning through the process. Right? You can actually click on this bit here and it shows you the reasoning on the right hand side. So you can see what it's doing and what it's working on.
Now if you want to do two different workflows at the same time with GPT 5.6 six. Then you can have a tab like this open and then you can go to a new tab over here. And what you'll notice is if you click the plus sign on a new chat here, we can select between different apps and different workflows. So if we want to create an image, we can select that. If we want to search the web, we can do this. You can kind of use it like Google and you can say, okay, web search, search this, search that, blah blah blah. You can also do deep research. Now deep research is not something that I usually use honestly at this point. I don't think it's that useful. But let's say, for example, you were trying to write a report. Well, then it could be very useful. If you were trying to create something very in-depth and it's a one-off task or it's something where you need loads of sources, then you can use deep research combined with GPT 5.6. So for example, if we type in deep research and we say okay, research the latest trends or research the best methods for using Hermes agent for AI automation. What it's going to do is it's going to take quite a long time to generate that deep research report. It may ask us for more clarifications whilst it plans out what it's going to do over here. And then you can see here that it's ask actually creating a plan here that we can edit before we start doing the deep research. So when we're doing deep research with chat GBT, we can see the plan before it gets started. Now this is important to check. It will actually autocomplete if we don't do anything for 30 seconds. So you can see a countdown over here. We can cancel it and we can also edit this. If we can edit it, then it actually stops the counter. We can give it a follow-up plan over here to change the plan and we can go from there. Now, if you're happy with the plan, if you're like, "Okay, I've read this. It looks pretty cool. This here's what it's going to do step by step." So, it's going to survey official herization documentation, review the community tutorials, blah blah blah. Then we hit start and that actually start researching the background.
And so, what I like to do when I'm using GPT 5.6 and any sort of AI really is I like to just pin the tab, especially if you're on Chrome, minimize it so it doesn't get in the way of everything else I'm doing. And then we can move on to the next workflow. This is super useful because when I'm doing multiple workflows, I know, okay, the images might take like, you know, sometimes 60 to 60 seconds to like 5 minutes to generate. So, I'll just leave that running in the background. And then for example the research, well that might take like 10 15 minutes as well minimum. So you can see it's gradually working through the checklist. You'll check everything off as it goes along. So when we're doing stuff like this, just you you can leave it running in the background, come back to it later, pin the tab so it's easy to navigate to and you don't forget about it. And that way you can still get lots of stuff done, but you don't need to be distracted as you go along. Now you can also see here that we have multiple images generated from that previous prompt and we can click this button to get it. We can share with other people. We can change the aspect ratio. So whether it's for example like portrait, story, landscape, widescreen etc. And we can click through between each one. But if you look at these with that prompt that we generated before with that prompt, it's quite easy to create beautiful images that look better than designers. I mean, for example, when I've used teams of designers, number one, they're never going to get the design done as quickly, and number two, it's not going to look as beautiful as this. And you see how it's like change the colors, the layout, the styles, etc. Looks really nice.
Well, the beautiful thing about that is that you can automate the sort of the way of testing different ideas in a way where you can come back to it later. You can get many different variations in different styles. Pick through which one you like the most and then you're good to go. And you can do that for portrait images or whatever. You can also comment on these as well. So you can add comments. I don't really use that so much. And then if we have a look the deep research is being done in the background here. So you can see it's like doing loads of searches like 119 sources searched and that's in the space of like two or three minutes. We can also update this as well if we want to. And you see how it's ticked off the tasks it's done so far. That's really useful as well. So, so far we have automated our social media. We have automated image creation. We've also automated the process of doing deep research. So, that's a really cool way to use chat GPT.
Another way that you can get more out of GPT 5.6 is you can use projects and you can also pin stuff. So for example, if we have a look over here, by default, this will be hidden. But if we click on open the sidebar, we can see our previous history of what we've done. We can also rename the conversations and we can pin our conversations too. Now we can switch close the chats here. So we can tap on that. If we want to start a new chat, we can click on this. And then over here we have our list of projects. So we can create a new project and go through anything that we want to whenever we want to. So for example, if we have a look here, we've got different conversations of the different projects. At this point, you may be saying, okay, what is a project? A project is kind of like a way of organizing all your files and all your systems together with GBC 5.6. So if we have a look at the AI powered SOP machine, what this essentially does, it creates SOPs for my team, step-by-step instructions I can train my team on. And so that's a great way to automate training resources. You could also, for example, automate courses. If you did a lot of images, well, you could actually have a folder specifically for images. And then you can come back to that project anytime you need to. So if we have a look at this project here, we can share it with other people. We can also rename it. We can switch between the sources. So we can add sources. And this is a way of adding training to the AI so it understands everything about you and what you're trying to do. Adds context because for example if you just use chat cheap straight out the box it doesn't understand a lot about you. It doesn't know what you're working on. It doesn't know for example your brand your voice your projects your businesses your goals your websites. There's a lot of information it doesn't have by default. And so when you add these sources and you upload for example we've added our sales training over here. Then by understanding that it can create better recipes in the future. It can better uh understand who we are. It can use that as context to create better um setups in the future. The same for example with that social media setup that I showed you before with that social media process that we created earlier. Well, we can actually use that in the future to make sure that we have a project where we can automate social media content whenever we want to. So, we've got the chats, we've got the sources, very simple. And then over here, if we click the three dots, we can click on project settings. So, we can change the name and we can add custom instructions to this. So, when we're adding custom instructions to the AI here, what you're essentially doing, and we can drag that down so we can see the full set of instructions. What you're doing essentially training the AI on how to respond every time you use it. Now, you might say, "Okay, but why don't I just uh prompt it every time or why don't you use those prompts every time?" The problem with that is that when you're reprompting over and over again, you could waste like 20% of your time just retraining the AI on exactly how to do stuff and that is quite time consuming. It's quite draining and it's more efficient doing it this way where you have the instructions inside a project that's ready to go whenever you come back. You can also change the memory over here. And you can change what it has access to as well. And then you change the name. You can also add like a color and emoji to that project as well. So it's easy to navigate. Then you can change the effort level just like normal. And you can switch the models as well. Over here you can see previous conversations inside that project. And that's basically how to set up a project.
Now you also have something similar to that called the custom GPT. So if we go to more and then GPTs over here, you can see that this is basically like a whole different market of AI automations ready to go straight out of the box. And if we have a look for example at the scholar GPT just as an example here, we can start a chat with it. So we can use other people's automations and then we can use this for better research. So this one is specifically for research and that sort of thing. And we can create a new chat. We can pin this so we save it for later. We can review it as well. So we can leave a review for it. And we can chat with it as well. What you also see when you're using this is that you can switch between the thinking effort and you can also add files. So if you want to add like files that give the AI context. Let's say for example I'm creating a research paper for this scholt and I'm like right okay well here's the brief of the project based on that brief make sure that you create better research so there's just so many things you can do with this it's pretty much endless with GPTs and these sort of things and you can search through for anything you want so if you do a lot of SEO for example you can search for SEO over here and it will search the marketplace for SEO related automations. If we click on my GPTs, I've actually created loads of GPTs that are custom made and I can see how many people have actually had a chat with it, how many people have used it, how often they use it, and who can access it. So, we could create a custom GP2 over here. We could call that social media content. And then over here, we could give it custom instructions on how to respond. So we could say for example, write short, punchy, direct. You can get as detailed as you want. And you can actually test over here before you make it go live. Then you can add more instructions. You can add a description of what it is. And also you can add conversation starters over here. So if you add a conversation starter, it could be like automate or write a new social media post. And then we can add knowledge inside the GPT as well. So we're building our own automation that we can come back to bookmark. we can share with our team, share with the world. And this can automate anything. It could be automating social media, could be automating research, it could be writing instructions for your team, it could be just generating nice images. Right?
So, as an example of this, if we went to my X account and we use this for training so that my AI understands how to create content for me, I can paste that in over here. And then we can use this as a training so that our AI understands exactly how we write. So now I can say okay based on the knowledge that you have right social media posts so that I can automate them quickly in my style. I'll give you the topic, you do the research and create the post for me. Right now from here we can also change the model. So we could switch the model to GPT 5.6 as you can see. So we can switch over here and then we can change its capabilities. So we could say okay it's got web search, it's got canvas so we can edit the outputs. It's got image generation if we wanted to create the outputs as well. And now we've got a custom trained GBT that we can use to automate images. Now for me personally, if I have more time, I would probably spend about 5 minutes crafting the instructions here because I know if I get it right once, then in the future, it's going to save me more and more time. The better it is, the more likely I am to use it. So, we can click create over here in the top right, and we can share this. So, we can say, okay, only me has access or anyone with the link has access or we can even add it to the GPT store. It's kind of like the iPhone app store except that instead of creating an app, you're creating a custom GBT which is a AI automation. You share it with the world. Pretty nice. So, if we click on any with the link, anyone with a link can now access that post to create social media content. Now, on the right hand side here, we can test it and then if it's not very good, then we can iterate it later. So, whilst that's saving, you could be thinking about, okay, let's test it with a couple of posts and let's go from there. Now you'll see that it gives you the link to open that GPT and this is the view mode. So before we had the edit mode, this is the view mode. So if we go inside this chat here and we're like, okay, create a social media post about GPT 5.6, right? It's going to search through the knowledge that we trained it on a second ago. It's going to search through the web and get the research as well cuz we enabled that. And then we've got the whole post here and it's written it based in my exact style. Pretty nice headline as well. That is classic Julian Goldie style headline to be honest. Then we've got the bullet points over here and we have a nice little um setup here as you can see. And then we can copy it and again you can post that to X or whatever you want to do.
So that is GPT 5.6. We've talked about projects. We have talked about automations. How to build your own custom automation share with the world. how to generate images, how to use the chat, etc. One thing that I would say inside the chat as well. This is something that helps me a lot is there's going to be some conversations where you're like, do you know what? I need that conversation. I need to come back to it a lot. As an example of that, I've got one where I actually use voice mode and I chat with the AI in Italian and we just go back and forth so that I can learn a bit more Italian and that sort of thing. And so what I can do is I can pin that. We can we can pin that and then we can come back to it later. So it's kind of like a bookmark but it's a bookmark for GBT so that I can easily navigate and find stuff that I want to use later.
Now over here you're going to see the scheduled tasks right these scheduled tasks these are ways of basically scheduling a task and then coming back to it later. Now you might say why would you do that? Well let's say for example there's something you do every single day. Could be for example like posting a blog. If you do that every single day, you might as well automate it and then have it running in the cloud so that you can come back to it later. Now, they have a bunch of examples and templates that you can use right here. And then over here, you can schedule tasks for later. So, you could say, "Okay, every day, uh, 2:45 a.m. do some deep research for me." So that when you wake up, the deep research is already done, and you don't need to wait for all that time for this to be available. So you could schedule in tasks run in the cloud like having a 24 assistant and then you can come back to it later.
You can also do voice mode with GPT as well. So if you click on the voice mode here, this has improved substantially I would say over the last week they just had a big update inside OpenAI and chat GPT for voice mode which means you can get way more out of it as well. What that means essentially is like smarter responses, faster responses and it works way better. I'll give you an example. So when I used to tell it to test me in Italian, it would I'd have to ask it twice every time. It didn't make any sense, right? Whereas now, for example, when I'm using this system, it's quite easy because it would just test me straight away. I don't have to retrain. It's just more intelligent when you use it. And we've got the schedule tasks over here. We can come back to it. We've also got a separate tab for work and for chats. Well, this depends on what account you're on. So whether you're on like a personal or a workspace account. So we've got the schedule task there. Then we also have plugins. Now with plugins you can basically connect chat GPT to external apps. So why is that useful? Well, let's say for example you wanted to work across Google Drive and take the outputs that you have and plug them into Google Drive. You can install the Google Drive plugin and from there you can use it later. So let's say for example we were talking about blog post before with blog posts it could be like right okay schedule this task every day where you write a blog post for SEO about my topic and then plug it into Google Drive for me. So if we click on install plugin over here we can connect to Google Drive as you can see we can select the Google Drive they want to connect to. So I'm just going to connect to that in the background. Um we're going to give it access to that now. And that is how easy it is to connect Google Drive to Chat GPT. Perfect. Now, if we type in Google Drive, we can work across it. We could say, okay, create a new blog post about SEO just as a random example. And now it will start coming back to us and it will say I'll create a polished Google doc with an SEO focused blog post, clear headings, practical steps, and a strong CTA. So you can see an example right here.
Now also when you're using something like Chat GPT, you're going to find like it makes mistakes and you're going to find ways to get better outputs from chat GPT in general. So if you look at the prompt that we had earlier which is this one this is a nice prompt for generating images when we're using this prompt very in-depth very specific in terms of how to use it. Now basically every time I find a mistake when I'm using chat GPT I will then change the prompt and save the prompt for later so that I can come back to it and every time it makes another mistake we'll improve it again and again and again and again until by the 10th or the 15th time of improving that prompt is actually really really good and it's a custom template we've got ready to go. So the way that I would recommend saving that for later and this saves me a lot of time. So, for example, if I type at and then we have a prompt like this, it can easily type that out automatically. You might be wondering how I do that. So, I actually use something called Text Expander. Text Expander is great for just saving my prompts for later. And then I've got them as shortcuts ready so that I don't have to type them out or copy and paste them from a document. It's like I'm in the chat. Boom. Let's just type that one word out and we've got it ready to go. And so, inside text expander, you can save your prompts. You can add a new snippet like so. and you're good to go on that. That's how easy it is. Would definitely recommend using them. So, you can see now that it's created the Google Doc in the background. As you can see here, we can open up and boom, Shakalaka, we have a Google doc with the whole blog post in it ready to go. And so, the great thing about this is you can link chat GPT. So, it actually does work in the real world, not just inside this tiny little sandbox. that is a chat GPT conversation. So you can link everything together, you can get it working together. It's quite easy to set up and it doesn't take a long time which I think is is way more powerful than anything else that you can do with this system.
Now from here we've created images, we've created uh tasks, we've linked it to apps, we've created custom GPTs, we've learned how to use a voice mode, we learned how to pin conversations and save them for later. And you also learned how to create custom projects, custom GPTs, schedule tasks in the cloud, etc. You also have the library here where anything you create you can come back to later. Honestly, I don't like it that much simply because I think it's quite hard to come back to and it's kind of annoying and that sort of thing, but if you want to check it out, you can. And it's got like your files and images and everything else you've created. Then you can search across chats over here. And from there, we are going to start moving through to codeex. And codeex is where you're going to level up massively, my friends. Because really, you've only tasted about 10% of the power of GPT 5.6. With codeex is a much more powerful way to use agents essentially. So, if you look at chat GPT, it is literally a conversation. It's like a chatbot that's just very very smart and it can link to apps and nothing else and that's great. But you can get so much more out of all of this when you start understanding how to use codeex. So you can get it as you can see right here and you can just download it and then install it once you're ready with that. Once you've done that you can use codeex as you can see over here. Now with codeex this is a way of using GP51.6 ICS in a system where you can code and build anything out. So like we were talking about before, you got these custom projects over here and you can actually set up a new project anytime you want.
So if we click on new task over here and then we are going to give it access to a folder like so. and we can change the folder over here. So, we can switch between all of these different guides, folders, everything else. So, if we click on that project, for example, we're good to go on there. Now, you've also got something called sites. So, sites allows you to take your idea and turn it into a website just by typing in the prompt. So, if we have a look at this one, this is a website that we've created inside the sites tab and we can share it. we can preview it etc. So if we click on the share button here, we can click on like that. Like so. Publish it. Wait for it to publish. It will take a couple of minutes. And then we can copy and share that with anyone as well. So we'll come on to all this and exactly how to use it in a second. So, let's delete that. And also, what you'll see when you're using codeex is you can switch between ask for approval, approve for only me, or full access. If you want to play it really safe and you don't want like a lot of people watch this and be like, okay, I don't want some agent taking over my computer, that is totally fine. So the way that you can avoid the agent doing anything you don't want is you can set up ask for approval and then every time it wants to edit something or use the internet you approve it and you give it permission. So you control all the permissions. If you do approve for me this only asks for actions detected as you know for example destructive that could be editing a file that could be deleting a file that sort of thing. And then we have full access. So, this is unrestricted access to any file on your computer because I'm very comfortable with doing this. I'm going to go with full access cuz it will speed up my project. But if you don't want to do that, you can just select ask for approval. And then from here, we can select a new project. So, we can say okay, let's start from scratch or let's use an existing folder. So, we're going to start from scratch and we're going to create a new project in Codeex where we'll build something cool in a second.
So from here we're going to type in okay let's create a website for the AI profit boardroom community and whatever project you want to build you can build with this. You can also switch between local and cloud. So with cloud this will actually work in a sandbox in the cloud. If you select local this will work locally on your computer. Sandbox in the cloud means like it can't work or have access to your local files. It's going to create something in the cloud. You can access it later. So that's kind of like a safer way to work where you work in a cloud environment and then working locally it works directly on your computer. You can also see how much usage you have remaining. So how much usage you have left to use which is great. This is good because then you can see how many tokens you've used. And the more you use AI and the more powerful the workflow and the more detailed the instructions you give it, etc. the higher the effort levels that you use, all those things use up tokens. And the more tokens you use up, the less like the less time you can actually use it for. You're going to get less out of your subscriptions. So, I'm going to go working with locally. We've made sure we've got soul selected. If you want something really smart, then you can go with so extra high. If you want medium, you can go with that. If you want something like really instant, you can actually toggle this and it will switch between the different models of uh 5.6. 6. So you can see for example, Terra is light. If we switch to medium, then it's automatically set up to use the model soul. So you don't need to switch between the models. It's just automatically set up for you depending on whether you want faster or smarter. Which ones do you want? And again, you'll use usage limits faster if you select smarter. So we can just go with medium as you can see right here. You can also go to advanced and then you can actually change the model and everything else. But for most people, if you're watching something like this, you're probably just going to go with the defaults. You can also switch between fast and standard. So, how fast you want it to reply. Now, obviously, if you select fast, it's going to move 50% faster, but at the same time, it will use more tokens. So, again, you see how like when you're using stuff like this, you want to be efficient when you're using it. If you are not efficient, you will run through tokens in a way that you know uses up a lot of tokens and usage and that sort of thing. So from here what we are going to do now is we are going to start typing in our project. So if we say okay create a beautiful website for the AI profit boardroom. We'll just hit enter as an example. And you can see that it gives us the details of the environment. It's now going to plan out what it's going to create. It's going to use different skills depending on what we want to build with. And I can come on to skills later and then it's going to start coding that out. Now the cool thing about this is it can work run later. So what this means is it can run in the background when we switch between tasks. So if I switch over here, guess what? That's agent is still working in the background which is fantastic. And also we
can pin projects. So if you have certain projects you want to come back to, you can pin them. If you are coding locally and you want to see the actual files themselves that have been generated, you can click on reveal and finder. It'll take you directly to the website. If you click on allow, you see how it just asked me for permissions there. I'm going to allow it so it can edit stuff and see that sort of thing. Now, you can see it's asking for details. So, just to save time here, what I'm actually going to do is just paste in the information from this page so that it's fully trained on exactly what we want. So put here's the full details and then it's trained on what we want to do and everything else. We can also chat with chat GPT in the background as well. So the cool thing about this is that we can have a project here that we're coding with but then we can also use the chat history from chat GPT. See our recent chats over here. Uh so for example when we created that blog post a second ago it's over here. So all our previous chats are right there from chat GPT. So if you're like oh I want to use the functionality of chat GPT inside codeex you've got it over here and you can see your view you can view your chat history you can see where we created the cat with cheese etc and we can minimize that whenever we want to now you can see the timer has started so it shows us how long it's been working for what it's building out and how it's growing from here and then that is our project being built in the background now we also have for example if we close that we have the pull requests feature here. This is more for GitHub. For most people watching this, you probably don't need that, but if you do want to sync your projects to GitHub, you can set up GitHub over here. And then in the bottom left, you'll see your name and your profile. Over here, you can see your usage. You can see how much usage you've used. You can also reset your usage, too. And that basically just gives you an idea of, okay, how much have we used? How many tokens have we used? And this just counts down. So it shows you, for example, I've got 99% of my usage left to use. We have a pet that we can use at any time. So for example, with the pet, we can see what's being worked on, and we can click back to the conversation. So all the stuff that's being done in the background, we can use with the pet. So if we click on that, we can see what it's doing. If we click on the create the boardroom, we can see, okay, we can go straight back to that. So it's just an easy way to see what it's working on and then also to navigate back to chats. So, if I was working on five projects, we'd have five pop-ups here with the names of each product. And also, you can see that the circle here is buffering. So, we can come back to this later and see what it's working on. So, we have that working in the background.
And then we have the site section here to create websites. So, you can turn your ideas into websites and that sort of thing. If you want to create a new site, you can just click on create a website at. As you see, if you click on sites, create in the top right, then you can say, okay, create a website that, and you just give it your idea. So, it's like, okay, create a website that tracks my habits dayto-day, for example. Let's type that in. Then, if we hit enter, that is now going to create a whole system for sites that we can use later. And then, if we go back to sites, we will see the project being coded out in the background as well, which is pretty useful. Then then we have the plugins. So just like we talked about before plugins and one thing that I think is fantastic is really really cool to use is that we can actually use computer use. So if we go to plugins over here and then we click on the featured we can click through and see what's been done, see all our apps and see what we're working on and everything else. So we've got everything that we've installed at the top. So for example, Google Drive, GitHub, all these other ones. And then anytime we want to use one of these plugins, we can click on try it now. And what this means essentially is like you can have multiple different plugins being used. You can use the power of like a coding agent, which is what we're using right now to create everything that we want. And we can use the power of of computers as well to test this out. So if I say to computers, okay, go to google.com and type in hello. And you see how we're working inside the same project here. And it says I'm using computer skill because this task requires operating the browser UI directly. And so we can see that that is beginning to use the skill of navigating the website. Now what it's going to do is request permissions. So, it's requesting permissions to be able to look at my screen and give me access. And we can see that it's working in the background over here. So, it's navigated over. So, what this means essentially is that it's loaded as a tool. It's run through what it's going to do and it's beginning now to use computer use. So, you can get your AI to do stuff in the background. So, we can see that it's navigated to Google over here. And now you can see this is not me controlling. I'm not using my hands, but it's typed hello into Google for us and it's completed the task. And the cool thing is well, you see how we enabled that pet before the pet. Even if we navigate the website, we can see what codeex is using in the background. This is really useful because basically if I need to like go off and do something else, when I've got that coding agent working in the background, I can see all of the previous conversations over here and I can click through to them. And so no matter what I'm doing, whether it's inside Chrome or inside any other app, I can see what a codeex is doing in the background and I can come back to it later. This is also super useful because some of these tasks are going to take like 5, 10, 20 minutes to do. So how do you get a notification that's done? Well, you can look in the top uh sorry, in the bottom right at the P and then just click on them and go through to them over there. Really, really useful stuff. So if we have a look now, it's prepared the directions. It's giving us some ideas. And it said, "Okay, should we go with build one, two, or three?" I'm going to go with three. And then it's going to start building that out. As you can see here, it's like, great choice. Let's build that bad boy out. And then we've got the habit tracker website being built out as well over here.
And you might be thinking, okay, is this just for creating websites? Literally, no. This you can create this with anything. Let me show you an example what we've created recently. So we have this agent OS dashboard that links all of our agents together. We even have GPT 516 built into the system. And so what we can do with this is we can use all of our agents together inside one beautiful dashboard. We can link it together with a memory system. And this is a SAS tool that I've basically built out shared with my AR profit bordm community. It can basically automate and build anything. Really powerful stuff. But the great thing about this is that I'm not a coder. But we're at the point now where if you're using Claude or if you're using codeex, whatever idea you have in your head, you can now build. That could be an app, that could be a game, landing page, could be a tool, could be all of those. And you just build them out in the morning and you just tell it what you want to do. You might say, "Okay, does it work like magic?" You still want to test it. You still want to make sure it actually works. You still want to test and iterate because the first version you get back is never going to be perfect. But if you keep doing that every day, number one, your skills at using AI, which is probably the most important skill to learn in 2026, are going to improve. And then number two, you're going to get better output the more you test the projects you use. And you'll learn how to build better projects based on what you learn along the way. So this is the website that's being built over here. We have the other section over here as well.
And then if we go over to plugins, we've got all of our different plugins. You can also search for different stuff here. So, an example of a really aentic workflow could be like, okay, well, I want you to create a blog post and then I want to link it to my website and publish it directly to my website as well for me. We've also got schedule task here. So, in the same way that you can schedule tasks on chat GPT, you can also schedule them inside codeex, but the difference is that codeex is much more powerful because it can edit files locally. It can, for example, code stuff out for you. It could work with other apps in the background. And so this can run 24/7 as an AI agent and do things without you. So, for example, it could be like, okay, publish content to my website every single day for SEO based on what we're already ranking on. And then you don't need to log into your website. You don't need to have a team of editors anymore. You've just got Codeex doing all the hard work in the background for you, which is pretty unbelievable. So, let's check out where we're up to with the habit tracker website. What you're also going to see is that it creates images. So, it can create images for us. It can code for us and it's basically taking like all the best features of chatbt using them aentically in a system like this. And then we can see everything that's being built and it saves it locally to your computer. It's quite different as well to chat because chat GBT like it doesn't link or edit anything directly on your computer. This does this is working 24/7 on your computer. Um it can control your browser and it kind of just steps out the confines of a little conversation with chat sheep. So that's basically some of the best workflows right here. So we've talked about the chat, the sites, the pet, the plugins, schedule tasks. You can create new tasks anytime you want. You can see everything that's being built in the background. Super useful. So, it's now building out the website as well. And what you'll see on the right hand side here is that we have the changes that have been made if anything has been made. And then we also have this sidebar here. So any files that you have you can view on the right hand side. Then you have this little environment feature. And then you have in the middle the chat itself. So if we click on this we can see that we're working locally. You can change this as well. We can see our usage and we can also see our sub aents. So we have three sub aents working. We have one for design, one for design command and one for design modern as well. We can see what's running in the background. So these are terminal commands. So obviously this is a coding agent. It's using the terminal to code. But the great thing is this is you don't need to worry about the coding yourself. It's just doing this all yourself. Uh it's all doing this for you. So what this means essentially is like there is no technical barrier to creating anything that you want anymore. Now before you make something project or production ready, you might want to get like a team of developers to check it out for you. But there's nothing stopping you from creating your own projects without a team of coders or developers. this really I mean in the same way that I was talking about how the the workflows for content creation basically replace a marketing team the workflows for images replace a designer team and the workflows that we've got here for coding replace developers and so if you look at all the systems that chat have built it's quite scary because a couple of years ago I had a big team and now what I see is that we're actually like reducing the amount of team that we have we're hiring less and less every week, especially those sort of like menial admin tasks we're just replacing every single week with AI. And the difference is that AI doesn't take weekends off. It doesn't sleep, doesn't get tired, doesn't make the same mistakes as often, doesn't require retraining as often, doesn't complain, and there's no drama, right? So it's it's much easier to manage a team of agents that are just chilling in the background doing a bit of code whilst I'm talking to you versus having a team of of people who can be quite energy draining and quite timeconuming to manage honestly. So this is the way the world's going my friends. It's a scary place to be in but you can see how it's working and how you can use all of this. Now it is saving the site over here. So that's basically it. That that is codeex in a nutshell with chat GPT GPT 5.6 how to use it how to get the most out of it how to build with it how to make it run faster how to code locally even if you're a non-coder and you've basically gone from like zero and no knowledge to be a being able to build and automate anything that you want.
Now one final thing that I want to show you is what else you can do with this. So for example codeex can also create videos for you. Let me show you an example of that. So there's a skill in the same way that we installed Google Drive before. You can actually take a skill from GitHub like you see here. So this is Remotion and with Remotion, you can teach codeex to be able to create videos for you. So if we take this skill, for example, this setup here, and we're just going to go to Remotion on GitHub, and there's tons of skills like this, but this is just one example. We can go over to codeex here and we can train it with a new task on how to create videos. So I can say okay based on this skill create a video about the AR profit boardroom for our website. So we can hit enter like so and that's going to start creating the video now too. So, it's going to now start using remotion and it's actually already got a pre-built skill for hyperframes as well. So, it's going to use a combination of reotion and hyperframes, which are both video skills you can teach your agent. So, that understands how to create videos for you. And then this will just start creating a video in the background. In the meantime, we have the website being done over here. And we can review it. We can also open up the website and we can share it. So, if we click on open, you can see that our website is ready to go. We just got to sign in. Now, we're going to log in. As you can see, and we'll start checking this out. And here is our website built, designed, fully automated. It created the UI, created the page, created the whole system, built it out step by step, and that was super easy. And if you're like, "Okay, I want it to be red." No problem. Go back into codeex and ask it to do that. You might be like, "Right, okay, I want it to have a video at the top." No problem. Give it the details of the video and tell it what you want inside the chat. Anything you want to change, you can. No problem at all. And then you can open it inside your favorite app. You can share it with people. You've got all the assets, everything else. And if you want to actually view the folder itself, you can just click on reveal and finder. Boomshack lacker. You have the video uh you have the folder itself there. If you want to publish this, you can actually use something like Netlefi and Netlefi is a web hosting website. So, for example, if we have a look at the aiprofitboarding.com, I've never logged into this website. I don't even know how to, but my AI hosted it on Netlefi, deployed it to a custom domain and has my personal access token for Netifi so that it can publish everything that we create as well. And that is it my friends. That's pretty much it. We are at the end of the course. What else can we do? I want to just show you one final thing which is the video that I've created previously using GPT 5.6. So if you want to see more tests and more things of what we built, you can just go to the leaderboard at Goldique Bench. This is my benchmark for all the stuff that we've created with each model and we've tested each one. So we built out 50 different projects as soon as GT5.6 dropped. And if you scroll down, you can see that we have this video promo here. So this video was fully created with an AI agent. You can also give it access to your AI avatar. If you've got an avatar, for example, on a gen, it can add your face and voice over to it, too. And this video was fully automated using a combination of codeex and reotion. Looks beautiful, nice design, sleek, interesting, looks better than any of my editors could create, beautiful animations, and again, it replaces the need for a video editor. So, you can see how Codeex one by one is replacing everything that we do in pretty much every single job. And you might say, okay, what's left for us? Or, you know, this fast forward 6 months, 1 year, 2 years, what's left for us? I would say you're going to see the the biggest difference between those who win with AI and those who don't is the ones who actually implement everything that I've shown you today. And then also it it really comes down to the idea. I think the idea is everything. So for example, every time that I get feedback from someone inside the profitable community or they're asking me to build something or improve something, then I go into our agent operating system here and I just edit it. I change it. I make it better. I improve it. I add something new in. And so it's really about the idea. And the way that I would look at this is you analyze all the problems that you have, all the frustrations that you or your customers have and you start building those things out with codeex. So for example, one of the biggest problems that my community had was that they were using all these agents and they didn't know how to link them together. So we solved that with the agent OS. And then for example, a lot of business owners inside the community, they wanted to automate lead generation. So we built out a system that generates leads, sends emails, manages the inbox, and writes a whole email campaign. We also had a lot of people who were like, do you know what? I want to create videos like you, Julian, but I don't want to uh spend loads of time creating them. No problem. We'll create the video agent for you. So you can just type in a prompt and create whatever you want. And then a lot of other people are like, well, okay, I just I need my agents to understand me, my business, my goals across every single agent that I use. No problem at all, mate. We'll build out the memory system. So like every single problem that you face or find, I would start solving it with codeex or with claude because we're at the point now where that is all we can do. You know, we're in a we're living in a world where everything that can be automated is being automated. Everything that we do as humans is being replaced by AI, particularly anything that's behind a desk. I mean, you saw how smooth and easy it was to use codecs to control our computer and type for us and do anything else. So, if you want to get more training on this sort of stuff, feel free to get it inside the air profit boarding link in the comments in description. Thank you very much for watching all the way through to this point.
Inside the classroom, you can get our agent OS system that I just showed you. So, if you want that full system, you get it here. We updated daily new video tutorial, new guide, zip file to install it. We had new daily guides, as you can see. So, for example, a lot of people right now want to use AI more, but they run out of tokens because they're, you know, creating so much cool stuff. So, what we have here is a token minimization playbook. You can also ask questions inside the community, get help and support in real time and we help you as much as we possibly can. I actually create a video tutorial every day answering all the questions personally that we have inside the community and then inside the calendar you can jump with coaching calls, ask questions, get help and support, share your screen, etc. Inside the classroom, if you're a complete beginner, you can go from beginner to expert over here. And if you like stuff like this, we have loads of courses, loads of trainings inside the classroom as well. inside the map you can meet people in your local area and you can connect with people using tools like codeex and the agent OS and if you're wondering okay you know what are people's experience with this sort of stuff so we actually have 196 pages of testimonials of wins of people growing and learning with AI automation using the systems inside the AI profit boardroom and just having a great time learning on this journey so feel free to get it link in the comments description or just go to the aiprofit.com pump. Thanks for watching.