--- title: 'GPT 5.6: Soul, Terra, Luna Breakdown and Agent OS Integration' source: 'https://youtube.com/watch?v=N9FlkmPfuLU' video_id: 'N9FlkmPfuLU' date: 2026-07-14 duration_sec: 0 --- # GPT 5.6: Soul, Terra, Luna Breakdown and Agent OS Integration > Source: [GPT 5.6: Soul, Terra, Luna Breakdown and Agent OS Integration](https://youtube.com/watch?v=N9FlkmPfuLU) ## Summary The video provides a comprehensive overview of OpenAI's newly released GPT 5.6, detailing its three model variants—Soul, Terra, and Luna—and their performance across various benchmarks and real-world tasks. The presenter demonstrates how to integrate GPT 5.6 into an agent operating system (Agent OS) for autonomous task execution, compares it against Anthropic's Fable 5, and shares strategies for leveraging the model for AI-driven SEO and content creation. ### Key Points - **GPT 5.6 Release Overview** [00:00] — OpenAI released GPT 5.6 with three models: Soul (flagship), Terra (everyday), and Luna (fast/cheap). All share a 1M context window, 128k max output, and February 2026 knowledge cutoff. - **Goldie Bench Testing** [01:30] — The presenter runs Goldie Bench tests comparing Soul, Terra, Luna, and Fable 5. Soul ranks #2 on the leaderboard, with outputs described as 'Fable 5 level' quality. - **Model Performance Comparison** [05:00] — Soul takes 6 minutes to build a 3D space game vs. Luna's 1 min 37 sec. Cost: Soul $1.18, Luna $0.12, Terra $0.43. Token throughput: Soul 109 tokens/sec, Luna 198 tokens/sec. - **Agent OS Integration** [10:00] — GPT 5.6 can be used inside the Agent OS without an API key via OAuth. The system supports goal mode for autonomous background tasks, a workspace for saving builds, and memory integration. - **SEO Automation with GPT 5.6** [15:00] — The presenter demonstrates an AI SEO system using Google Search Console data, keyword research via Luna, content writing via Terra, and deployment. A video agent creates edited videos for ranking. - **GPT 5.6 vs Fable 5** [20:00] — Side-by-side comparison of 5 game builds: Soul wins on action/precision, Fable 5 wins on atmosphere/detail. Overall, very close; presenter prefers Fable 5 for Claude Desktop experience. - **Community Q&A Highlights** [25:00] — Topics include using Agent OS across multiple devices via VPS, creating AI avatars, running a marketing agency with AI, and setting up an Obsidian memory vault for agents. ### Conclusion GPT 5.6 is a significant release from OpenAI, offering three tiers that cater to different needs. When integrated into an agent operating system, it enables powerful autonomous workflows, especially for SEO and content creation, though Fable 5 remains a strong competitor for detailed tasks. ## Transcript All right, the big day is here. GPT 5.6. Let's go. We've got it inside Codeex. You can see inside the models here. We got Assault Era Luna. We're going to be testing out putting it through Goldie Bench, seeing how it performs as well. Uh comparing it against Fable 5, creating some awesome stuff inside the Agent operating system as well, the Agent OS using this. So, let's get straight into it. If you have any questions as we go along, feel free to ask. Let's get the party started. Can you add GPT 5.6 to codeex inside the agent OS? Make sure goal mode works and is can run in the background autonomously. I'm already logged in on codeex. So, make sure you use a Can you start running the tests on Goldiebench using GPT 5.6 soul? Add it as a new model. Deploy the builds as the build out. Make sure you test the actual gameplay or the working demo to make sure it works. Focus on these six builds first and then once they're working and ready, then do the rest. Can you make sure you're on Goldie bench? Be interested to know if anyone watching this has tested about GPT 5.6 is what you think as well. From what I've heard so far, is really good, but it's not on the same level as something like Fable 5, which is quite interesting. Also saw some people running it autonomously for like 7 days, which looked absolutely wild. Create a guide comparing them to create a guide on GPT 5.6 breaking down the models and everything else. Yeah, I think so. There's a lot of people doing that inside the AR profit boardroom. So, there's a lot of members like asking about that and building and sharing what they're building with it as well. So, we definitely check that out for a starting point. But, yeah, can be done. Heat. Heat. Hey hey welcome. Just testing out G5.6 today to see how it performs. Testing out soul especially, but we'll compare them side by side as well. It be interesting to see okay how does each one perform key creator guide on G5.6 versus Soul versus Terror versus Luna. Show the difference between each one. Show the benchmarks for each one. Show the actual graphs and everything else. And then also make sure that you build a few things and compare them side by side in terms of how they perform and talk about the use cases for each one as well. Loving it so far. All right. Exciting times. I'm excited to see it as well, actually. Heat. Heat. Does anyone know if GT516 is available on? I assume it is. Just made some changes to it as well. Don't override those. Aman says, "I've got an Aentic IDE and app I've built out 5.6 found four items between the two apps that were really meaningful. One of them found a bug that he's been fighting with. Nice, man. Yeah, it sounds like you're making good progress there. That's awesome. I'm excited to try it now. It's going to be fun. We're going to build it out in a few different ways. So, I think we'll put it inside codeex. We have a section for goal mode as well. Maybe a image generator, too. I think that'll be fun. Thank you. Yeah, that was based on your feedback. I know you asked to add that feature in, so we're like, okay, anything that's useful, just let me know. I'm one thing I'm actually thinking about removing is the uh is the scores on there cuz I'm like well you kind of have to recalculate it every time a new model comes out and a new model literally comes out like at least twice per week. So I'm thinking about the best way to handle that. And also just trying to make the UI on Goldiebench much nicer as well. If you have any ideas on improving the UI, let me know. But yeah, great ideas. Thank you. Don't worry about bugging me. Stop. Hello. Thank you very much for joining. Appreciate you joining. If anyone's got questions whilst we're going along like feel free to ask. Happy to help. You can see we're improving the UI on it now. So, it looks a lot nicer. Gradually getting there. I don't know. I'm testing out now. I'm interested to see Soul beat Fable or not. It's going to be very interesting. Um, yeah. But from what I've seen, it it creates amazing stuff. The other interesting thing I saw some people doing was running the one second. Yeah. The other thing that's interesting is running the goal mode autonomously for like days. I saw some people doing that. It looked pretty wild. I've not tested that out. I think he's going to use a lot of tokens, but some of the stuff you can build with it when it's running for days is just unbelievable. Oh, this is looking nicer now, isn't it? Much nicer. If I were you to make a session suggestion, it would be to move the best AI to its own column and then change your table so you and sort and filter by the columns. Interesting question. Move the best AI to its own column. Ah, I see what you mean. So, like the best AI out, right? Let's do that. and allow people to filter and sort the leaderboard too, right? Is your agent running your computer? It looks like it because I don't see you typing uh in in well we have loops running in the background so like I'll just say okay test this try this etc and then it will go from there what we testing today. So, we're going to run the Goldie Bench benchmarks and we're testing out Soul, Terror, and Luna. Going to compare them side by side. We're also going to test Fable 5 versus GBC 5.6 as well and just see how they all do. If you if you have any tests you want me to run as well, just let me know. I'll take a look for you. Vault, hope you're well. Hope you well, too. Thanks for joining. I'm doing very good. I'm excited today. I mean, it's been a great week, isn't it? Fable five back. Grock 4.5 out yesterday. That was pretty good. Um, and then today, of course, we got 5.6 out. So, it's been a great week. I'm excited. See what we got here. I really like the idea of a Jarvis type interaction, but I'm concerned about confidential files. um is the way that I have that set up is we will have a system where basically our agents are siloed away. So we do have like uh a Java style app here which is Apollo and I can tell it stuff and it will speak to me in real time and it works really well. So this is pretty powerful. But when you're using it as well, so you can see it's like literally looking at what I'm saying right here and understanding it. Now when I'm using it, it's siloed away from everything else, which means it only activates when I'm on it. And we can switch it off at any time. So we can click it off or on. So, the cool thing about that is when you're using this system, you don't have to worry about um you know, giving it access to stuff you shouldn't give it access to because it's all controlled inside this system and it's all inside one place. So, that's the way I really like to use it. It's much easier, much more simple. Um, and that way you can still make the most of a voice activated AI agent, but you don't need to worry about like, okay, you know, um, you don't need to worry about like it having access to stuff and it just switches off. So, as soon as I switch back to here, it stops listening and it will switch back on if I switch back on that tab. So, that's the way that I recommend it. Way more chill, way easier to use, etc. Is it weird that my agent OS is similar to yours, but the difference is the mode? What do you mean by the moat? Or what do you mean by the the difference? How do the loops work that allow it to do all the things I'm seeing? What is the main harness you're using? Is the agent accessing both codeex and clawed at the same time? Yeah. So, basically, we give skills to our AI agents. So, to show you a really simple skill, we have the SEO skill over here. Okay. And so this 13step process just continues in the background when I ask it to create AI content. And the great thing about that is that it's just running in the background. The same with, for example, Goldiebench tests. So it will just continue looping round until the the Goldiebench tests are done. Uh so for example, we've actually got that running in the background here. So when you're building out loops and that sort of thing, you you just want to I think the easiest way is to give it a skill and then it completes the skill and that's how it gets everything done. And then it's using Claude, but it's accessing codeex because we give Claude access to codeex via the agent OS. So Claude can operate codeex for us if we plug it into our Asian OS. So for example, if we go to codeex over here, we are using GPT 5.6 inside this system and then codeex can operate it for us. It can build stuff for us in the background. It's got everything inside this workflow here and it's all running in the background. Thank you. Are you silo away in VM or Docker? I don't really use Docker, but you can. Um, yeah, I just I'm I'm not that technical, so I don't really like to use it. And also, like I find with Docker, as soon as I stop using something, I switch it off and then I just don't use it again. So, a great example of that is Open SEO. Like, Open SEO is plugged into our agent OS, but because it's inside Docker is so siloed away that I don't have it running all the in the background all the time. As you can see right here, I can switch on anytime, but it means like I have to open up Docker and then I have to click play and then wait for it to load and then make sure it works. Um, so I don't really like using stuff inside Docker. I find it's like quite timeconuming. Julian, what plan do you need to be on to get access to GPT 5.6? Oh, maybe you just don't have access in your area. Let me have a look at I think when I'm inside chat GBT. Let's have a look here. One sec. I've got two accounts. So, just need to check here. Let's have a look over here. So, if we go to settings. Yeah. So, we're on pro here and we have sold down there. So it seems to be I've got access on pro but you can get access on open router. It's just you'd have to pay for the API. So the the loops are just skills but yeah they're they're kind of like crown jobs. You know crunch job is just a skill on a schedule. So yeah it works in a similar way. I mean there's many different loops. Like for example if you have a look we got this system for loops as well where we can define a goal and it will just loop around and get that done until it's completed. So that's an option as well. Okay, I got it now. Yep. How do you give Claude access to codeex? Is that in the in the school? Um, you literally just say to to Claude like, "Listen, plug in codeex into the system and then it will do it for you via and you just say do it via OOTH." Um, and that's how we can put for example codeex into the agent OS alongside Claude CLA. Nice. You got it. How do you have it siloed? No. Uh, no. So, I have it like so it only switches on when I go to this tab. But as soon as I switch away from that tab, that's how it silos away. So, it's siloed away inside a tab in the agent OS where we have the Javis style app. And then we can switch between them anytime. So, slow roll out. Ah, interesting. Also in Texas. Oh, we got a whole bunch of people in Texas today. So, just waiting for this to build out. Let's see how we're getting on. Just going to check something here. Well, it's only it it's siloed in a way where it's only activated ated the agent is only activated when we click on that tab. So if we don't click on that tab, the agent is not activated. So that's what I mean by siloed away. But also what you could do is you could when you're setting up Apollo with the agent OS, you could actually tell the agent like listen limit your permissions so you don't have local access for example. And that way as well you could reduce the amount of access it has. So when you're setting up this uh when you're using the installation file from the air profit boardroom you could say okay set up Hermes Apollo but limit the access in terms of the files so that you only have access inside one folder or you don't have access to local files at all. And that way you can silo it away as well. Donjee. Not the Javis agent. I'm asking about the agents. Oh yeah. Yeah. No, those are not siloed away. Those are not siloed away. Like they have access to everything. But again, like you can limit the permissions. You can limit access to local files. So if you want to silo away certain parts of it, you can. Also, another option that a lot of people do inside the AR profit boardroom is they will run the agent OS on a VPS and that way it's siloed away on a a separate device. That's enough way around it. Thank you. Welcome, Edward. Thank you very much for joining sir. Wow, it's early early for you, but I appreciate you joining at this time. Good on you for waking up early. All right, we should have a better version of Goldie Bench now. Let's take a look. Yeah, it looks nicer. Looks a lot nicer. Whoa. Awake for for 19 hours. Got to get some sleep, man. Take care of yourself. But usually I sleep about nine hours a day. Sleep about 8:30 9. Wake up about 5 or 6 a.m. All right, back in a sec. Heat. Heat. Sorry about that, Guys and girls, let's see who we got here. Haha. Wow, it's 8:29 p.m. in Louisiana. Sitting in bed working on a laptop. Awesome. Edward, I know, I know, but I'm that guy that doesn't like to sleep. Fair enough. So much energy. Impressive. Been watching more and more. I will be a member of the community soon. It feels right. Thank you so much. Uh oh, wow. How old are you? I am 37 this month. 36 right now. Hello Julian from Austria. Thanks a lot. Thank you very much. Krazy Design. Thanks everyone for joining. All right, let's see where we're getting on. So improving the UI on this. Ah, it looks 10 times nicer. That is looking way. I mean, it's not perfect, don't get me wrong. It's definitely not perfect, but it looks a lot nicer. Let's see what else we've got here. Uh, so you can sort by what it's best at, the model, the rank. This is looking better. Thanks everyone for the ideas on improving this. 3:35 a.m. in Zurich. Hello Henrik. All right, we're building this out. That's done, I think. Let's see how we're getting on. Where are you established, Julian? What hour is it to you? So, I'm based in Asia. It's 8:30 a.m. here. Hi. Thank you very much. You're impressive in your I would have guessed you're 25. Ah, that's being too nice, but I appreciate that. You're impressive in your discipline in life, man. It's inspiring. saw once how you showed how to track all your time and I just respect it. Thank you very much. Yeah, that's uh that is something really powerful. If you're watching this and you don't track your time or you know you haven't even done that for a couple of weeks, it's well worth it. Well worth it. And we got Geraldo here from Brazil. Welcome here. How's he going in the World Cup? If you're into I don't know if you're into football, but you might be. Not so good. Not so good. I hope it gets better. England playing on Saturday, which I'll probably watch a replay of and try and avoid the news in between. See how we're getting on with these tests. Oh, so we got GPT 5.6 I think nearly ready. Let's see what we've built with it. Whoa. All right. All right. It's looking good already. So, we got six builds with Wow. Wow. Nice. Ooh, this is some of the best I've seen. This is Fable 5 level. Very good. Very good, my friends. This is looking really good. Boom. Shakalaka is what I say to this. Controls are backwards on this game, which is a bit confusing. But it looks great. Looks absolutely fantastic, doesn't it? Try it with Grock 4.5. Yeah, I've done quite a few tutorials on Grock yesterday. It looks awesome. Then it says, "That's good. That is I would agree with you, mate. I'm with you there. It's looking very good. You should try the new plug-in sites. What's the plugin? Is it designed for building websites or something? This is I'm I've got to say this is looking out. There's a couple of things I would tweak on that which I'm going to tell Claude to fix now. Heat. Heat. Codeex desktop chat GBT as well. Nice. That is looking very impressive. Absolutely loving it. Can you add this to the agent OS as well? Can you make sure that you add the builds inside a GPT 5.6 six soul folder in the workspace of the agent OS as well and make sure it actually works. Looking fantastic. That has been the easiest test I've done so far, honestly. Really, really good. inside the agent OS goal section. Can you make it so that when I doubleclick it, I can also rename it? Can you also make sure the full URL shows? Okay, nice. Let's have a look. So, codeex sites before use site says continue. Create a new site. Create a website that advertises the AI profit border. Don't use the skill boss skill. In fact, delete that. You know, the key difference as well with all of this is that you can use a warf with chat whereas you just can't do that with Claude, right? You're not going to be able to use, for example, Fable 5 inside Hermes. But you can use GP 5.6 inside Hermes. Let's see how Codeex sites is getting on. She's working right now. This is looking good. We're going to have to create a guide on GPT 5.6 versus Fable 5 as well. Perfect. Can you also make sure that I can like bold and format the text as well when I edit it? So, for example, if I press command and B, I can bold it. Let's see how sides is doing. Still working in the background there. Why is open rooster so slow? Is everyone using GPT 5.6? Oh, you missed some good stuff. Well, we'll we'll show you in a second, but we built some awesome stuff with GT 5.6. It's looking really good. See, it's coming through nicely here as well. Can you create a guide on this? Talk about all the ways you can use GP5.6 for AI SEO inside the agent OS. add some screenshots and data based on our top performing websites that you've already got access to. Let's check out the SEO projects as well. See how we're doing there. Whoa, that's crazy. Let's go around. That's doing good. We like that. so many sites beginning to take off now. That is good news. There we go. It's good to know a bit of romion. You never know when you need it. I was in uh where was it? Bucharest a couple of years ago. I did speech there at one of the events for SEO. Used to have a Romanian singing teacher. Nice. Polish piano teacher. There we go. We should start a band. We're going international. all of these builds are looking really good actually from what I can see. Let's take a look at Goldie Bench. See what's going on over here. Controls are a little bit weird. I'll tell you that for free. Overall, looking good. The the wife won't be happy about that. How about starting a band? Great. And have you deployed it to the workspace folder inside codeex? Just right up there with Fusion right up Yeah. Wow. No more fans allowed mate. Ah, sad times. Yeah, I would agree. That's Do you know what? This is why I tend to put everything inside the agent OS cuz like I wouldn't like to give codeex access to all the websites we have directly inside Claude. Otherwise, what what could happen is like they get confused and then they overwrite each other and it get super messy. It's like no no no no no let's just make sure we have all my main projects inside claude and then if I want to use codeex I prefer to use it inside the agent OS that way it's kind of silo the way you don't like overwrite your projects or for example deploy a site. Exactly. Exactly. Cuz what's happened before is like I've had two website uh sorry two different coding agents and then you deploy a website but let's say for example codeex edits it as well as claude and then claude deletes the improvements made by codeex and overrides it. You kind of go like one one step forward, two steps backwards, which we don't want obviously Thank you. Yeah, definitely check it out. Exactly. Yeah, you don't want too many. I think the more you can simplify, the the less room for error, which is what we want. Awesome. Have you deployed the latest tests? How much better is GT 5.6 than GPT 5.5? I would say minimum, absolute minimum, I would say 50% better. But from what I'm testing, it's it's way way way better uh in so many different ways. You'll see on the test in a second. Um I don't actually have like 5.5 comparisons cuz we created after that release, but yeah, pretty good. Oh, thank you so much. I really appreciate that. Thank you. Super kind. Yeah. All right, let's get into this. Today we have three new models from GPT 5.6. Massive step up. I can tell you now. I'll show you some of the tests that we've gone into, but we have a full solar system of new models directly with GPT 5.6. If you don't have access to it right now, by the way, it is rolling out slowly. So, some people around the world are getting it sort of step by step, if that makes sense. Now, OpenAI's new GT 5.6 has three new layers. So, Soul, Terror, and Luna. Luna is the fast cheap one. Terra is the everyday model. Soul is the beast. If you want to see what Soul can achieve, let me show you some of the latest creations from GBT 5.6 that we've tested on Goldiebench. And you can see it's actually coming in at number two on the leaderboard when we've tested out. It's unbelievable. So, let me show you some of the stuff it's created. It looks really, really cool. Uh, we've tested out relentlessly already built out 33 different projects with it. And you can see this on Goldiebench. So, let's check them out. For example, we have this game and it's looking really really good. You might say, is it Fable 5 level? I would say at this point it's it's on par. Like it's really really good. It can create some amazing stuff. I have a little feeling that if I was to choose one model and only one model that I could use forever, it would definitely be Fable 5 over GT 5.6. But Soul is awesome as you can see and creates some amazing stuff. Let's have a look at another one. So, this like an open world game as you can see right here. Um, it feels really nice to use. I was testing out Grock 4.5 and that impressed me yesterday, but this is way way better. It's just more fun. The game play is sharper. It feels just a lot nicer visually and it's a lot smoother and a lot more interesting in these open worlds as well. So, it looks fantastic as you can see. Let's have a look at another one. It's like a Doom style game as you can see. And one of the things I'll say here is like quite often models struggle to really build out like a smooth game like this. And this has has worked perfectly and it it just feels really nice to play. It feels smooth when I'm looking at the creations it's made. Is like a racer game as well. So Soul is the absolute beast. In fact, I probably wouldn't use any other model apart from Soul um if I had the choice. I think that's what I'll try and stick to for now. But if you're wondering about the other two models, I'll come on to those in a second. But just to show you like the the creations here are absolutely unbelievable. Here's a full video that it created. So it created a full video animation here. Looks super nice. This is you can see here for example has a progress bar at the bottom. It's created all of this. Like I haven't touched it. I didn't create any of this. It just from one single prompt does the test and it looks fantastic as you can see right here. So this is fully like a video fully created with GT5.6 soul. So you can create videos, you can create games, you can create landing pages, you can create awesome stuff as you can see right here. And uh yeah, unbelievable. All of it was really really good. There was nothing I looked at where I was like ah not sure about that. Like all of it that I checked out was just fantastic and it's it's a really really good model which is the first time that I've said that about GT this year apart from the image model. Their image model was goated as well to be fair. So let's run through it. What is GPT 5.6? How does it work? Etc. So this just dropped today. Open up made GPT 5.6 generally available not as one model but it's free. So same architecture, three sizes. All three share 1 million context window, 128k max output and February 2026 knowledge cut off. You can see the prices here. So for example, of course, like soul is a lot more expensive to use than Luna, which is a lot cheaper. One thing to bear in mind here is when you're using it, like you can just use it with a so that means you don't have to use it with an API. So for example, if I was using my agent OS and I'm using codeex inside here with GT 5.6. So we can switch between the models. I can use a free model as well like omni root inside codeex if I want to. But if we went to the workspace here, what you can see is that we can flick through between everything that we've built. It's inside one place and this was all created with a flag like I didn't pay for the API to to build this stuff which is absolutely awesome. Now it's the same for example inside Hermes we can use GPT 5.6 is a separate profile and the great thing about that is it doesn't cost us any extra whereas for example if you using uh Fable 5 or any form of Claude inside Hermes agent or any sort of agent like openclaw you'd have to pay. So there's a big difference between these. So how would you use these? Well, if you look at the positioning here, Soul is more for like flagship hard multi-step agentic work. Terra is the everyday production model and then Luna is like high volume latency sensitive work. So, Sun um Sun uh for Moon, biggest to smallest. And this was actually delayed. So, there was a government review first. It dropped in preview about a week or two ago. And then from there, they've finally released it to the public, which is fantastic. So you can see here if we look at the benchmarks on intelligence index fable 5 is still outperforming soul by one point. So it's 60 versus 59 versus 55 vers 51. This is from the artificial analysis index. So soul is literally right up there with fable 5 cost per tasks. So you can see here soul is a lot more expensive like almost twice the the price of terror. Luna is 80% less than soul. And then we've got Terminal Bench as well. So Soul Ultra is right at the top there. I don't think I've even tested Soul Ultra. That could be fun as well. So Soul itself is really, really good. And then if we have a look here as well on Terminal Bench 2.1, Fable 5 is 83.4. Soul Ultra is 91.9. Soul is 88.8. Now, from what I've tested so far, it it does look really good. And the thing I would say here is like don't believe all the benchmarks. Don't listen to all the benchmarks. just test it yourself. That's why we created Goldie Bench so that you can have a look at what I've built with it and then test it out. So, you know, on our actual creations, it's right up there. Like, it's it's one of the best models you can use undeniably with GT5.6. So, if you're wondering like how long does each one So, we actually gave each model the same task with one shot. Uh, this was for a 3D space game. And from what we saw, Soul takes a lot longer to build. So 6 minutes versus 1 minute 37 for Luna. The total cost was $1.18 versus 0.12 from Luna and 0.43 for Terra. Code written was much bigger from Soul. Measured through was a lot slower with Soul as well. So you can see 198 tokens per second versus Souls at 109 tokens per second. and builds that ran errorree. They all ran error free which was fantastic as well. If you're looking if you want to see like what the actual creations look like. So this is Soul versus Terra versus Luna. So Soul's version here as you can see like 3D game pretty fun to play etc. Let's have a look at I mean there's a huge difference between it's not even comparable at that point. Look at Terra's version Soul. Like Soul is just a much better model. And then you got Luna which is kind of the same level bad as terror and at the same time has less going on less interesting. Right. So Soul is is literally like the best model by a long long way. We also built out some landing pages. So if we test these live and by the way if you want to see this full guide it's available at agentes.guide. So this is the page that we built with soul. Looks super nice. Like look at the colors. It doesn't feel like, you know, AI slop to me. Like, it looks super nice. So, we built out a full landing page for the air profit boardroom. Looks beautiful. Super interesting. Looks fantastic. If we have a look at the version from Terra, the medium model, this is not bad still. Like, it's kind of it kind of feels like Opus 4.8. Honestly, when I have a look at this, when I'm testing out, it still looks good, but it just it doesn't have that nice UI that Soul does. Soul's output is way nicer. And then if we have a look at Terra actually did a very good job for such a cheap model that did a great job at building a landing page. So fair play to to Luna on that job. And then we built out an analytics dashboard as well. Let's test these out. So this is the one from Soul. Looks great. The buttons on the left hand side don't actually work. Just to be 100% honest with you. This looks very similar to be fair. I mean, they all look very similar at that point. Like, I can't really see a big difference between them. Now, you might also say, well, you know, sold most expensive model, like, isn't that a waste? But actually, if the outputs are like 10 times better, uh, especially for the first test, then it's 100% worth it, right? That's the way that I look at it. It's kind of like, would you use Haiku or Sonet or Opus or Fable 5 if you had a choice? Of course, you're going to go with Fable 5. So, here's the way that I would recommend using it to get the most out of this stuff. You know, the old way would be like one model for everything. Like just a flagship model for every request, big or small, summaries and drafts at frontier level, slow responses and tasks that needed depth and a bill that gets more expensive uh with usage, not with difficulty, right? And then with this method, I would recommend that actually you switch to you change the model depending on what t you needed. As you saw, for example, like Terra and Luna did a pretty good job on the website. So if it's like high volume work, well you can use Luna. If it's just day-to-day work, you might use something like Terra. But on the hard agentic stuff like building out big projects, coding something really interesting, then you can use Soul. And that way like you keep the cost down or you keep the token usage down without things getting expensive. Now you also might say, okay, the the cheap tier, for example, Terra or Luna, it's not going to create something nice. But I've shown you three different tests, and on two of them, the outputs were very similar. You might also say I I should just wait and use whatever's in chat GBT. I would switch the models depending on what you want. And I also don't think the best way to get the most out of AI is using chat GPT. So if you just go back and forth inside the chat of chat GPT here, you're not going to get anywhere near the level of uh output that you would get from something like this. Right? So we've got the agent OS here. And if you want to use GPT 5.6, I wouldn't use chat GPT. I'd go straight into codeex. I would create some amazing stuff like this. Everything saves inside my workspace. We can check it out anytime. Easy to use, easy to navigate between everything and it's all there ready to go inside one place. And then if we want to switch, we can go to Hermes. We can set up a profile with GPT 5.6. It's ready over there and everything's linked inside my memory system, my memory galaxy. That is a way to get maybe 10 10 times more out of your AI, whatever AI you use. Whereas, for example, you're just going back and forth inside the chat or using codecs, I don't think you're going to get that much out of this. So, Luna for volume, terror as the default, soul for the hard 20%, long aerentic sessions. You can actually run soul for hours or days and especially with the for/goal feature and get a lot out of it. So, if you go over to codeex over here and then you go to the goal mode, you can run uh the goal mode for for like literally days. You can say don't stop until you get the job done. And that's a really powerful way of just using autonomous agents. So yeah, really, really good stuff. All of them very impressive. Definitely the best release that I've seen from chat cheapy this year. Apart from the image update, you might also think, okay, when would you not use this stuff? I mean, if you have a choice between Fable 5 and Soul, I would still go with Fable 5 personally, but I think you can get a lot out of them by just having them inside an agent operating system as well. So, what have you learned today? You've learned about the three sizes of chat, soul, terror, luna. You've learned how to get the most out of each one and when to use each one. Which one is the agentic leader? And also, you've seen how soul performs in terms of creations and everything else. You might also be wondering, okay, how does it perform versus something like Fable 5? So we can actually have a look here and compare Fable 5 versus GT516 Soul side by side. So if we have a look, let's just check the arcade game for example. So this is the output from GT516 Soul. Looks pretty nice. This is the output from Fable 5. It's actually very similar. I would say this just feels slightly smoother from Fable 5, but not a big difference. So, if we have a look at this game here. So, this is the crypt game. I like the way the the, you know, the enemies and the characters appear. The controls are a little bit weird on this game. But apart from that, pretty nice and lots of characters, lots of interesting stuff going on there. We have a look at Fable 5. Feels a little bit more linear, but again, like I would say the graphics are nice. It looks cool. It's easy to use. One thing I will say is like you see how the staff doesn't move when we use it. That seems a bit weird, but apart from that, it's pretty good. It's a little bit buggy there as well. So, I mean, very very similar outputs. Like, I can't see a big difference. This is the only one where I see a huge difference between them. Like, I would say that Soul absolutely crushed it on this example. Like, looks really good. super smooth. If we have a look at the version from Fable 5, we can't really turn around. So, it's kind of strange. It's like you you can only really move backwards, forwards, and sideways, but you can't change like the camera angle, which is kind of weird. Pretty hard to use. Yeah, but yeah, very very compatible. So, if you want to get our systems for using all of this, you can get that inside the AI profit boarding. We've got the complete agent operating system with the model routting playbooks. We have the agent OS. You get five uh sorry, four coaching calls a week. There's 1,000 pre-built agents inside there. You get new model briefings all the time. 4,000 members across many different countries. And all of the systems I've shown you right here are inside the air offer board. So, we have a codec section here. And if you want to use free models with codecs as well, you can choose that inside the drop down. You can also switch to goal mode. You can see all of your sessions. You can check out your workspace here. Here you can see everything you've created. You've also got, for example, the orchestration system here so that you can have paperclipip orchestrating a team of agents that could be working with Claude and GPT 5.6 and everything else. And then you can also plug in Hermes with Soul over here too. And then we additionally have a image generator which is powered by GT5.6. So if you want to get all of this, it's inside the AI profer room link in the comments description or just go to the profer.com. This is my community for helping you save time, grow, and scale with AI automation. inside the classroom. You can get access to all of my best trainings. If you want the agent OS system that we talked about today, you can get that over here. If you want to get our token minimization playbook, so you can use GT5.6 for more without paying more, you can check out the token minimization section here. And we have loads of great guides and new tutorials all the time. Inside the community, you can ask questions, get help and support. I personally answer all the questions inside here daily. And then inside the calendar, you can jump on weekly coaching calls, get help and support in real time, ask questions, etc. and inside the map you can meet people in your local area. So feel free to get all of this inside the AI profit boardroom link in the comments description or just go to the aiprofit bomb.com. Thanks for watching. Trey says 5.6 is doing things to my website that are 100% logical but I didn't specifically asked GPT to do. I had a problem with some uh code and GPT went and fixed it. Nice. Does that mean is is that good or bad? It sounds good, but also like it might be bad. Josia says, "Great see here." Great see here too, Josia. Thanks for joining, sir. Appreciate that. Ah, thanks so much for joining in by the way. Just asking chat cheap how to get the best out of your workflows. Awesome. Well, I mean, feel free to DM me inside the proper forum if you have any questions or anything like that. Always happy to help. So, I really appreciate you joining and thank you. Going to move on to the next tutorial in a second. Hey, Don says, "Hey, quick question. Feel free to ask anything you want. Red says, "Evening, Julian. Enjoying soul." Yeah, it's really, really good. Very impressive from what I've seen. We already posted the benchmarks by the way for everyone on on Goldieb, but it's looking fantastic. Not gonna lie. Not not been a big fan of chat cheap open eye this year to be honest with you, but I would say this release is special for sure. So are you using GPT 5.6 basically for free in your OS? How are you doing that? Yeah, it's through oorthth. So you can so for example with GPT 5.6 in the same way that you get access to codeex you can plug it into your OS and that way you can use it for free. So it's pretty cool. I was, you know, I did a SEO podcast on on my other channel the other day and um they were saying how like a lot of people don't go outside that much nowadays because of AI and you know people are talking to AI so much all day. It's it's kind of like the internet. You know the internet kind of made people go out less and made people a bit more introverted. It's like that but even more so because you're always like interacting with it, right? It's pretty interesting to see that Today I'm going to run through GT5.6 and an Asian operating system that doesn't even require an API key. You can basically use a Wolf with your existing subscription, plug that inside one dashboard and it's absolutely wild for pretty much automating anything. Now, GPT 5.6 just dropped. We've already plugged it into our agent operating system as you can see right here. Um, we can actually switch between APIs. So, for example, if we run out of tokens on our existing plan, we can switch over to omniretex inside here. And then we can also switch over to GD5.6 like so. Then we can speak to it inside here. So, let's just test it out. We can add a message. And this is pretty powerful stuff because for example we have a system now where essentially and you can see it responds here. We have a system now where essentially we can save everything that we create with this agent OS inside one place. So for example we created this analysis as you can see everything that we've created and built with GPT 5.6 is inside one place and also that connects with a memory. So it's a super powerful system to just get more out of your agents. And also the great thing about this is that you can now plug GPT 5.6 into agents using OAF as well. So for example, Hermes agent can run directly with GPT 5.6. So this is what I call the GPT 5.6 command engine. And the thing I would say here is like if you're using GPT 5.6 ICS like if you're just going back and forth inside chat chet like so you're just not going to get the most out of it right like you can't see anything you've created you all your conversation history is quite messy it doesn't plug into a memory system also if you're using other tools like for example we use claude we use anti-gravity we use codeex whatever we're using we can have inside one system we can quickly navigate between them all in one place one click away whereas for example if you want to get more out of codeex then you have to go to codeex over here and then you have to for example, navigate to Claude and it's all quite messy. So, this is a much more powerful stack. And so, if you're watching this, just promise yourself one thing, which is you'll finish this guide and set up one piece of it today. Because even if you just set up one piece of this, the moment you make that transition, everything about how you work with AI changes and this is just going to make your whole AI engine way more powerful. So, let's talk about the GT command engine and five things it gives you. So, first of all, you got one login. So this is all running on my chat GPT subscription. So I can just log in. I can use all these tools and it doesn't cost me anything extra. I don't need to use the API or anything like that. It's all plugged into one system. Also, we can switch between the different tiers. So we have Soul, we have Luna, we have Terra, and these are all different parts of the solar system that we can use to get the most out of this stuff. We have goal mode. So we can go inside goal mode here. And this is a loop where we can basically hand it a goal and walk away. We have the workspace where every build it makes is saved and previewable in one click. And we have one command center. So GT516 next to Claude, Hermes, and Gro on the same screen. One dashboard runs your whole model lineup, which is really, really powerful. So let's talk about this. You basically ask, it's going to use your chat GT login. GT 5.6 builds with whichever model you prefer, and then it finish saved inside one workspace. And the difference here is like the old way is you would used to live in like chat GBT tabs. You copy and paste in between them or you're paying per API token which gets expensive. You close the session all the context is gone and it's not plugged into one memory system and then it's not running in the background. You're one model and if all your builds are just scattered across your downloads folder and your tabs, right? So for example, if I go into codeex here, it's super messy. I can't organize everything. I can't visually preview anything that we built and it's it's just nowhere near as good. Whereas with this system, with the Asian OS, we've got everything inside one workspace, right? We can preview everything that we built. We don't lose track of anything. Everything is saved in one place. It's much easier to navigate. So with this, we have one dashboard, GPT 5.6 in one tab on the chat GPT login. You already have Soul, Terror, or Luna ready to go. We got goal mode ready to go. Every build is saved and previewable. Everything sits next to all of our agents and nothing is lost as well because everything goes inside our memory system. So whenever we use any of these agents, it's all saved inside our memory and then we can navigate across that inside our memory galaxy which is super powerful. So a lot of people ask me like how are you using GPT 5.6 inside your systems here. So we actually give the agent OS to our members but basically when you install this you can just ask your agent that sets it up to plug in your OOTH into this system and that means you don't need to use an API key and then you can basically with one drop down automatically switch to the best model depending on what you want. So Soul is the flagship one really powerful terra is the everyday driver. Luna is super fast and cheap for high volume work. And inside the agents, it will just live inside one section. The other cool thing as well, like if you're running out of tokens, you can just switch to Omniroot or HY3. Both of them are free models that you can use with codeex. So if you're like, ah, you know what, this is a high volume task. I don't need a frontier model. You can switch between them as well. Also, you may see say like the cheap tier doesn't create good stuff. But actually when we've tested out if you're just building out a basic website there's not a huge difference between each of these. So it really depends on what you're doing. Like if you're creating a massive project super complex big SAS of course you're going to use soul. So it depends. Most workflows don't require a frontier model. And then also we've got the goal mode. So what we can do with goal is we can go inside here. We can set a title, set a task and then we can set up auto approve. And basically this will just loop around as a system until it's completed. So you set a goal, you can close the tab, it builds in a sandbox, and then once it's done, it goes into your workspace. So all of our goal tasks here are saved inside one workspace, and we can just come back to that whenever we want, which is fantastic because basically we can run our agents autonomously in the background whenever we want. So if we wanted to do this, for example, we can plug in that task, launch a goal, and now it's going to start running in the background inside a sandbox. Pretty powerful stuff. You also might say, "Well, an AI running on its own in the background sounds risky." So, it's actually sandboxed in, right? Each goal runs auto approved, but it's sandbox. So, you can also switch off the auto approve as well. If you want to switch off, you can untick it there, which is great, too. Now, also, the great thing about this is that every build is saved and previewable. Now, you know how I said, uh, for example, Soul is is the best in terms of Frontier work, like it can create the best stuff, right? and create the nicest stuff. Uh we actually tested out on like 40 different builds so far. Really, really good. Then the other thing I said before was like you can't see a big difference between the dashboards or the things that you create. So this is Terra and this is Luna. There's not a big difference between the builds. Like very similar in terms of the way they're created. And so depending on what you're doing, you don't always need a Frontier model. You don't even need to use the model that uses the most tokens depending on what you're actually doing. Now, you also might say, okay, with an aentic system like this, will it work for your business? And the great thing about this is you can customize it as much as you want. So, we have loads of members who are non-technical, building out their own custom workflows, creating amazing stuff. And the way that they're doing this with the agent OS is building it out and then just saying, "Okay, well, I need to automate this. can you build that in? And they just ask SA agent to add that to the agent OS and it doesn't. So for example, like previously I'd have to log into Suno create music. That would take like 5 to 10 minutes of me waiting. Now I can just have this tab here and I can create music in one click. If I want to do SEO, I can go to this section. If I want to use a video agent that can like create uh fully edited video that looks beautiful with an AI avatar and nothing else inside one system, we can. Really easy to do. And so when you're doing this, it's pretty easy to build out these systems and custom workflows. And that way it works for any business. But if you want to get all of my pre-built workflows and the codec set up, that's inside the agent OS as well. Also, you might wonder, okay, when you're using the Agent OS, which is the best model? Well, if you need speed, Luna is the fastest. It's actually four times faster than Soul when we tested out a build with it. So, what's really interesting here is Soul can create the best quality stuff, but it takes a lot longer. So, 6 minutes versus 1 minute 37, right? Big difference in the amount of time it takes. You might also say, doesn't running an agent OS use a lot of tokens? So, if you're worried about that sort of stuff, we have token minimization playbooks inside the AI profitable boardroom, but also you can switch to free models from the drop down here. So it's like okay well if GPT 5.6 runs out of tokens or you don't want to use it or you don't want to rinse the tokens on it no problem you can switch to free models like omnirute. So that way you can switch depending on what you want to do and it's really easy and simple to just switch between them and make the most out of all the models. So just to recap here some people think you need to use an API to add GP 5.6 your agent OS. You don't as I've proven. Some people think the autonomous agents running in the background are like a gimmick or don't actually work. At this point, we're at the the world, we're in a world where we can do everything in the background. So, for example, that task we were just talking about a minute ago, that's been running for 3 minutes in the background whilst I was talking to you and navigating across the agent OS. So, it's really powerful because you can just give it a task, walk back, we can come back in an hour, that should be done. If it's not done, it will keep looping around until it is done. And also, you might say, well, this is just one more tool to check. But actually, it sits all inside one screen. That's the difference here. Like, if I was using GPT 5.6 inside Codeex, I would probably never use it, even though I'm on the pro tier. Uh, simply because I just don't want to navigate to loads of different apps and mess around clicking between tabs. Whereas, if I've got everything inside one system and uh, goal mode can just run autonomously without me, well, that's fantastic, right? I just get more done inside one place and I'm way more productive. And then the way that I see this is like this is the model we already pay for. Why not get the most out of it and plug it into a system that's super useful? So if you want to get our agent operating system set up with GPT 5.6 wired in ready to go, you can get the GPT 5.6 Codeex engine, goal mode, the workspace, one dashboard for every model including cord Hermes Gro codec side by side, the local Hermes engine, the AI mastermind with a group chat with these agents, agent cambban, the memory system, the token efficiency playbooks, and 4,000 founders that you can network with as well. You can get that all inside the AR profit boardroom. This is my automation community for helping you save time, learn, and grow with a automation. We actually at this point have 196 pages of testimonials and wins and awesome reviews from people learning and growing with this stuff. Lots of people sharing their agent OS setup, how they've customized it, how they've built it exactly how they want. So if you want to get it, it's inside the profit boardroom. Inside the community, you can ask questions, get help and support in real time. I personally answer these questions as well. And then also inside the classroom, you can also get access to all of our best tutorials and lessons. Like you can see right here inside the new daily update section, you can get the agent OS. We update this daily with a video tutorial and it's zip in file to install. And then also if you want to learn about token minimization or for example omniroot or anything else that we mentioned today, you can get all of these tutorials inside here. Inside the calendar, you can jump on weekly coaching calls, get help and support real time. And inside the map, you can meet people in your local area who are building with AI agents like you. So feel free to get that link in the comments description or just go to the aircraft.com. Thanks for watching. I thought Grock 4.5 was only a cursor, but I did put it in. Yeah. Uh we we have it inside the agent OS. Created a ton of tutorials about that yesterday. Do you have this stack laid out in the school? Oh, yeah. Yeah. So, you just go to the classroom here. Then you go to new daily updates and you can get the Asian OS right there. So, it's all right there for you. The only tiny hiccup is that you're not able to bring your memory from chat. Um, well, so the way that you would do this instead is like you have the agent OS system and everything goes inside the memory. So every time we use all of these agents, they can use a memory from this system. It's all inside one place with a neatly organized context. And then well, you get a certain amount of tokens. As far as I'm aware, there's no time limit on GBT 5.6 ICS unless I'm I haven't seen the news on it. But yeah, as far as I'm aware, you can use the OF like we've been using it for for a couple of months now inside Codeex. I'm not sure about time usage, but token usage they limit, right? So, but that's why I talked about omniroot. So if you run out of tokens with chat GPT on your subscription with our agent OS you can use omniroot or tensen hy as you saw in the drop down and so you can switch between free models if you run out of tokens and also we have token minimization playbooks inside the AI buffer board as well. Yeah, exactly. So, Don says, "We're kind of saying the same thing, which it basically comes down to how many tokens you have." Yeah, that's it. And if you run out of tokens based on your plan, no problem. You can switch over in the agent OS to 10 cent HY3, which is a free API, or omniroot as well. So that way you can still use codeex and you can still use agent OS without running out of tokens. Thank you. Yeah, Omni Route is really good. Super impressive. We got two hot tamales out. GPT 5.6 versus Fable 5. Who wins today? We're going to be testing them out side by side and showing you exactly who wins. So, Open Eyes new flagship model is GPT 5.6 six soul and we're going to be comparing that against Claude Fable 5 and seeing how it performs side by side. So these are two new Frontier models very different actually. So basically GT5.6 Soul is Open Eyes flagship model top tier. And then you got Claude Fable 5 which is Anthropic's newest model as well. And if you're wondering okay how do they perform on benchmarks here? The output price is a lot cheaper on Soul. And if you're wondering how do they perform on benchmarks, let's run through it game by game. So, we've tested them out side by side. I'll run through the results in a second. So, this is Soul versus Fable 5. Let's open them both up. So, Soul's output here looks really good. Really smooth game. Kind of like an open world dragon game as you can see here. And actually works. doesn't feel laggy or buggy at all really. I mean, the graphics could be better, but overall pretty nice considering this is one of the nicest generations I've made for this specific test. Now, let's compare that to Claude Fable 5, which is a bit brighter, but not quite as interesting when you're roaming around. on. So if we have a look for example, there's less characters, there's less stuff going on, not quite as interesting, not really as entertaining as a game. And so I would actually say that on that particular round, GBT 5.6 Soul has won. Then we have Dragon Realm and we have the Doom game next. Right. So, if we compare these side by side, this is Soul. Looks super nice. Really smooth to use, easy to control, easy to navigate, quite fun to play as a game. A little bit easy, but you know, you can level up and that sort of thing. But overall, pretty much perfect. Like it it doesn't feel buggy in any way. Again, smooth, easy, looks visually nice. Now, let's compare that to, for example, Fable 5. So, Fable 5's output here. Quite nice to use. The graphics don't look quite as good. There's not as much going on again as well, so I can't see any characters. There's a couple here. Here we go. Quite nice to use. Uh, I would say they're pretty much on par, but I'd say the graphics are slightly better using Soul. Next one. So, this is a racing game, 3D racing game. And so far, I would say that it's very close on the Doom game, but Soul just beats Fable 5. And then for the Dragon Realm game, Soul won by a long way. So, so far, Soul has won two out of five tests. Let's have a look at the next one. So, this is a 3D racer from Soul. Looks super nice. Quite fun to play as well. And the graphics on this one don't look quite as good. It is fun to play. Moves super fast, but I would say that it's very, very close right there. The graphics are worse on Fable 5. Soul did a better job on that, but at the same time, the game play was way more fun with Fable 5. So, I'd say that's very, very close. Maybe Soul Just edges because it has created nicer graphics. Next up, we have the crypt game. Again, Soul has done a great job on all of these, especially this one. Fun to play, easy to use, nice graphics, nice colors, smooth, everything else. Now, if we have a look at the Fable 5 generation, it's good. Like, I mean, it is pretty good. It's a bit harder to navigate. The staff doesn't move as well when you use it. It's a little bit buggy in places, too. It feels very linear as well. That's the only thing that I would say here. Next up, we have a Skyrim style game. So, let's have a look at this one. It's called Frost Fell. Now, this one is quite hard to control, weirdly enough. So, I'm trying to go straight forward, but it's walking sideways, which doesn't make much sense. And then you can see the dog is like kind of moonwalking over there as well, which is weird. So the game play looks nice, but when you actually try and play it, it's very hard to navigate and you you kind of have to switch the keyboard around. Like the keyboard changes. If we have a look at Fable 5, this is smoother. It's nicer. Looks pretty cool. Let's try and go over there. Yeah, it's looking way nicer. So, I'd say the Fable 5 won that one. I mean, all of these are super super close, like marginally close, honestly. And this is the output from Soul. Doesn't seem to work when we're using Oh, it's okay, actually. How do we change the camera angle? We can change it like that. Yeah. Okay. Yeah, it's looking good. Looks nice. And then this is the output from Fable 5, which I would say is a lot smoother, more interesting game. Like if you look around, looks visually more appealing, more fun, etc. Yeah, it looks really cool. Super nice. The Fable 5 won that by a long way, I would say. I mean, honestly, that is pretty much a draw like across the board. very similar. If I had to pick one for coding projects for me, I think my brain is trained to be like, for example, I think number one, I wouldn't use both inside codeex and and claude. I think if you were working on two pro one projects at the same time across two different apps, you're going to have situations where you overwrite the code or you delete stuff accidentally and that sort of thing. So, if I had to pick one, I'm still going to stick to claw desktop. I think it's easier for me to get my head around and build multiple things. So, for example, we're building out all these GPT56 features inside the agent OS using Claude and I think that's a much smoother way to do it. What I do think is probably the best way to get the most out of all this is you have an agent OS system where you can have everything inside one place. So, for example, like you can see right here and then we can have for example goal mode working autonomously on one project. OS and that was using codeex. But then what we can also do is whilst we're doing that we can have Hermes with codeex implemented over here and we can have Claude ready to go inside here and we could be using for example our Ultra Code in the background and so like I think the best way to get the most out of them is to have them combined side by side. If I had to pick one, I would go to Claude Fable 5 just because I prefer Claude Desktop. And either way, you can plug both of them into your agent operating system. The other thing I would say here is that when you're using, for example, agents like Hermes, you can't use clawed oorthth. So, you have to pay for the API. Whereas, for example, if you're using GP 5.6, ICS, you can easily use a olf to generate images to go back and forth with Hermes and use genic models for free with your existing subscription. So that's a better way to use Hermes. So it depends what you're really using it for. But if I had to pick one, Fable 5 difficult, very difficult test. And actually, if you see the differences here, Soul was quite good at like the action and precision. Fable 5 was better at like the the atmosphere and making things feel like a bit more detailed and interesting. Let's have a look at some more build here. So if we go to Goldie Bench, we can compare the models side by side. So, if you filter by GP516 Soul, we can see Fable 5 side by side here. But again, like how similar do they look? They're very, very similar in terms of the output. So, let's check the flight simulator. This will be a fun one. By the way, tested this out with HY3 the other day, Tensen HY3, and it didn't achieve anywhere near this sort of level. So, I mean, these two models are just in a class of their own. I wouldn't say GLM 5.2 gets close either, to be honest. So, it's creating awesome stuff. Really, really nice use. This is the output from Soul. And then let's compare this versus Fable 5. I mean, Fable 5's output here, it feels again a lot smoother, a lot more detailed, which is what we were talking about before, which is, you know, Fable 5 is great at the details. It's great at the atmosphere and everything else. And then GPT 5.6 is is good at like making the game play smooth and everything else. Let's try this one. So this is Neon City. I mean, how cool does that look? And then this is Fable 5, which is it's pretty cool in a different way. They've both got a totally different vibe when they code stuff out. Be interested to see as well how do they perform on the video creation. So we actually gave Remotion, which is a skill for creating videos into both. This is the output from GPT 5.6. It looks super nice. Like look how clean and nice UI is. It's smooth. It's got a nice progress bar at the bottom. And basically this coded out a full page as you can see. And it looks fantastic. Design and the UI is really really nice. We compare that to Fable 5. Let's open it up. I Yeah, this one looks nicer. Like it just feels smoother. It's more interesting. It's more entertaining. It's more animated. Everything is better. So yeah, both fantastic models. Glad we have both, you know, access to both. I think GT 5.6 ICS is scoring higher on the leaderboards. But again, if I had the choice, it's not just about the model itself, but often it's about the harness and where you you store it, right? So, for example, using claw desktop is so nice and smooth as an experience, whereas codeex, it just doesn't feel quite as nice. It's not quite as fun to use. Other thing I would say is like the sites option is quite interesting as well. So, this is a new feature inside Codeex where you can build websites. But yeah, that's it. Now, if you want to get both models running side by side, you can get that inside the agent OS. So, we have a section for Claude, you can use ultra code, you got artifacts, everything you built in the workspace, the chat, and then we also have codecs over here. The other interesting thing is, you know, if you're using GT5.6 six quite a lot with o you might run out of tokens. Now if you run out of tokens you can actually switch to omnirute which is free or hy 3 which is free as an API inside codeex using our agent OS system. We have goal mode running autonomously. Over here we have the workspace over here as well where you can see and preview everything that you built with this stuff. And then we have the memory system that links them all together. I would say if you want to get the most out of Fable 5 and Claude and Codeex, plug them all into one powerful system like this. So you get way more out of it. It's way more fun to use and also you can build and improve and customize this exactly how you want it dayto-day which I think is the best way to get the most out of these models. So if you want to get that it's inside the AI profit boardroom link in the comments description or go to the profitab.com. Inside the community you can ask questions, get help and support in real time. I answer these questions daily with a new video tutorial inside the classroom. You get access to all of our best training. So, for example, we have a full 6 week beginner to expert master class over here. And we have uh new daily tutorials over here that we update with new advanced stuff. As you can see inside the agent OS section here, you can get the new updates, the video tutorial, the zip file to install it, and we add new daily updates as you can see. And then you can also get four weekly coaching calls where you get help and support real time, share your screen, meet other members, and inside the map you can meet other people using codeex fable 5 and also building agent operating systems as well. So hope to see you inside there. Cheers for watching. Bye-bye. Let's see what questions we got here. What was the prompt for the 3D racer game? Um, I think that'll be on Goldiebench. Think if you go to tasks here. Yeah, you can get the the prompt over here. Yes. So you can see uh the prompts and what we asked. Can I use AI to Yeah, you can. Yeah. Yeah. We got lots of training on that inside the alpha forum. What was the prompt for that again? Again, it was it's on uh Goldie Bench. Is it remotion the skill for videos? Yeah, it's a free open source skill on GitHub. Hi Julian. Can I design a multi-lm orchestration system? Yeah, for sure. I mean, that's what we do inside the air for boardroom. So, we have the uh agent OS over here. And then we have agent orchestration systems over here. So, for example, if you wanted one LLM to root out everyone else, you could set up paperclipip like so using this system over here. You could also, for example, have a group chat where you go in between your agents. You speak to them directly. Or you could also, for example, drop an idea inside the idea pipeline and go from idea to implementation with shipped apps as you can see in like one click. So you got lots of different options right there. Fable 5 or chat GBT 516. I would go with Fable 5. Yeah, good questions today. Happy to help. Thank you. Thanks very much for the great questions. Appreciate that. What the Glad you like it. Today I'm going to show you how to rank number one with GPT 5.6 AI SEO. So GPT 5.6 just dropped today. Open eye shipped three models, Soul, Terra, Luna, and I'm going to show you some of the most powerful ways to get the most out of this to rank number one based on what's working for me. So for example, if we take a look at this keyword best AI school community, you can see it's ranking number one with this keyword over here and also this keyword inside Google AI overviews. You can also see us ranking on Google directly here too and then also on the first page of Google. So we know how to rank for AI SEO. This is something we're actively testing. These are the best systems we've set up to get the most out of GT5.6 to rank your website and get more traffic. So, let's get straight into it. And there's three pieces to this system. And the first method is using Google Search Console. So, Google already tells you for free every search where your site site shows up. The keyword, your position, how many people see it, how many clicks, and the agent OS that we built over here can actually use that system to start ranking your website. So what we can do is we can go into the research section here. We can type in a keyword that we want to find keyword data for and we can get our agents using GP51.6 inside this system to analyze our website and see where we're actually ranking. So for example, you see that it links to one of our pages here. It shows us how many clicks we're getting and how many impressions we're getting for that keyword and our average position. Right? So it gives us all the keyword data we need to make sure that we understand okay where are we ranking and then also how can we rank better in the future. So it's a pretty powerful system. Now as we go through this we can actually use this topic to create more content which we'll come into in a second. And then we have the brains which is cheap 5.6. You can use soul terra or lunar three models at three different prices and Luna can find thousands of keywords for you right terra can write the articles. Soul can do the hard part and then you got the full machine. So you can research, write, deploy in one dashboard wired to five different websites or to whatever platform you want to create content for. And this is a powerful system to get your website ranking. So let me show you an example. So we can take one of these keywords as you can see right here. And we can take one of these keywords and use that to create content over here. All we need to do is click a button and then we can generate five articles and deploy them to our website. So it's a really powerful system that we can use for AICO to get rankings quickly. We actually have a a full free SEO course as well in the description if you want to learn more about this stuff with 250 free lessons on AI SEO. So that's one method for using this. You can also use the video agent. So if you saw before, we're ranking with videos uh content. So, we can actually plug in a keyword that we want to rank for over here. And the whole agent will do the research, find the avatar, edit it all together, and plug out a video that's fully edited like this once it's completed with an AI avatar, the full voice over the top, the full edited video. Looks beautiful, easy to use, simple, plugs into our memory system as well. So, really powerful. And then what we've also got is we have codecs with GC 5.6 plugged in over here. And one of the most powerful ways to do SEO is that you could say, okay, build me, you know, a certain tool and then create the website for it. And we actually tested this out earlier today. So you see this was done 35 minutes ago. So just as an example test, let's say you wanted to build a racing game and then rank your website for that racing game. Well, you can see here that we've got the final outputs and we've got the chat here. So we can see the full page that's been created. And if we open that up, we can see the full website that we've created over here. So, it's really good at like creating websites. You can use them for design. You can actually use them to actually build out the output. So, this created a full game as you can see here using codecs in GPT 5.6, but you can also create the landing page and and rank the page as well, right? And it can all run autonomously in the background. So the way that the goal mode works here is that you can give your AI agent a goal. It will loop around and keep working on that task until it's fully completed and deployed locally as you can see right here. So the great thing about that is that we could say, okay, create content for my website and keep going until you finally start ranking for some content, right? And it would just loop around autonomously. It could run for days without you using GT5.6. which is pretty wild like that especially if you're using soul. Now also what you can do over here is everything that you create you can save inside one workspace. So if you have a look we've got the game and the page that we've created previewable inside this section so we never lose it again. We can also open it up in a new tab so it's really easy to organize. We can see the source code. We've got the preview and we can copy that or we can save it as well. Same for the website that we created and the game and nothing else. And then all of our creations with GPT 5.6 six are saved inside one workspace because the great thing about this is that you can easily preview and come back to things whenever you need to. So we created out another blog post here as you can see and then we can come back to that later. So it's a pretty powerful system for doing a SEO because you can create websites, you can rank them, you can have goal mode working in the background on a loop, you can use the SEO system to find the keywords and then generate content and also deploy it to your website. And you also have the memory system where you can get the context of all of your projects, your clients, your goals, your vision, your voice, your previous content that performed well inside one place. So for example, we zoom in here. All of these different dots are different memories inside my galaxy. And the agents can use that to create new content. So for example, it understands everything about my agency over here. And it has this all stored inside one system that we can come back to for SEO later. And the great thing about that is when you're creating content, it's all unique to you. It's all created specifically for you, your business, your voice, your brand, your goals. Really, really powerful stuff. So, the final way that I'd recommend using AI SEO with this is that you can use Hermes agent and then that can create content for you as well. And also, if you're doing link building outreach, you can actually use the outreach tool over here that we've created and then that can find the leads for you. So you could say okay uh find uh AI automation related blog that accepts backlinks or guest posts and then it will find the leads for you. As you can see right here it will be fully enriched with the AI. You can then add those to your next campaign. Inside the campaign section you can write the email campaign and then from here we can actually start sending the emails too. So we get the agent to manage our inbox, to manage our sent inbox, and to send out emails for link building as well. So it's like a full SEO link building outreach tool that you could also use for lead generation inside one place, which is beautiful. So those are some of the best ways to rank with GPT 5.6 based on the systems we've created. The pipeline is you pull the live search console data, you score every keyword, you route the work to the right model, you write all the content, deploy it to your website, and you're good to go from there. It works in five simple steps. So here's an example of what it would look like. And the great thing about that is the data now tells me exactly which keyword is one nudge from number one. We can have Luna find lots of new queries. We can have Terra write the content. We can have Salt build out the websites depending on how difficult the task is as well. You might also say is this agent OS system like difficult to install or set up? So, we've actually got over 196 pages of wins and testimonials and people learning and growing with AI automation. Loads of people installing the agent OS from our community. And for example, like Jacob, he set it up in 10 minutes. David said Julian is cooking harder on his agent OS than anybody I've seen. You've got, for example, Paul who actually created and set up the agent OS on multiple different devices. So, the point here is like you don't need to be technical. I'm nontechnical. Everyone inside this community is nontenical for the most part and anyone can do this. Anyone can set this up. You might also say, does AI content actually rank? Can it get you good results? So, we have multiple websites ranking with AI SEO as I showed you before and it works really well right inside this AI SEO solar system that we've got. And if you look at the old way versus the new way for doing SEO, the old way is like you pick keywords on a hunch or you go after keywords from an AI tool that's like 6 months old, you have a team of writers that are difficult to manage and quite hard to control and there's no quality control in that process as well. Well, if you are quality control on all the content you create, it's quite messy and timeconuming. With this new system, the data is picked from Google search console. Your agents create all the content. They follow a very specific skill so that your content is quality controlled and also created at the right standard and then everything helps you rank number one in an optimized quality controlled way. So if you want our full SEO stack from today the AI SEO solar system that's just one part of the agent operating system which is the dashboard I run everything on right. So if you join the profit for boredom, you can get everything I've shown you today. The SEO section, the Google search console data set up, the content pipeline. We have open SEO setup inside there as well. We have GPT 5.6 with codecs inside there. Every CLI that you already use, you can plug in and you get all sorts of crazy stuff inside there as well as a memory system as well. So you can get that inside the forum. You also might say, okay, SEO is about writing more content, but only if that content is actually good. And that's what our systems help you do. You also might say okay uh I need to rank on Google I don't want to rank inside AI but at this point this whole system optimizes for both of them and other people might say well doing this whole system like takes up a lot of tokens and actually if you're trying to limit the number of tokens you use inside codeex we allow you to use omniroot or hy and these are two free options you can use for creating content completely free um and they're easy to set up as well inside the system and that way if you run out of tokens or if you don't want to use chat or you don't have a subscription, no problem. You can use free systems instead. So, if you want to get that all, it's inside the AR profit boardroom link in the comments description. This is my AR automation community that helps you learn, scale, and save time with AI automation. Inside the community, you can post questions, get help, and support in real time. Inside the classroom, you can get access to all of our best trainings and lessons. If you're interested in AICO, we have a full AI SEO automation section over here. And if you're a complete beginner, you can go from beginner to expert over here. And if you want new daily updates on all this stuff, you can get that inside the new daily update section. We have the Agent OS system here. You can see when it was last updated. We update it daily. You get a zip file and a full video tutorial on how to set up. And then we add new daily guides based on what's just dropped and what's actually used for. So this is all inside here. You also get four weekly coaching calls where you get help and support in real time. Inside the map, you can meet people locally who are building out with AI agents like you. And if you want to get it all, it's inside the air offer boardroom. Additionally, if you want to get a free one-to-one SEO strategy session, you want to get us to rank your website for you, you can book that in at goldie. Agency or click the link in the comments description. And basically on that free SEO strategy session, we'll show you how we take websites from zero to thousands of traffic per month, how to get more leads from Google, how we rank our own websites, you get a custom tailored SEO game plan for your website. You'll discover the secrets to SEO link building. Answer any questions you have in real time. You'll learn the best link building strategy for your website plus how to out rank your competitors with link building. So if you want to get this all completely free link in the comments description or go to the aircraft one.com sorry to uh to Goldie agency. Thanks. Let's see what we got here. Motion love your content. Just launched uh AI assistant. What can you recommend as SEO strategy? So I mean like the strategy that we've just talked about today helps a lot, right? like that's that's pretty powerful. Uh for me, well, I have a I have an SEO agency and a community. And then for example, like we have all sorts of of different things, but we have a lot of training on that inside the AR profit warning if you want to learn more about that. Yeah, you can get our best trainings inside the community on on how to fix that. Today we're going to be talking about agent operating systems, how to get the most out of them, how to automate basically benefit how to build amazing systems that actually save you a lot of time and this is based on all the community questions I get recently from the aircraft boardroom. So we're going to jump into this. If you're wondering what is an agent operating system so this is basically a system where we can have all of our agents working together inside one place. So if we want to orchestrate our agents we can do it over here inside paperclip inside the group chat where we have all of our agents talking together. We have the pipeline where we can go from idea to implementation in like one single click using powerful systems. We also for example have Claude and Fable 5 plugged in. We have Hermes that can run with any API or OOTH profile. We have Hermes Oracle, Hermes Astros. We have a voice powered AI agent that's basically like Hermes Jarvis essentially and it can run in the background autonomously. It can build stuff for us. We have a a lead outreach tool. We even have SEO systems that can basically find keywords from our Google search console, publish them to our website. We've got, for example, a video agent that can create fully create uh edited beautiful videos. Like you can see, we've got a video uh music agent, our memory system, which has all of our agents plugged in together with the same context. So, they all understand everything about us. And this all inside one place where all of our systems are just one click away, right? It's the most powerful system that I've ever seen for building with AI agents. And today we're going to answer some of the latest questions about it. We've even, for example, created a version of codecs where you can basically plug in your OOTH over here. You can switch between APIs. So you can use GPT 5.6, you can use your O. So you don't need to use API. And then you can go back and forth for inside the chat. You got go mode that can work inside a sandbox. We tested out this morning to build out a website. game. We have all of our creations with GT 5.6 inside one place as well, so we can preview and save everything that we created. And it's just the most powerful thing that we've ever built. So, we're going to talk through some of the latest questions from the Appium community and answer them so that if they have those questions, you probably have them as well. We're going to try and help you as much as we can. So, we got a question from Kevin who's asking about how to use an agentic operating system across multiple different devices. So, let's say for example you have like three different machines, you want to set up this agent operating system across all of them. So, one of the best ways I've seen and I've seen a few members set this up is with a VPS. So I personally run it locally but if you have multiple machines and you want to give access to each one you can see that for example you can use a combination of a VPS which is a uh for example like you could use hostinger and you could have Cloudflare running with that to give it access to any device and that's pretty easy and simple to set up. You can also see another example here. So Paul actually just asked his agent to add it on multiple places so it's available on both devices and that worked as well. And then we've got another example right here where it's been set up on an agent OS as well. And Jacob actually said like it was mind-blowing how fast the setup was. So the VPS is the easiest way to share access across multiple different devices. Laurel was saying, "Okay, first what an amazing community. I've learned more in 4 days in the entire four months, which is awesome." So, congrats to you. So, now they're asking like, "Is there a way to link the AI across multiple machines?" And again, you would use a VPS for this. So, this seems to be like quite a common and complex problem, but yeah, just use a VPS if you want to give your agent OS access across multiple devices. Okay. got a question here on AI avatars. So, James was asking like, you know, how do you create your AI avatar, etc. And he was seeing like a few issues with the avatars he's created. So, we actually do use this tool. Hey, Jen. But what we did is we created like multiple different recordings. Um, trained it on that and then tested them all out and saw which one was the most realistic looking. So, I think that's the best way is like just test out different recordings and go from there. You can also use a different version of Hen and depending on which one you use, it's better and better. Jake was asking about the Hermes mission control. So, we have a Hermes mission control over here as you can see. And we have basically a chat mode where you can talk to your agent and go back and forth with it. And you can switch between different profiles as you can see right here. And then we also have the talk mode where you can have a conversation with your agent. We have follow which is a voice activated version of Hermes. And then additionally, we have Hermes Oracle that pulls in the latest news on a 24-hour basis. We have Hermes Astros that pulls in the latest news and finds relevant topics based on the keywords we monitor. And then we also have the studio where we can generate images and voice and video here. So it's like a much more powerful way to use Hermes because we've segmented each of the potential workflows you could do with that. So, if you want to get that, it's inside the classroom and then just go to the new daily updates and you'll find the Asian OS system over here. Got a question here from Joshua which is you know he's been using Sonet 5 uh so it looks like he's been using GM 5.2 he's put Sonet 5.2 but I think it means GM for a lot of blog writing and document creation and he's looking for a cheaper model. So it looks like he's comparing G Sonet versus GM 5.2. For me personally I think the Sonnet is the best writer but if you want something cheaper GM 5.2 is great. You can use the coding plan and then you never seem to run out of tokens. So if you give it a good skill on how to do SEO and you train it on that skill so it's quality control properly, then you'll get pretty good outputs from Jill and Papa 2. It'll never be as good as something like Sonet for writing, but it's right up there and it's a lot cheaper. Hey. Hey. Hey. So Aiden was asking like what's the best way to run a marketing agency with AI? said he's watched the agent OS video, wasn't sure about the best way to run it, etc. Should he be using paperclipip or Hermes or that sort of thing? So, the way that I automated a lot of my SEO agency with AI was first looking at what I do dayto-day and then trying to automate from there. So, for example, if you spend a lot of time analyzing data, then you could set up a workflow just inside Claude or Claude desktop that automates that for you. And then once you feel more comfortable with AI if you're a beginner then you can keep building on top of that. So for example week one might be like okay automate task one build a great system and workflow for that and then automate task two and task three etc. For me personally some of the things that made the biggest difference was automating a lot of the marketing. So, for example, and I'll link to the tutorials on each of these, but automating funnel pages, setting up the video agent, and also automating the newsletter creation. All these things helped me grow my marketing agency, get more leads, and save a lot of time. We have workflows on each of them which I've linked to over here. But you just want to start where you are and analyze, okay, what you working on dayto-day and then how can you automate from there? And I don't think you need anything fancy for that. Like just keep it simple and then build up We got a question from Scott about like automating or or using the agent OS whilst you're on the go. So, for example, like you could be in the gym and just use the agent OS from your mobile phone. And you actually can't do that with a VPS. So, for example, like John Mesa, he connected it to a mobile device as you can see right here. And he's got the mobile version of his Asian OS. So, you could use that. And for example, even if you were walking, you could set up Hermes Apollo with the voice implementation. And then you could use the voice implementation whilst you're on the go too. A lot of people are looking for partnerships and people to work with as well inside the Asian OS uh inside the air profit corner as well. So, it's a really good place to like meet people as you can see a lot of people super active inside there which is super useful too. We also have, for example, Mike is building out his Obsidian vault, which is a memory system for an agent operating system, too. So, if anyone's watching this, you're like, how do you set up a a memory system? How do you give context to your agents? So, what we have here is Obsidian, which is just a bunch of markdown files. And this is neatly organized into a knowledge vault where each one of these dots is a new memory and it has all the information about who I work with, who's in my team, what brands I have, what projects I'm working on. And that's really, really useful and super powerful as well. And so what you can do here is you can have all of this connected in one place. Now the way that I started with this is I literally went into Claude and chat GPT and I said right based on everything that you know about me create a project folder and markdown files inside that inside the Obsidian folder explaining everything you know about me right and use uh par as a system P AI which is a system of organizing files and everything else together and so that's how we created this and then we asked claw to like keep improving it, keep making it tidier and everything else. So Mikey has done the same thing as you can see right here, which is pretty cool. And he's put all of his writings into an obsidian vault as well, which this is like a really cool way to really customize everything that you've built, plug it into one system, and make it as useful as you can. And I think that's a beautiful thing about the Agent OS is like you can customize it as much as you want. All the systems inside there, you can make them personalized to you. Even the UI, anything else, you can make them personalized. So, fantastic example, Mikey. And just keep sharing your journey. And that's basically it for all the latest questions. If you want to ask me questions like this, feel free to post inside the AR profit community. I create a video tutorial every day like this answering all the latest questions on the agent OS and also just helping people as much as I can inside the community. Inside the classroom, you can actually get access to the agent OS over here. And we add new daily tutorials like you can see based on everything we've built. We've already plugged in GT516 which just dropped today inside the agent OS as well. And we have a full beginner to expert course over here if you want to learn more about this stuff. Inside the calendar, you can jump on weekly coaching calls, get help and support in real time, and it's just an amazing community with great vibes with lots of people sharing cool stuff. So, it's a fantastic place to meet people, learn, and grow together. There's also a map where you can connect with people locally who are building with AI agents like you. And that's all available inside the AirPuff link in the comments description or go to the airoff.com. Thanks for watching. Fables the orchestrator works hard to use on his own. Yeah, I agree. Julian, can we talk after in school? Um, I actually got to go to the gym pretty quickly, but feel free to direct message me with any questions you have and always happy to help Josiah. So, feel free, man. And that is pretty much it for me. I think we've done five tutorials today. We have been through AI SEO with GT 5.6, uh, GBT 5.6 6 versus Fable 5, the Q&A from the community, the Agent OS plus GBC 5.6, and also a breakdown of GT5.6 Terror, Luna Assault. So, thanks so much for watching. Appreciate as always. I will see you on the next one. Cheers. Bye-bye.