Hello. Hello people. Just let me know if the sound's okay on this one. I'm not 100% sure if it's going to be good sound or not. Got a bit of a different setup today. Mixing it up. And uh we're going to start building some stuff out with Fable 5 today. If you got any questions as we go along as well, feel free to ask. Appreciate everyone who joined today. Also, I'd be interested to know what do you want to build? Like, what do you want to see me build? Um, we're going to be adding more stuff to the agent operating system, polishing some stuff off. Should be a good one. Heat. Hey, heat. Hey, heat. Jerome, welcome sir. Thank you for joining. How are you doing today, sir? Let me know if you can hear me okay as well. Just so I know. Thank you. Awesome. What time is it for you where you are right now? Must be about 1:30 a.m. No. Wow, that's dedication, mate. Appreciate that. Fred says, "Interesting welcome here, Fred. Thanks for joining on LinkedIn." Yes, I agree. Late night vibe coding, why not? Greetings to you too. Whereabouts in Europe be based? Going there next month actually to Italy and to Milan. So far so good, but I think it's it will probably run out pretty quickly. I can imagine there's no you can't see how many tokens you've used on Fable 5, right? That's the problem. They should have like a a section here that says how many tokens you've used on Fable 5. Maybe there is somewhere something somewhere that I'm just not aware of. But yeah, so far I can't see any sort of option showing Fable 5 here uh for for the token usage. Exactly. Exactly. You should be able to monitor it. Hello Flo. Thank you for joining sir or pho debates. There we go. Thanks for joining Pho. Definitely. All right. Apparently, Gemma 4 12B is 90% faster according to O Lama. So, we'll see. We'll see how it goes. Before when I was testing out, it wasn't it was really slow. So, if it's 90% faster, that's going to be amazing. I can't wait to test that. Fixing some bugs on the Agent OS and figuring out what next to uh what next to build into the Agent OS as well. Hey, hey, hey. Hey, Thank you. Pho, what you working on currently? Hey. Hey. Hey. Empower, welcome here. Nice to see you, too. Yeah, it's good to have Fable 5 back. To be fair, I I did test it and I was like, I don't know if it's as good as the original, but it's still good to see it back. Still still good to test it, etc. So, yeah, it'll be interesting to see how it performs. Um, it's only available for another what, 4 days now. Um, I actually tested it on the benchmarks on Goldie Bench. It it wasn't as good as I think we all thought it was when I tested it side by side, but it's still a decent option. It's still worth checking out. I think it's ranking number eight on uh the charts actually. Maybe slightly higher now. It might be like number four. need to update the leaderboards. I retested it yesterday. Hey hey hey. Heat. Heat. baby. Hey J Parsons, welcome here, sir. How's the Asian operating system going? doing great today. Got Hermes cam going. Nicely done. Thank you very much. We created the music with um our music agent inside the agent OS. Thanks. Exactly. Using a combination, we made the music with a combination of uh Hermes and Suno API. So the way that I would look at it with Hermes Camb versus Paperclip is like you use paperclipip if you are building a team of multiple agents. So if you're using Cloud, if you're using Hermes, etc. together. Then at that point, you would use something like uh paper bunny. And then for camban, it's more for like short little tasks and also when you don't have multiple agents working together. So if it's just Hermes and it's just for like a short task, it's quite linear, then you go with something like Hermes Bye. Hey hey hey. Nice. Thank you so much. I appreciate you watching. I'm glad I helped you. Um I mean it's up to you, right? If you're if you're looking at Hermes, you're like not sure I'm I'm comfortable with what it can do, then I would say in the first place don't install it. So it's up to you. And you know, never do anything that you're not comfortable with. So if you're not comfortable with Hermes, don't install it. Don't use it. multi- aent thing inside Hermes. Uh what do you mean? So if you if you're using the cambban, you can use multi- aent profiles, but you wouldn't use multiple different uh agents from different suppliers. So like you wouldn't have Claude and Hermes working on a campband together. Yeah, paperclip can be a little bit buggy sometimes. It's just like a open source project. I understand why. Yeah, it sounds about right. That's that's always the case for BPS. Hey, hey hey. Heat. Hey, Heat. Hey hey hey. Heat. Heat. I mean, I do every day. We've I I showcase so many open source projects every day. So, I don't think we need to worry about that. Not on this channel. No steps today. I'm in a different setup today. Should be back to normal, I think on Sunday. But yeah, next couple of days just in a different setup. So, does your system hook up to a VPS? Yes. So, we have a tutorial on how to do VPS when you install the agent OS. It's on the installation page. So, you can use VPS if you want. Personally, I run it locally, but if you do prefer to use a agent OS on a VPS, they can fo says AI could be in an incredible tool for people who enjoy studying and learning, but we need to be careful when it starts listening. Sounds sensible to me. I do. Thank you so much. Appreciate you joining. Can't read anything on the screen. Nice. Could just zoom in a bit. There we go. Maybe better now. The quality is already there on 1080p. It's just cuz we're using a different screen. Heat. Heat. down. Hey, feel me. Heat. Hey, Heat. Heat up here. out. Are you feeling Hey. Gemma 4 is actually good. It's the first time I've seen it run it quickly. Unbelievable. I'm excited for that. Heat. Heat up Exactly. I think that's the key is like document everything you can along the journey. If you do that, you're always going to win, you know. Thanks so much. And also it's fun, you know, it's fun to have a community. It's fun to to share the journey with everyone. So why not? That's the way that I look at it. It's actually one thing that I regret about SEO. When I was learning SEO first time around, I should have documented everything. should have documented all the websites I did, all the projects that didn't work, all the projects that did work. As soon as I started doing that, we started getting a lot more reach across all our channels. So, if you're watching this, you are building with AI, try and document the journey as much as you can because people love stories. Reesh, thank you so much, sir. I really appreciate that. Super kind. Pho says, "Insane." Absolutely. That's what this channel is all about. Today we're going to be answering some of the latest questions on our agent operating system, covering all the latest stuff, all the stuff people have questions about, how to use it, etc. This is our agent operating system that links everything together. So, we've got all of our systems in one place. Everything is just one click away. For example, we have a a voice activated version of Hermes over here. We have Hermes Oracle ready to go. We've got a studio that can generate images, videos, and voice with Hermes. And what I do every day is just answer the questions inside our community to help you as much as I can. Also got, for example, GLM 5.2 plugged into Claude Koda. But here we have open montage that can generate uh films with AI on the spot, which is pretty amazing. and it looks absolutely awesome. And um we also have a memory system so that we can plug everything together and just have it all working together. So today what I'm going to do is answer some of the latest questions from the community which we've got over here and we're going to kick this off and basically if you're wondering okay like what is this community? This community is where we all learn, we grow together, people post questions, the community helps each other. I answer these questions each day video tutorial. Let's get straight into it. So first one we got here from AU which is I'm switching from claw CLI on local to Hermes on a VPS. Can I run all the skills I already have on it uh by cloning them as markdown files? Good question. And last anyone could give me five things that I should configure to get the best outcomes using Hermes agent. Great questions. So great questions here. Number one, for switching from claude CLI on local to Hermes on a VPS, you can take all your skill files, so all your your markdown files for the skills and transfer them over to Hermes. Hermes will learn them in the same way that Claude does. It might not be as smooth depending on the API that you use for Hermes, but yeah, you can switch over all your skills and transition them from claude to Hermes directly. And then when it comes to five things to configure to get the best outcomes. So I mean the things that have worked best for us and it depends what you work on dayto-day, you know, what you're doing etc. But if you have a look at our setup, some things that worked really well for us was the Hermes Oracle inside the agent operating system. uh being able to generate videos and images with Groco was pretty cool. The outreach systems actually giving it its own inbox so that it can send emails as well was pretty amazing. So that's three things. And then the other two that I'd recommend give it access to firewall so it can easily do web search and it sounds like you have a lot of skills and like you're always learning. So what you can also do is use forward slashlearn with Hermes agent and with that forward/learn feature it can take a guide read it learn from it and then add it to skills so that you can build out skills even more. So that's five things in total. We got a question from Ronald about Windows agent OS, right? So you can you can take this agent OS system and you can plug it into for example Windows or wherever you want. So his question is before installing agents uh do you install on WSL 2 or just install Claude and Hermes through Windows PowerShell? I'm not a Windows expert, but I think you do install Windows through power uh sorry, you install Hermes through PowerShell or Hermes, I would just go down the normal route that's recommended. I think you can install, for example, like cloth desktop, which is the official app, and then from there you can use that on Windows. Again, I'm not a Windows expert, but that's the way that would look at it. We also have a guide on how to install Windows with Agent OS which I'll link to below. We got a pretty interesting one here from Patrice. So, Patrice was talking about a project that he's built out. Basically, it's like a interactive 360 tour of an island uh by a drone, which is pretty insane. And this is this is live right now, as you can see. Um, which is absolutely insane. This is the sort of stuff people share and build inside the group. So, I mean, look at this. It's it's wild what you can do with this. Um, absolutely unbelievable. Thank you so much for sharing. Number one, it's a really amazing impressive project. Two, I'll keep building out like looks great. Look at that. It's even moving. And then you can move this around. Wow. Love that visual stuff you can do with AI. You know, when you vip code something that just looks really cool like that. So we got a question from Gary here about SEO as he's running a company and he was asking about you know he wants a SEO that can do the keyword research the posting etc. So we actually have that inside the agent operating system. So if you go to the agent operating system inside there we have the SEO section and inside the SEO section you can do your keyword research over here or with open SEO and then you can generate and deploy the content directly to your website with your agents. So if you prefer something that can do everything for you, the keyword research, the content generation, and the deployment, you can do that with the agent OS. If you don't want to do it yourself, I think something like Surfer SEO could do it for you. And then for the ads and the video ads, you could do that with Hey Jen, which I've linked to over here. A lot of SEOs in the group though, so that's good. I'm glad you joined. So Chad was posting about using autonomous multi- aent system using paperclipip Hermes alarm open router. It sounds complicated already. Um so basically they're imagining something like this. Hermes does the test, writes today's models, paperclip reads it, agents confirm which model they you implemented with. If a model stops working, the agent falls back automatically. I think the if you're using Hermes agent, it actually has backup keys that you can plug in. So you can have like two or three different models plugged into Hermes as backups in case one fails. That might be a simpler way of doing it. You might have already looked into it, but basically the way it works is you can have a primary model fallback. So what that means is it automatically switches to a different provider when your main model fails. And then you can also have an auxiliary task fallback which means that for stuff like vision for web search etc. Hermes will automatically switch depending on the setup as well. Uh so that might be simpler than the whole workflow you've worked on. There's also some GitHub extensions where it automatically roots through free models. So if something fails, it automatically roots the next one, the next one. That might be a good option too. Mike was asking about what are we doing with Fable 5, what are people building with it, etc. So, we've actually tested it pretty relentlessly with 42 different builds on Goldie Bench. Uh, you can see some of the stuff that we built with it over here. It's number three on the leaderboards. I mean, it can create some pretty amazing stuff as you can see like this like a open world game. We also have, for example, this game over here, um, which is super fun. And so, like, we we just test it out. Um, and it's pretty cool. It is pretty cool. So, if you want to check out some of our builds, they're actually on the Goldieb website. Usually the way that I work with Hermes with local stuff is I just tell it where it is and that seems to work faster. So if I'm like, "Okay, I've just taken this screenshot. It's in my downloads folder. You can check that out locally." That seems to be the fastest way that I know. And also, for example, if you take a screenshot, you can also drag it into the terminal chat. Here's another question from Mark. So Mark was asking, you know, we've got this amazing system for the agent OS. How do you run it on your phone? Really good question. So one of the easiest ways to do this is you can run it via Tail Scale. If you have Tails Scale set up, it can connect to your phone. And John Mesa actually set this up as well. So if you look at his post here, he connected the agent OS on VPS for mobile and desktop using this setup and it worked really well. So fair play to him. He created a full tutorial on how to do it. Uh so you can use Cloudflare or you could use Tail Scale as well. Personally, I don't do it simply because I don't want to uh worry about security stuff when it comes to this. David says, "Uh, giving props to Julian for following his dreams harder than anyone I know." He said, "I've been in the SEO industry for 20 years. Seen enough um of him, but as I watch him build and use AI faster and more in depth than anyone else I know, I have to give him a pat on the back to this man." So, thank you very much. I really appreciate that. I I do know people who work harder than me and push it more intensely than me personally but um I try my best. I just try and help the community as much as I can. So, thank you so much. Is it We've got a question here. So like for example about the AI websites that we have like for example if we have a look at the SEO tools that we have here and all the content that we've deployed here one of the questions we got was like you know where do you host these? How do you host these? So out of everything that I've tested the easiest setup for hosting websites that we generate is with Netlefi because you can host a lot of websites on there. AI connects really quickly to them with a personal access token and also you can see the analytics. The analytics are built into the system as well. So it's got everything you need but it's as simple as it possibly can be. So I would recommend Netify. I don't recommend something like WordPress just because it can get a bit too distracting to manage and it's quite timeconuming. I've just found Neti is a lot more simpler. Daniel was asking for the simplest way that I know to automate SEO. And I would actually use this SOP. It's as simple as I could possibly make it. Um, and basically what we're doing here is we have a system where we have a prompt for generate the keywords. So there's a property inside there for generating keywords, generate the content outline, and then publishing the content as well. So if you just want to automate the SEO in the simplest way possible, it's just a three-step process. It can easily help you find keywords, help you create content, and then help you get it posted as well. So I'd recommend that instead. Then later if you want to try something more complicated then you could use something like the hos. This is pretty interesting. So, someone actually use got the OMI pendant. Now, if you're not sure what OMI is, basically it's a system where it can automatically record what you're doing daily. And the pendant is actually a wearable which I've never tested out personally. Uh I use the the screen share that can basically like look at my screen. It can uh listen to what I'm doing and then it can record those memories into my Obsidian file which then plugs into the operating system as you can see right here. So he's planning to use that with his agent OS and how it works etc. So for me personally, the way that I use this is I use something called the infinite knowledge context engine and basically OMI will listen to what I'm doing dayto-day, see what I'm doing dayto-day and then it will write notes on that like every few minutes. From there we will plug that into Obsidian. And so from obsidium, we can plug our memories into the memory galaxy. So the way that I use it is it's a three-step process. OM takes some notes on what I do daily that plugs into Obsidian and Obsidian plugs into my memory system. I try and keep it as simple as I possibly can because I tend to find if you complicate these things, they're very difficult to manage. Whereas the full memory system that we have, the agent OS is fully automated and I don't have to maintain it. I've never updated a file inside there manually and my agents can just manage it. And then the way that it works is my agents will use that memory to give me context, but also update that memory every time I speak to them. And that system works because they know everything that I work on, what my goals are, what my projects are, and how it goes step by step. So, Liliana was asking about the agent OS. So, they installed the agent OS But there's been a lot of updates right since then. So if you want to see all the latest updates that we have for the Agent OS in in chronological order, we have that with the dates over here. So if you want to see the video tutorials of the updates in chronological order, we have them inside this section. And also inside the zip file of the agent OS itself, you can see a list of all the updates and all the modules included inside there. For actually updating the system, you can use this system. So if you've got like an old version you need to update it, that's inside the new zip file. So you can use the update MD file to update it. And then also one thing that we have here just for the community in general, but also for this uh specific setup, we have a change log. So we ch we track all the changes that we make to the agent OS day by day. And then you can see the progression of the community and also the agent OS itself in chronological order there too. So that's it. Every single question answered, my friends. That was a lot of questions, but I enjoyed it. So, thanks so much everyone for asking the questions. I personally answer them every single day. The whole community helps each other as well. So, there's always people online. There's 3,900 people in the community. There's 120 people online right now, which means you can get help and support 247 pretty much. If you want to ask me questions and get me to record a video tutorial like you've seen today, then feel free to post your questions inside the air profitable community link in the comments description or go to the air profitable.com. Inside the classroom, you get access to all of our best tutorials, lessons. If you're a complete beginner, you can go from beginner to expert over here. If you want our new daily updates, you can get that over here. And we actually have a video tutorial on the Agent OS. You can see the last update date and then you also get the zip files and a guide to install it. And then every time something useful comes out, we add a step-by-step guide as you can see right here with a video tutorial and we're just constantly updating it. Inside the calendar, you can jump on weekly coaching calls, share screen, ask questions, get help and support in real time. And then inside the map, you can meet people in your local area who are building with AI agents just like you. So feel free to check that out. Link in the comments description. Plus, you can DM me anytime. I'll see you on the next one. Thanks. Bye-bye. Yes. Uh so we got a question here from Rob which is can I use my subscriptions inside the OS like claude Gemini not API? Yes. So you can use your CLI with Oorth inside the Gemini CLI and you can use your CLI uh with Oorth for claude as well. So you can use both of those inside there. FA says fancy the worst. Take care. Have a great day. Thank you very much. We got another question here. Uh, how well are LLM benchmarks matching your experience? Do you have a favorite benchmark? My favorite benchmark, my friend, is Goldie Bench. And the reason for that is because I've tested out everything personally. And I honestly don't pay attention to benchmarks anymore. I just test it out and see what we get. And everything that we create inside here is like way better, way more interesting um than anything that I see on a benchmark. And also much more useful. I mean, we created this with Fable 5. It looks absolutely awesome. And uh you know, this is a a really easy way to just check, okay, what's working and what's not. And then we run those comparisons against everything else. If you go to the compare section here, you can compare all your favorite models and you can see the tests that I've created. And I don't know anyone else who's doing that like to compare, you know, what people are building with the same prompts with the same setup and comparing them side by side. I don't know anyone else who's testing them out first firstand like that. You know, you can see benchmarks usually the benchmarks come out from the company themselves. So it's like, well, how do you know it's real, right? How do you know it's legit? You don't. And so that's why I much prefer to just test FNL on Goldie bench instead. Thanks, Rob. Appreciate it. All right, back in a sec. Well, let's get that set up for you, my friend. Um, so Aean says, "It'd be nice to see to have exactly what you've built, but also be able to see what each model is best at and the ability to sort by that." So, let's do that right now. Um, I'll ask the agent that built the website to get that set up for you. And we'll just get that built in on Goldie Bench right now, my friend. And then there's a second one there. We'll plug that in as well. There we go. Thanks for the great suggestions. I appreciate that. All right, I'm back in a minute. Take a little break. Hey hey hey. Hey, Hey. Hey. Hey. Heat. Heat. baby. Hey hey hey hey hey hey. Yeah, that's it. Welcome evidence. Heat. Heat. Hey. Hey. Hey. Hey, Hey, hey hey. Heat. Heat. Ibrahim says, "What's that song is catchy?" That is from the music agent in the agent OS. So, we got this music agent over here and then we can just create music with it whenever we want. Plug that in. But yeah, I agree. I like this one, too. I can listen to it on loop, no problem. Good for game in the zone. Hey, hey, hey hey hey hey hey hey hey hey hey. That was uh this one is called Dcript. All right. Today we're going to talk about how to create an agent operating system. If you're wondering what an agent operating system is, this is for example all our agents working together. We got Claude, Hermes, Open Claw. We could build out our own SAS tools with these agents as well. So for example, we buil a lead generation and outreach tool using Hermes. We have the Hermes Oracle that can find the latest news. and basically automate anything that we want from here. We can also, for example, have custom builds for video agents or SEO or even music itself, right? And so what we've done here is build probably the most powerful system I've ever seen with AI and we've plugged it into one system called the agent OS. And then with the agent OS, we can basically keep improving it every single day. It becomes a system that improves every single day. And what I'm going to show you today is how to build it because I built this from scratch. I literally have no experience building SAS tools. I have no experience building this stuff out. I have never done this before. But literally from I would say about eight weeks ago, I started this project and it's become an absolute beast. It's become the most exciting thing that I've ever worked on when it comes to AI and it it's become more powerful than anything else I've seen out there. So if you want to build something similar, I want to show you how you can build it using a five layer system. So you can build your own agent operating system one layer at a time. And you know most people are running like AI is fi 14 scattered tabs all over the place. I wrote mine as one operating system stacked in five layers. And I will talk about those layers in a second. We talk about foundation memory routters agents and then the loop. And here's exactly how to build your own um mostly for free if you want to. I mean because I already have the chat uh the Claude subscription for Claude code I don't pay anything extra to build this stuff out right like it just come it's just basically I use the tokens I already got included in my plan and then we just keep building out more better stuff inside it and it becomes the ultimate tool that we can build and automate anything with and if we create too much then we can just hide it inside the settings if you want to share it we can we actually shared it with our community so we've got 3,900 members inside our AR prof. We've shared the agent operating system with everyone there. And then we've got this whole memory system as you can see that basically links everything together so that all of our agents have context. And so it works in five layers. Let me talk you through each section. This is the mission control as you can see right here. Um the actual mission control dashboard. And the way that I look at this is like every layer is built on tools you probably already have. So for example, I built it with clawed code. I've got a llama in there. You can use Obsidian. You've got, for example, GLM 5.2 plugged in as well into our agents. We have uh a router to decide which model wins, and I'll talk more about that in a second. And then we have a full agent operating system. And the main thing to note here is like, you know, I I ran AI previously is 14 different tabs. And then I built it into an agent OS. So I had chat GPT in one tab, Claude in another, a coding tool in the third. Super messy. Every session I had to reexlain my business. uh I had to explain my clients my voice from scratch. Files got lost, contents got lost, nothing got remembered and everything, you know, pretty much became a mess from there. Uh that was the old way of doing things. Right now we have a way where I stopped collecting tools and started stacking layers. So now it's one agent operating system with every model and every agent on one screen. They share one memory. So they know my business literally as soon as I start them up. So if I go to Hermes, I'm like, "Tell me what you know about my business or tell me what we worked on last week." It will give me a day by day um recount because it understands everything about me, my business, what I work on, etc. And so it's one agent operating system with every model, an agent on one screen, they share, what memory, so they know my business. Um each job goes to whichever model wins it, mostly free ones. And the agents do the real work um 247 whilst I go off and do everything in the background, right? And also the great thing about this is we created a a project which we've shared with everyone inside our aircraft boarding community. And you know at this point we've got 194 pages of wins and testimonials from people using this stuff too. So look at what Shin said right? He said out of all the school channels I've joined um the agent OS system is by far beyond worth the value it provides. So Hockey said I've installed the Agent OS and your system is amazing. Rick said he joined the school group and he's already got superpowers and that he built it in just 30 minutes. You got, for example, so many people just sharing their wins, sharing what they've built with it, sharing their customizations. Like for example, Sheena, she built a self-running lead machine using the agent OS as well. So you can customize it and and build it however you want as well, which is a fantastic thing as well. So, if you're watching this, you know, I promise yourself that you'll set something like this up because the sooner you do, the more fun it's going to be, the more powerful it'll be, and the sooner you enjoy building with AI instead of feeling overwhelmed by it. That's the thing that I would say is like it makes it so much easier to control when you've built the system yourself. Whereas, for example, if you're chasing every single model, every update, every new tool that comes out, it's going to be super messy and overwhelming. You're going to have shiny object syndrome. And I can I I imagine a lot of people watching this right now feel overwhelmed, feel like there's too much going on. So we've basically set it up in five layers. So let me talk about how this works step by step. And also I want to say here like the old ways you have 14 tabs. You have loads of different tabs everywhere. You reexplain yourself over and over again. You lose all your files and all your context. You're using all these different tools monthly. It's very hard to keep track of them. Nothing talks to anything. You got no memory, no handoff. and you are the glue holding all of your AI tools together by hand. That's simply where it come what it comes down to for a lot of people who are watching this sort of stuff, right? Whereas the new way with this system with the agent OS is like every model and agent is on one screen. You got shared memory so it knows your business. Each job is rooted to the model that wins it. Agents ship real work whilst you sleep. It compounds and gets stronger every week. And the system is a glue. you just point it and that's the difference. The system is the glue. It's not you just trying to juggle a million different tabs of AI tools all together. So the first step really is the foundation and this is every model inside one screen. So if you look on our side panel here, what I'd recommend you doing first of all is every tool you actually use, it's not every tool that you have, but every tool you actually use, plug into the agent section here, right? Have a sidebar like this. So layer one is really the ground that everything stands on. one dashboard with every model wired in, not 14 tabs in one place. And you could have a free local model for the everyday work. You can have free APIs for more. You can have the CLI you already pay for including Claude plugged in as well. So you're not paying twice. And then you've got everything inside here ready to go. So if we want to speak to Claude, no problem. We got it down here. If we want to speak to Hermes, no problem. We got it down here. Right? That's the point of this whole system. Now, if you want to how do you get started with this? You just go into Claude and you tell it what you want. So, if you're like, "Right, I want a mission control dashboard that looks beautiful with Claude and Hermes built in." You just tell it what you want. And if you want my full setup, you can get that inside the app boarding. But if you want to build it from scratch yourself, you can. And so, you might say, "Okay, well, doesn't this use a lot of tokens when you're building out an agent OS system?" No. And that's the biggest myth about it because the foundation can run stuff on free local models. It can run stuff on free APIs. And that way, you don't need to worry about tokens at all. Now layer number two is the memory and the memory um is really just one shared brain. Now we use obsidian for this as you can see right here this our memory galaxy it's improving and growing. The universe is expanding every day my friends right and so it's one shared brain and the thing that I would say here is like a model with no memory is a stranger you reintroduce yourself to every single morning. So layer number two in this whole setup fixes that because you got one shared vault, your business, your clients, your voice that every model and agent reads before answers inside the system because they're all linked. And so you stop wasting 20% of every prompt reexplaining who you are, already knows. And this is your obsidian vault. Hermes, my video agent, my SEO agent, Claude, etc. They all plug into this beautiful vault. And then one vault feeds context to every agent. So they all know you and your business for the answer. Right now, how does this work? Well, basically we've got Obsidian. And with Obsidian, you just have a plain folder of notes, basically memories as you can see right here. So this marks everything that we're working on or you know the latest news in AI, etc. Uh and you have that folder of notes who you serve, what you sell, how you talk. You point the agent operating system at it. Now I get my agents to do this for me. And if you're using agents a lot, I recommend you get it to set that up for you. You don't need to do that yourself, my friend. You can just say deployed. Okay, listen. Download Obsidian, set up a folder that's well organized inside Obsidian with markdown files that documents everything you've done for me. And that's it. That's as simple as it gets, right? And so, you might also say, well, I don't want all my data going to a cloud inside Obsidian. No problem, mate. You can use local um instead, right? So, Obsidian can be local for you and then it just runs locally. And also you can just use local models and then no data goes to the cloud at all. Now layer number three is the routter. So you want to pick the right model for the job. And layer three is the intelligence so that you stop asking which model's best and you start sending each job to whichever one wins that job. So for example, it could be a free open model for like just generating text. It could be for example a frontier model if you want to build amazing stuff. Um, now the way that I actually do this, just something we're working on longterm, but basically if if I just need to go with something basic, I can just go to the local section here, right? Or I could go to, for example, something quite cheap like GLM 5.2. And then if I need to build something that's absolutely mind-blowing, no problem, mate. We're going to use Claude and we're going to use Fable 5 and we can create some amazing stuff with that. Okay. Um, so that's what we do over there. Now, also one thing that's been super useful along this process is I test every single model and I build over 42 things with it. And then what I actually do is categorize what it's the best at. So, for example, you can see the leaderboard for Goldie Bench here. And the way that I do this is I get Claude to help me decide, okay, which models are the best, which ones perform the best, etc. Uh, where do they score against everyone else? So, you can see Fusion is right at the top here. Hermy's mixture of agents is pretty good as well. And then we got Claude Fable 5. We also categorize what it's best at, right? So you can see, for example, Claude Fable 5 creates some amazing games, like pretty amazing stuff. We have a look at Claude Fable 5 and we scroll down here. You can see that that we can see u we can build like this insane sort of games. You can see right here. It's quite fun to build with. The the graphics is super nice, easy to use, etc. Um and it looks cool as well. And we've got all these builds that we share with people so they can see, you know, what's good and what's not. And also the cool thing about this as well is that when we're building out with this, I mean, look at this. It's pretty crazy. It feels like Skyrim. Uh, so when we're building out stuff like this, we can share it, we can compare it, and we see what's the best for for which category, which is super useful as well. Uh, so that way you can just compare and and see, okay, what's the best for which thing. And we also have the build side by side here. So we can see okay what have we built with each model inside the agent OS and then how do they compare side by side and that's super useful as well. So let's get back to this and and that was part three which is the routting. You just want to understand okay which model you're using does the best work. From there you pick the agent that gets the works done. So for example, you can have one agent that plans it and then you get a agent shipped email etc. So what I mean by this when we say the agents the work that gets done is that layer four puts it to work. So you can have agents that plan build write and research on the in parallel whilst you're doing something else. So the way this works for example is if we have a look at Hermes we have Hermes mixture of agents plugged into this system. Now why is that useful? It's useful because we can have two agents working on the same thing. Then we can fuse it with an aggregator and then we can see what we've built over here. Why is that useful? It's useful because these agents can work in parallel whilst we're going off and do something else. The same with the camb board. We could give our agents a task over here with the camb board and then this fans out work to different agent profiles so that we don't need to orchestrate it ourselves. And I haven't prompted for so long. Honestly, I don't really prompt anymore. All I do is I I tell my main agent what I want doing and then that goes off and orchestrates everyone else. And I'd recommend that for you as well because otherwise if you're prompting and giving feedback and manually managing each of your agents, it's going to be a mess. Uh also the thing I would say here is I've seen, you know, some people say, well, will this work for me or will this work for my clients? With an agent operating system, it will work for any industry. The reason for that is because you customize all the workflows to your industry. How do you do that? Well, you can build your own custom workflows. Like for me, for example, I do a lot of SEO. So, we built out the SEO section. The same for the video section too. And if you have a look here as well, we've got the memory section here too that's set up nicely as well. And so, like every part of this process is all customized to me because we have the unique memory system that I talked about before. We have the unique workflows that were built in. And that brings me to layer number five, which is the loop engineering. All right. And you know I was talking before about how I don't prompt my agents. I just command my agents work. The memory grows. The operating system gets smarter and that just runs around in a loop. And so this improves and that's really the top layer here. This is what turns it into an operating system because everything it does feeds back in. Winds get logged to the memory. New models get tested and slotted into gold bench. Uh new tools plug in as new agents. You don't rebuild. You just extend what you've already got. And that's really useful because the operating system you run in six months is far stronger than today's. Even if the models change, right? And the models will change. They'll be different next week, next month, next year. That doesn't matter because we have a system that's always going to stay the same and always be useful. So that's basically it. Now, if you want to build this yourself, you could. I personally spend about three hours a day improving it day by day, building new stuff into it, making it as useful as I can. If you want to skip that whole process, you can get the whole agent operating system from me with everything you see right here inside the AI profit boardroom link in the comments in description or just go to the air profitboarding.com and inside this community uh you'll get all the layers that we've talked about the foundation the memory the agents the loop the local hermes engine token efficiency playbooks as well so you can reduce the amount of tokens you use four coaching calls a week where you can ask questions and this is just a community focus on helping you grow save time and learn with AR automation now inside the daily update section here. You can get the full agent OS video tutorial, the last update date and the zip file to install it yourself. You also get new daily tutorials as new useful stuff comes out. Inside the community, I answer these questions personally, myself every single day. And then also there's always people online to help you whenever you need. Inside the classroom, you can get access to all my best trainings. Inside the calendar, you can meet people um on a call live, ask questions, share your screen, meet, see what other people are building as well. And then inside the map, you can actually 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 go to the airbomb.com. Thanks for watching. Back in a sec. Come on. Heat. Heat. Hey hey hey. Heat. Yeah. He Heat. Heat. Hey. Hey. Hey. Hey hey hey. Heat. Heat. Heat. Heat. Heat. Hey, Heat. Heat. Heat. Hey, Hey hey hey. Heat. Hey, Heat. I'm put I think that people who were creating all these tutorials and comparing ing open claw versus Hermes side by side like uh yeah, I think I think stuff like that helped it a lot. It was really the people using it and documenting it that I think got it to where it was cuz it's probably better tools than Hermes out there that nobody knows about. But it's just like, okay, who, you know, what gets promoted the most? Well, what gets marketed the most? That's the way that I look at it. Listening to uh a song that the agent OS created. So, you can create these like cool songs using this section here. Uh so, for example, like chill work music and then it'll create it and plug it in as you can see. So, like we can click on chill to work here. Uh, let's try this out. Super chill, motivating, inspiring, funk word to something like that. Right? So, you type it in like that. Then, you're going to hit generate over here. And you can see it's beginning to create the song. So, it's going to create that now. And it would take about I think about 30 seconds on the API. And then we'll come back to that music in a minute. We're the first egg already. So, what we can do is grab that. It's almost done. And then once we've done that, we're going to go over here, plug it in. It might not be good. Just warning you now. Sometimes it's not that great, but we can test it out. We can rev around the track together. And there we go. Hello. Um, well, we I mean that's a whole system, right? How how do we post so much in one day? If you want the full systems for that, go to the classroom over here and uh we have two options for this. So you can check out this section here which breaks it down like the full system step by step or you can check out this section right here and it will show you step by step the operating procedure, the video, etc. It's all good to go. That music is not bad at all. All right, let's get this party started once again. Today we have a brand new update from Gemma 4 from Google's Gemma 4 and it's now nearly 90% faster on a Mac with O Lama using MLX and this means that essentially you can run Gemma 4 way faster than ever before. I've been testing out I'll show you what we've built out in a second. We've also plugged it into our agent operating system, so we can build and automate anything that we want. We can preview it over here. Uh, honestly, is way, way, way faster, which is fantastic. Um, we're liking that. And we're going to show you exactly what we've built with it today, how it works step by step, um, and how can you can use it, too. So, the great thing about this is that honestly, local models up to this point have been pretty bad and pretty slow. And I'm on a Mac Studio. So if I'm on a Mac studio and it's not that great, then that is a bit of a nightmare, my friend. So what we have now is we have MLX with GitHub. So MLX is a way to run local models uh directly on your Mac, right? It's designed for a Mac because usually like if you're running local models without MLX, they're super slow. And then what we also have is we have Olama. And on Olama, you can use Gemma 4. So every I just want to be 100% clear here before you know people get confused or whatever just to be very clear here Gemma for free model you can get it on Olama Mox free open source project you can get it on GitHub and so with this combination that means that you can build and automate whatever you want using these free models and they're way faster and way better than they've ever been before which is fantastic. So you can see here that they've actually added MLX models directly to O Lama. So this means that you don't even have to get MOX directly. You can just download O Lama which is free and then guess what? You get Gemma 4 which is free. And that's basically it. And then you just make sure you install one of the MOX models. Now once you've done that then you can plug it into a system like this. So we have the agent operating system and this basically takes you know it's plugged into my memory. It's got all my agents plugged in as well. So, we could run, for example, Gemma 4 directly for free with something like Hermes agent. We actually did that right here. So, we've got Gemma speed we called it because it's 90% faster and we've got that plugged into Hermes, which is perfect. And so, this is a pretty powerful system where you, you know, local models are actually good, actually useful. So yeah, my for just got faster um up to 90% faster. I will show you what the actual tests my own um tests show you when it when it comes to speed. I think from what I tested it can be up to 90% faster. That doesn't mean it's 90% faster for everything. So for example, when I was testing it out, I think for most of the stuff that we tested out was about 60% faster. I actually got Claw to measure it for us. 60% faster, but still that's pretty good for a free local model. you know, we can have it um switched on and it's ready to go. And so this is a big big jump really. Uh and I'll show you what we've built with it in a second, how it works, etc. And by the way, this just dropped. It was only on the 1st of July that this update dropped. You can see the comparison here. So we it was running at 50 tokens per second previously and now with MOX, it's running at 95 tokens per second. So big jump and improvement right there. And you can see the on Hermes when we've got Gemma speed running or Gemma 4, whatever you want to call it. Uh we said working and it's working right there. So it's good to go. Pretty useful. We can use it with Hermes. We can plug it into free clog code. We can plug it into our local model builder. And then what we've done here is we've set up a system inside our agent OS where we can ask it to build something. And then from there we can preview it and we can also see it inside our workspace which is super useful too. And we actually separate that by model. So for example when we were testing out the local model qual we have the preview over here. When we had for example gemma 4 being tested we have that over here right which is great. So, if we go back into the build section here and we say create a to-do list app and then we just hit enter, that will start thinking locally and building locally with Gemma 4 using Gemma 4 MX and then it will actually show us a preview on that once that's completed. So, we can have that working in the background. We could have it running 24/7 if we wanted to as well. And so the good thing about this as well is like everyday builds, the little stuff, the stuff where you need a lot of work going on, etc. Um, you can have that running on your own machine now. And so I can ask it, it can build, I can iterate. I could even get it to build the basics and then get, for example, something like Fable 5 to finish it off. And then for the really hard stuff, of course, you're going to use Fable 5 or something Frontier. But for the little jobs, you know, even if it was just writing a blog post or something like that, well, you could easily get Gemma for to do that. I mean, you can see here it's coning out pretty quickly in the in the background here, and it's just running autonomously without us. There is no other way to join, I'm afraid. Sorry. Uh, but you can check out our other stuff as well. You can check out our tutorials. Now you might say, does it actually run 90% faster on coding? Yes, it does. Right. So we actually ran the same prompts with the old build through Llama CVP and the new MOX build um is is a lot faster as you can see right here, which is super useful. Al so you might wonder like how does this work? So basically you ask something and that as far as I'm aware like MTV basically drafts and guesses three to five steps ahead and then it keeps the right guesses and you get the faster output. So basically what it does is it autotunes how many tokens to guess ahead so it doesn't slow down. That's how this works. That's how it it runs faster than it does versus before. And it's only a quick line to install it, right? You just run, make sure you've updated it, and then you can run this for the MOX command. So that's how you can set up inside terminal. Now you might be saying this like what what did we build with this? It's pretty wild, right? So, what we actually did here is we built out something called Dragon Realm with Gemma 4, and it looks pretty amazing. As you can see, this is using Gemma 4 to build this. Uh, looks fantastic. The way that we actually got this to build out something as complex as that is we actually used Gemma 4 inside Claude. Got it to download the model, got it to orchestrate the model, and then we built this, which is pretty amazing in itself. Uh, what else did we create? We created like a nice little landing page as you can see right here. That didn't turn out quite as well. So, you can see like the actual game turned out a lot nicer than the landing page, but still it's it's good. It just needs a bit of tweaking. And then if we have a look at that to-do list we just built over here, it's ready to go. Let's just test it out. Yeah, actually works. We can delete tasks, we can add them, and that's ready to go. And then basically how this works with the local engine inside agent OS is that we give it a command, it builds out in the background, and then we can preview over here. So we got the preview, and then everything is saved inside our workspace as you can see right there, which is beautiful. And we can do that's just running for free. This is another one. Um, so this one didn't turn out quite as nicely as well. So I think for like graphics or some something super visual, it's not going to be as nice. Um, but for like basic builds, basic to-do list apps, basic apps, uh even basic landing pages, content creation, etc. That's where you use it. So you wouldn't use it, for example, when you're building out uh something super complex with a lot of graphics. Otherwise, it's going to end out like that. But if you want something that's not so visual, then Gemma 4 can do a decent job and it runs faster. So now we can run builds for free. It's a lot faster. We can iterate it as much as we want. We can have this creating stuff 24/7 for us as well. And it's a it's the first time I've really seen like local models do something decent and actually useful. So it's great. Now, some people are going to say local models are too slow to be useful. That was true a week ago, but if you're coding with Mox Gemma 4, it's now 90% faster. You might also say free means worse. I mean, this is not frontier level stuff, but if you just have a basic build, no problem at all. And then you also might say, well, this is for for coders with big rigs. I'm literally just on a Mac. Like a Gemma 4 is Gemma 4 is designed for mobile devices. So, I think you're all good there. And that's basically it. Now, if you want to run an agent operating system with Gemma 4 inside, I mean the speed up is just one model getting quicker. The agent OS is where that fast free local model becomes your everyday engine wired into one dashboard with everything else. You get the local engine, free clawed code, the agent camb, the clawed workspace, every CLI you already have inside there, an agent mastermind where you can actually have agents talking to each other inside one group chat, an amazing memory, plus a community that's supportive and helping each other grow and learn with this sort of stuff. Then feel free to get that inside the AI profit boardroom. We actually got 194 pages of testimonials and wins from people learning and growing with this stuff. So if you want to be a part of this and you're thinking maybe it's too technical for me actually we got so many people winning learning growing with this who are non-technical and loads of people are building out with the agent operating system so I know if they can all do it and they can all build with it then you can too I'm not technical either I can use this stuff so feel free to get the air profit board with the agent operating system link in the comments description or just go to the air profitab.com inside the classroom you can ask questions get help and support real time every time something useful comes out we create a video tutorial and a step-bystep guide on it. If you want the agent operating system, we've got the last update date here, so you know it's recent. And also, you get the video tutorial plus a guide plus a zip file. And then inside the community, you can ask questions. You can get help and support in real time. Uh you can meet people inside the community. We have four weekly coaching calls where you share your screen and get help and support and meet other people. And then inside the map, you can actually meet people in your local area using local models like Gemma 4 and everything else. And if you're wondering, okay, like what can you build with this sort of stuff? So you can see here, for example, Robbie, he actually built out an amazing system with his agent OS. And basically what he's got here is like a a a free local system that can pitch uh for job opportunities and it does outreach for him. That's the sort of stuff you can build with this if you know how. So if you want to be a part of that, feel free to join the AI profit boarding link in the comments description or just go to the aiprofitboarding.com. Thanks for watching. Bye-bye. Bye. Today I'm going to show you how to run Hermes for free forever with a new update to Gemma 4 that actually makes it 90% faster. So this is a new update that just dropped for Gemma 4 with O Lama on Apple Silicon using MLX. And so with MLX you can run models 95% faster using Gemma 4. Gemma 4 is a local free model from Google and you can now plug it into Hermes in like one single click and run free models forever. So let me show you an example of this. We already plugged it into the agent OS over here and then if we go to the different profiles that we have. I usually just have one for each different API that we use. And the one that we've got set up running for free is Gemma 4, as you can see right here. Now, we just made sure it works. It responds pretty quickly. We can also build stuff with it locally. So, you can see, for example, here we said like build out a to-do list app and it build out in a couple of minutes. And you can see how many tokens it's running on as well. And then we can see the preview over here as well. And everything that we built with this is directly here. Now the beauty of this is that it's way faster than it used to be. It it used to be so slow it's basically unusable. Whereas with this system you can now run it. Now you also might say okay Julian this is just for Mac Julian I don't have a silicon whatever. Um what you can actually do here is you can use a free API on open router and open router have the 31B model available. It's a free API that you can plug into as well. So, if you're thinking this is only for people on Mac or you can't use this if you don't have a Apple silicon setup, you actually can. You just use the free API instead. So, don't worry about that. But either way, you can run Hermes agent for free. The difference, of course, is that if you run a local model instead, then it's all private. It can work offline. You don't need the internet. You can use it on a flight, whatever. And so the great thing about this is agent workflows can now run on a free local model. Hermes is one of the most powerful agents in the world and you can now run it with um these free local models. So let me show you another example. So if we pull up Hermes over here and we say okay for slasharn and let's just plug in the details of a recent tutorial that we've created like this. So we can take this example we can go to type/learn um and then we can get Hermes to start thinking and working on that task. Basically, when you use forward/arn with Hermes agent, it can read the tutorial and it can add that as a skill so it never forgets it again, right? And that's just running the background with a free local model which is beautiful, you know, so you can have autonomous agents that are just getting stuff done in the background which is unbelievable when you think about it. And so, you know, with Gemma 4 fast enough locally, I I used to avoid Hermes agent with Agentic stuff, but now with this system, we don't need to avoid it so much because we can just have Hermes running in the background. And you know, this might not respond to you like within seconds. Bear in mind, it's working on quite a complex task there and it set up a a skill set. It would normally even with a a Frontier API that would take a few minutes, but the point here is like you can have it running in the background. So you don't need it to respond to you straight away. It's not a chat bar. It's an agent. So it can build stuff for you 24/7 whilst you're off doing something else. Uh so it's great. I mean we can set a loop. We can walk away. It can build stuff. It can work on the cam. We can have multiple agent profiles building with Gemma 4 using Hermes and we can run it free forever which is absolutely amazing. And you can see an example of how to set this up. So if you're wondering okay like you know how do you run this? How do you use it etc. Well, it's pretty simple. The way that this works is you can just make sure that you use Lama with Hermes and you can run the terminal command for it. So, if you go to Lama, make sure you have the latest version. Once you've done that, go to models. Then you're going to go to Gemma 4 and the model that you want to use, this is important to note here, is these are the normal models, the old ones, and these are the new ones. So, you can see this was just updated. Um, this is MOX you want to be using. So you can run the MOX models and those are the ones that will be 90% faster. And so the great thing about this is it can just run in a loop without you 24/7 using this free local model. And here's an example what we created it with it. So, you know, we asked it for to to write a tagline for a fast local AI coder, critique it, and then improve it. So, it can run the whole loop itself. It just went off and did that, which is pretty nice right there. Uh, we also said, for example, we we pointed it at file and then asked Hermes to read a landing page and tell us what sections it actually has. And it did that right here, as you can see. So, it was pretty nice. And then we did another one. Um, you know, another one you can use is like you can say, "Okay, read my latest emails, go through my messages, give me the ideas, uh, give me the the lowdown on what's just happening in our inbox and we can go from there." U, the other thing you could use this for is like you could have it running 24/7 on a scheduled loop and then it watches like the latest AI automation news and just comes back to you. So, there's so many different things you can do this. And then also bear in mind like Hermes now has a sub agents update too. And what that means essentially is that you can basically give it a set of tasks and Hermes will just go off and do those tasks with separate sub agents. So you could have multiple Gemma 4 workers working on multiple tasks in parallel together. And just to to be clear here, you can see that we got the task back here. So we said learn with the guide details and then it replies to us. It learned about Goldiebench which is our personal benchmark and it can just add that as a skill right which is pretty amazing. So it can run really nicely with a lot of aentic workflows in the background and it can run for free. Now the other thing to note here is that it can use loops for you. So if you're not familiar with loops like in the old way of prompting agents particularly you would ask it answers you ask again. You might have to copy and paste between it and your work. You kind of have to watch it on every single step because it costs tokens. Nothing runs whilst you're away and and basically that is a very slow inefficient way of using the AI agent like Hermes. With the new way, essentially what you can do is you can run aic loops on local gemma 4 and it runs itself. So you set a goal, it can loop around till it's done. It uses tools. It can check its own work every try. What it has a full team of agents on a camb board working. And also anything that you're doing stays local, private, and offline as well. So I'll give you an example of this. If we go to our loop system inside the agent OS over here. So we go to the loop, we can switch the builder and we can switch the builder to a local model like Gemma 4 from our list. And that means that we can say okay do this job and loop around until it's done. We can use a frontier model to judge if the work is actually done. And then we can say okay run around 20 times and make sure the work is done. And on each time that you loop around do completing the goal that we ask you to do, if it's not good enough, you're gonna have to loop around again and again and again. And the good thing with Gemma 4 is that you don't have to worry about tokens, you don't have to worry about subscriptions, you don't have to worry about API costs because it's running for free locally on a loop for you. Now, some people want to say, you know, a gentic AI is expensive. It's only for for big companies or people want to buy frontier models. But actually you can use Gemma 4 now with Hermes agent. So no problem. You also might say well a small local model can't really do the work. But we've shown you multiple examples of Gemma 4 using the systems with Hermes and and running them really nicely which is awesome. And you might also say well I'd have to be technical to set this up but it's literally like a couple of commands from Lama. So all you do is you do alarm run gemma 4 and then you run the latest model with Hermes agent. That's it. Simple as that. So, you don't have to be technical to use this either. And you might also say, well, isn't using agents difficult? We actually have 194 pages of testimonials of wins of people winning, learning, and growing with our agent operating systems inside the air boardroom. So, I know like if everyone here can do it and they're nontechnical, if I'm nontechnical and I can do it, then anyone watching this can use AI agents for free forever using something like Gemma for. You don't need to be like a rocket scientist to to really configure it. In fact, you can actually just ask Hermes to configure it for you. So, it's pretty simple and easy in that way. It's never been easier to be honest. So, that's basically it. That's a whole system. I mean, we've talked about how to use the agent OS with Gemma. We've talked about how to plug it into Hermes. We've talked about why you would do it now versus like a month ago. We've shown you proof that it actually works. I've shown you some things that we've actually tested with it, like for example the forward slashlearn system. And I've also shown you how to get it for free from Olama. So if you want to get my full agent operating system, we've actually got the local loop engine. We have the loop agents inside there. We have full playbooks on how to reduce the amount of tokens to use. And we have loads of people, including myself, running local models. You can get that inside the AI profit boardroom. And every system that you've seen today is inside there along with a memory system plugged into all of it. So feel free to get the AI profit boarding link in the comments description or just go to the aiprofitboarding.com inside the community. You can ask questions, get help in real time. I personally answer them every single day with a video tutorial. And also there's always people online so you can get help and support 24/7 whenever you want to. Inside the classroom you can see all of our best trainings and latest trainings. If you're a complete beginner with this stuff, you can get a full beginner to expert training here. If you want to get my agent operating system, you can get that over here. If you want to learn more about Hermes, well, every time Hermes releases something useful, we create a video tutorial and a guide showing you step by step how to use it. And if you want to get the Agent OS, it's right here with video tutorial. You can see the last update date, and you can also see the resources included with it. Plus, we have four weekly coaching calls where you can ask questions, get help and support real time, and it's all inside the AI profit boardroom link in the comments description or just go to the aiprofit.com to get access. Thanks for watching. What we got here? Wow, so fast. Yeah, it's pretty awesome. Tobias, you legend. Thank you very much, sir. I appreciate that. If you got any questions or anything I can help you with, please let me know. Um, why Gemma 4? Why not Quen 3.6? Quen 3.6 is is too slow to run on my Mac Studio. That's why I don't do it. If it was faster, for sure I would use Quen 3.6, but it's not. It It's too slow. In fact, even Gemma 4, which is much more lightweight, I can use that on my computer only now. Only today because of this update. Like before, couldn't use it because it just wasn't fast enough. And then an core says, "Nice. Thank you very much. baby. Thank you. Not sure what you mean there. Reason to inspire, but feel free to to clarify on that. Ryan, why so much tempo? What do you mean? What does that even mean? Feel free to clarify on that. Nice. I mean, it's always mixed, right? We just mix it up. Sometimes fast, sometimes it's slow. Heat. Heat. Hey hey hey. Heat. Hey, Heat. Heat. Heat. What are we up to Today we're just uh well so far we've covered quite a lot already to be fair. Uh so we have talked about how to run Hermes agent for free forever. We've already been through uh the new Gemma 4 update with MLX. We've done a full Q&A on the agent operating system and we've also talked about how to build your own agent operating system as well. So we've talked about uh four different things when it comes to AI. been pretty productive so far. We still got to cover free clawed code with Gemma 4. Additionally, we're going to be talking about some AI SEO projects that I'm working on and what's working right now. And then we're going to be testing out a new free local model called Laguna XS 2.1. And I'm just running some tests and plugging it into the AGOS in the background right here as you can see. Um, so we should be good to go on that. But yeah, pretty nice. Nice. Uh well, I mean, if you if you had to pick between all of them and you're looking for something cost efficient, I would go with GLM 5.2. It's basically Opus 4.8 level, but the difference is that I have never run out of tokens before when using GM 5.2. So, I would stick with GLM 5.2 if you can. fantastic model. If you want to, by the way, if you want to check out all the models, if you're like, okay, I'm not sure which one to use. I'm not sure which one gives the best outputs. I'm not sure which one fits what I want, then just check out Goldie Bench because you can see everything that we've built. You can see it all inside one place. It's all super easy and um simple to to have a look. And that way, you're not guessing. You're not worrying about all this different stuff. You've just got everything ready to go inside one place. So, I'd recommend checking out because like for each model, for example, if you're like, I'm not sure if GLM 5.2 is good enough, check out Goldie Bench and then you can have a look at everything that we've built and have a look for yourself and see do you like it or do you not, right? Um, and each one of these you can actually open and check out for yourself and see what you think. And then additionally, um, you can compare the models side by side. But if you want my answer, you can check out GLM 5.2. If you want to check it out for yourself, just check out Goldie Bench and then you can see everything that was built with it. That's amazing. Thank you so much and well done, David, for just learning and growing like that. Like that's that's down to you and I respect you massively for learning from the videos that we have and learning from the tutorials and then implementing that knowledge, mate. that that's what it it's really about. So well done to you. Big respect Today I'm going to show you how to use clawed code free forever with a brand new update that just came out from Google that allows you to use free clawed code with a free local model and then also it's 90% faster which is usually the main downfall of local models. So I'm going to run you through exactly how we're running this system right here, what you can build with it, how it works, etc. You can see a bunch of things that we built with it over here. So we've already plugged this into our agent operate system and this works in three simple steps. So first of all, the update and why this is important. So Gemma 4 from Google is now 90% faster on Apple Silicon with Oama using MOX. Big speed up. That means essentially that you can run free local models, but they're also going to run faster and better than they ever did before if you're using Gemma before. Right now, if you're wondering, okay, how do you use that? And how do you plug it into free plot code? So, it's pretty simple. You're going to get O Lama, which is free to download, and then you're going to use it with Gemma 4, which is also free to download, and then you're going to install it into Claude Code, which literally takes one click. So you download O Lama that is free. You download Gemma 4, that is free. You run it with claude code and guess what that's also free. And so you've got uh the power of an agent harness like free claude code, but you're using it with a free local model that run autonomously for you 247 if you'd like. And if you want to see a demo of this working in action, so we can go to our agent factory over here. You can see it's running with Gemma 4. We can preview everything that we built over here. So, it's created some pretty nice things as you can see. And then also over here, we can ask it to build something. So, if we're like, okay, just simple like, okay, create a to-do list app, we can hit build, and then it's good to go on that as well, right? We can also use it inside the workspace here. So, if we go inside the chat, we can say, okay, build out a space invaders game. It's going to use free claude code to build that out and then it's just going to work in the background and start coding as you can see. So you can use free local models. You can also use this with Hermes agent. I tested it. The the thing to note with Hermes is like it still uses a lot of tools which means that Gemma 4 is going to be quite slow with that. But at the same time if you're just running tasks in the background or they're running on schedule, you actually don't need it to be fast because it's running in the background and then you come back to it when you're ready and that will be in like two or three hours. So that's another way to use this update as well. And the way that we're using it with free clawed code, it actually allows you to to run this with an API too. So if you don't have a Mac, if you don't have MLX, if you don't have Apple Silicon, no problem. You can still do all of this for free. The way that you would do that is you're going to go to open router over here and then inside open router just type in free and you will see Gemma 4 available for free. And so you can take that API from open router you get free API key the free API for Gemma 4 and then you would get the opensource project free clawed code we've already plugged it into the agent operating system if you want to get the setup from me but that's basically how you can use it. So you can get free clawed code and then you can route open router through that and use agent harness with claw code which is pretty insane in itself. So either way, whether you have if you're if you have a Mac and you have Apple Silicon, you can use Gemma 4 locally. If you don't have a Mac, you can use Gemma 4, but you use the free API and plug that into cloud code instead. That's how this works. And so whilst we're waiting for that to run in the background, one thing that I'd note with these sort of models, right, is that basically when you're using them, it's pretty useful, but you wouldn't like try and go for something super complicated here. So if you're using the agent OS with free clawed code or if you're for example trying to build something complicated with free clawed code then you would use a frontier model and if you're looking for like a cheap version of that you could use something like GM 5.2 but if it was something basic like write writing a blog post using it agentically if it was for example creating a landing page but then you could use Gemma 4 and it actually be pretty good and again you can see how quick and easy it is to switch to Gemma 4. Free clawed code reads from this model and you can point it in one single terminal command to free claude code with oama gem for mox. That's pretty much it. Simple and easy. We also have a local engine here. So if we go to our local setup inside the agent OS, we can also use it here too. So we can ask it to build something. Then we can preview what it's built and then it goes inside the workspace. And that way you don't even need to use clawed code. So you could use it in three ways really like with free clawed code with the API or with the local engine that we built into the agent OS. So that is the model that we're using with the agent OS right now as the local model. Again like it's not as good as Frontier models, but if you want something basic that does the job, you know, you're good to go. And so if you look at the old way, you know, every claude session costs tokens. You hesitate before the big tasks. Your code goes to a cloud. The free local option was too slow and so you wouldn't run local models. You just rely on something like clawed code, but it would cost a lot of tokens. With this new setup based on Gemma 4, you get the real Claude code CLI, but it's free for every token. You can let it run 24/7 for you locally. Your code stays private and local. You've got fast local gem of 4, 90% faster. So, it's way more usable now. And it's just like one terminal command as I showed you before to switch it over. Now, some people are going to say free clawed code must be like a watered down version. Well, it's actually the same claude code CLI. It's just that you're pointing it at a different model. You might also say, well, a local model can't write real code, but I've shown you examples of what it's actually done, and it can create some pretty cool stuff. And you al also might say, well, this sounds complicated, but I've shown you how simple it is with one single terminal command. And also, you might say, well, using an agent operating system, that sounds great. But isn't it a lot of work? Um, actually we've got a lot of people inside the AR profit boardroom using the agent OS system and they're getting great results with it. So I know like if I'm nontechnical and I can use it if they're all non-technical, we can all use this. You don't need to be technical to use it. Like we got over 194 pages of people winning, learning, and growing with AI. So it's an awesome community full of winners. And you know, you can get free models like this to work pretty nicely. Now, also you might say, okay, is this the best local model to use? So, I'm actually going to do more tests with Gemma 4 based on the new version that's faster. Uh, we've actually got a local leaderboard and every single time we test a new model, we add it inside this leaderboard and we build 42 different tasks with it to test it out. So, if you're interested in seeing, okay, which is the best model, how does it work, what are the builds like, etc., We show you everything on Goldie Bench so you can test out for yourself. And don't take my word for it. Don't take anyone else's word for it. Just have a look and see what you think of of each of the stuff that you build with it. Uh and and go from there, right? Like see what you think. See if you like it or not. If you don't like it, no problem at all. Um but yeah, you can get some pretty good outputs from these local models. I'm getting more and more into local models now because I think like they're only improving. And also I think like it's it's more and more important just to own the system rather than for example if you saw what happened with Fable 5. If you if you're relying on these models that are closed and not open the problem is like they can get taken down or removed any time and then all your workflows just break. So we don't want that. We want to use like local models. We want to get better and better at using them. And there are new ways that are coming out all the time that I'm trying to keep you updated with which is why I'm showing you this today. So, what have you learned today? Well, you've basically learned uh how to use free code, why Gemma 4 is more important than ever, how to get it working. We've plugged it into our agent operating system as well, so it has a memory plugged into that system, too. I've shown you some cool stuff to build with it, plus how easy it is to set up. Now, if you want to get my full system with the agent aware, submission control, all our agents working together, etc. You can get that inside the AI profit boardroom link in the comments description or just go to the aiprofitboardroom.com. And inside the community, you can ask questions, get help and support whenever you need to. I personally answer all the questions inside the community daily as well. There's always people online 24/7, so you can get help and support whenever you need to. If you want to get my agent operating system and more tutorials on new local models, like we've got a new model that we tested out here with full guide on how to run it locally, then you can get that inside the classroom here. And we have new daily tutorials. You can get the video tutorial and see when it was last updated. You can grab the zip file to install it and it's all available inside the AI profit border. Now inside the calendar you can jump four weekly coaching calls, get help and support in real time. Inside the map you can meet people in your local area who are building with AI agents like you. Plus you can DM me directly as well. Ask me any questions. Hope to see you inside there. Cheers for watching. GMA 426B. Um yeah I mean like have you tried it with MOX? That would be my question to you because it literally just dropped yesterday. So I'm I have a feeling you might not. Um but yeah, before I couldn't run it on Max Studio ever. So with Gemma 4 on uh on this system instead it works a lot better. We'll be joining the profit boarding. Thank you very much. Which local LM are you using inside your agent OS? Gemma 4 right now. What's going on here? We just tested out Gemma 4. Um we got a question here which is like you know if you don't have a business can you get results inside the AR profit war room absolutely so we have a path for people who have no setup at all who are starting from scratch and we literally show you step by step the methods of work and we have case studies here with actual members who started off in just the same position as we right so if you're like okay well I don't know where to start I don't have a setup etc just check out this section right here Because inside this section, we actually show you exactly how to get results if you start from scratch. Yeah. So, I mean, that's the process, but yeah, we have this fully mapped out as a course inside the profitable boardroom, and we've tried to make it as simple and as easy as possible as well. Okay, that Heat. Heat. N. I use Max Studio. So, if you're wondering what setup I've got, I've got a Max Studio. me. Hey hey hey hey hey hey hey hey hey hey hey. No problem, David. Thanks for asking. Heat. Hey, Heat. Hey hey hey hey hey hey hey hey hey hey hey. What's your daily routine like these days? Good question. So, I normally wake up about 5:00 a.m. 6 a.m. something like that. And then as soon as I have enough energy, I'll start doing a video like this live. From there about 11:00 a.m. 11:30 I'd normally go to the gym. That's usually a couple hours. Got a personal trainer down there and then get back get straight back to it. Start building everything again. I mean like I have an agency as well. I mean to be fair the agency has a really good team so there's not much required from me but you know usually there's a few admin things to finish off in the afternoon and then finish work about I would say 6 6 p.m. something like that 5:30 6 p.m. And then depending on the situation, like uh yeah, it depends what I'm doing. Like for example, yesterday flew down. It was my mate's birthday, so I surprised him for his birthday party. That was pretty fun. It was a good party. And then the usually what I'll do is, you know, something with my girlfriend in the evening after I finish work about 6 p.m. So it's kind of like it's a it's a nice structured healthy lifestyle, good routine, very focused, very simple. Try and keep it as simple as possible, you know. Thanks so much. Uh so you would just if you want the agent OS right you would grab the zip file from the classroom here if you want to install it to Hermes and then from here you can give the zip file to Hermes and ask it to set up for you. So if you give it that resource right here and ask it to install it step by step it can just go from there. Thanks very much. Trying my best. I get this question all the time. Honestly, I don't know. I I literally don't know. But what I will say is like I don't really spend a lot on tokens. The the thing that we do is we have to get every single tool and test it, right? So, for example, GLM 5.2 drops, we got to go out and test it. If, for example, something like um Hermes drops mixture of agents, okay, no problem. We have to get the APIs to plug into that. So, it's not like so much. It's not that expensive, but it's a lot of different subscriptions. Nobody else in the world is going to run the AI like that. It's just because we have to test everything and showcase everything. So, uh I I wouldn't say it's too bad. Is this agent operating system free? It is given to AI profit boardroom members only. So if you're in the air profit boardroom, you get it included as part of your membership. If you're not in the air profit boardroom and you want access to the agent OS, you can go to the air profit board to get access. Um, for me personally, I'm not based in Singapore, but I can imagine there probably are a lot of good things to do in Singapore. Well, I'm not I'm not based there. I I've got a company over there, but that's it. What's your thoughts on Quadrant 4 memory? Uh, that's not something I know much about to be honest with you, so I'm not going to try and comment on something I'm not familiar with. We try and keep memory systems as simple as possible. So, what I actually use is just Obsidian. And we use Obsidian directly with uh the the operating system that we have, the agent operating system. And that's just like bunch of markdown files, you know, simple memory, super easy to do. So for me, I'm not working on any projects. Like again, I'm a very focused person. So I I say no to like 99.9% of stuff. Um and I just focus on my own projects. So for me, I'm not looking to to work with anyone on anything. But what I will say is if you're looking for people to help you, join the eye profit board because it filters out a lot of people who are serious about AI and there's a lot of people in there who are looking to build and collaborate and work together. That's what the community is all about. So if you want to uh connect with people, you can inside the AI profitable. Today I'm going to break down all the systems that we're using right now to get results like this for our website. So this website's basically gone from nothing to hundreds of clicks per day. It's the same setup for this website as well. So you can see this website's on a nice trajectory right here. And we have another website that's doing really well with SEO here. So, I'm just going to break down what's working for me, the systems that we use, how we build them, how we set them up, and what's working for us right now. So, one of the first things that we're doing here is we're basically looking at trending topics. Let me show you an example of this. So, for example, Hermes, they recently released Hermes from agents. Massive, massive improvement in frontier level intelligence. And so, what we try and do is try and rank for that keyword straight away. So if you look at this particular keyword, you can see us ranking here and here for that particular keyword, which is fantastic. And then you can also see that we are ranking down here with our blog as well. And so with a combination of video and a combination of blog posts, we can start ranking for that pretty quickly. You can also see for example uh if we go to images here that we are ranking number one on the images and also if we go to AI mode, I can imagine we are ranking right at the top there. Uh yeah, so this is a system that we're using to quickly rank our content for trending keywords. And the reason that we go for trending keywords is number one, they have a lot of demands. Number two, we already create a lot of content about it. So it's just about making sure that we sort of categorize it in the right way. And number three, it's super easy and simple to do and there's not a lot of competition. So there's many systems for doing this. I mean, here's another example of a website we started on the 21st of May. So it's at zero clicks per day 21st of May straight out the box using this system is getting like 23 clicks a day from SEO right now this is for a website that I know some people will search for so I know it's got a bit branded search as well but again that actually helps your SEO so it's pretty powerful to do that at the same time so what do I mean by branded uh trending keywords and how do I go after that? Well I've shown you an example with the Hermes mixture of agents case study. So what I would actually do is normally I would look through the content that we're creating and I'd be like right okay out of these topics what would be really good to try and rank and and to try and create a blog post about right so for example there may be some people searching for like Fable 5 AI SEO I think that's probably something important but mixture of agents versus Fable 5 I think would be an interesting one too and then what I try and do from here is I try and like actually make sure that people are searching for that keyword. So, how do I do that? Well, I just go into Google auto suggest because there's not going to be much um data inside your existing tools like HFS. You know, it's not going to have much data on trending topics because this stuff is so new. So, we're going to have a look inside here and we're going to see okay what keywords around this particular topic are people searching for. So, Hermes mixture of agents, Hermes of agents tool, Hermes mixture of agents skill, right? So if we plug in this keyword for example again you can see us ranking because we want to rank for that trending keyword and that's the system that we use over and over again. If we type in skill here as well you can see us ranking again for that keyword. But look at this we actually got the featured snippet for that particular keyword. Right now that is something people actually searching on Google. You can see it on the auto suggest here. And so we know okay if we try and go after that keyword we've got a very good chance of ranking it. And also we can create content for it very quickly which I'll show you in a second. Right? So we got the feature snippet for this particular keyword. And again that's a trending topic but literally it only got announced less than seven days ago. So already ranking number one for it. Already got the feature sniff for it. Pretty nice, right? And then we funnel traffic from this page to our funnels. So for example here it's got a CTA and that goes to our community. And so we send traffic from our blog post relevant to our niche and we send it over to the offer we have for that niche which is the graph boarding for AI related stuff which is pretty nice. So I've shown you proof it works shown you how it's set up. I've shown you how we can get the featured snippet and also how we can rank for that keyword and also how we pick trending keywords so we have a less competition for but we know we can rank for quickly. Right? Pretty simple system. Now you might be going okay well how do you actually create the content for that particular keyword once you find the topic you want to write about. So the system that I built in here very simple system basically we can generate content that we want to rank for around that particular topic based on the case studies we already have. So if we went over here and we like right okay what are people searching for that we have case studies around this is probably quite a good one. to the XMCP server. So you can see like people are searching for that on Google which is perfect. So what we can do here is we can take that keyword and then we just need to grab the case study which we've got over here. So let's grab that case study like so. And what we're going to do over here is also get the rest of it which we've got down here. Once we've done that, we'll plug it into this system. So, inside agent OS, we just need to plug in the keyword, which we've got there. And then we're going to plug in the case study. So, we're going to select a new case study because it's brand brand new stuff that we don't have anywhere else. So, we want to take that, plug it in over here, and then we can also save that and reuse it. So I like to save my case studies just because there might be like different variations of that keyword that you want to create content around again in the future. So if we have a look at that Hermes mixture of agents keyword again, you can see that for the same case study, you could create three different articles, right? You got Hermes mystery of agents, Hermes mystery of agent skill, Hermes mystery of agents tool. Perfect. That's really good. So what we're going to do over here is we're going to plug in the keyword, we got the case study, we're going to hit generate five articles. And the important thing to note here is that we are adding information gain. We are not using chat GBT to write the content. We are adding information going with our case studies. Why is that important? It's important because if you're just using AI without a case study, without any sort of experiments you plugged in, the problem is that AI can already generate that, right? Google overviews can already generate that. So what makes your content different is the case studies, the experiments, unique data that you add. It's why for example we're constantly updating stuff like the goldie bench website because all the experiments I deal with AI I'll document them here and then again we'll use them for case studies in the future as well right so it's really important to have information gain and just document everything that you do espec whatever industry you're in because then you can use that for SEO in the future and I don't see anyone else talking about that but it's so easy to do why would you not do it I mean it's the same reason that we have our memory system over here right we have a memory system that automatically documents everything I work on plugs into a memory database that gives all my agents automatic memory and context and I don't need to do anything. So it's simple to set up. It's easy to automate. You just have to make sure you do it. So from here we're going to click on generate five articles and what that's actually going to do is start writing the content for our system over here. Right? Um so once we've done that then it will deploy that content to our website over here and we're good to go on that. Right? And so we can see our history. We can see what we've created. And this is all inside one setup, one system. Pretty easy and simple. And then the finished product once it's completed is something like this. Right? So this is for the keyword Hermes desktop installation, which again we are ranking for that keyword on the first page over here. Right? So actually works. I've shown you proof it works. And it's just taking our case study and our information again and then organizing that into a nice guide that's super useful for people to read. And so the great thing about this as well is like we don't have to log in. We don't have to use WordPress. We don't have to configure any plugins. We don't have to be technical. We don't have to design the page. We literally just come up with the keyword, give it some unique data, and then the agents just go off, design the page, publish it, etc. And that's exactly what we're doing for websites like this and also for websites like this that can grow so quickly. And I think that's the future of SEO. It's like you just, you know, if you if you look at this point, anyone can create content. Anyone can uh use chat GPT to automate SEO. It's like what's the difference? It's the unique data. It's unique systems that you have. It's a unique keyword research you do, which is what I've shown you today. Uh another system that we have as well is like we can pull in our latest data from Google search console. We can look over the last 7 days, 28 days or 90 days, look for a particular keyword and then generate research based on that. So we have these really powerful systems that we can use for AI SEO. So if you want to get my full system for this, you can get it inside the air profit boardroom. This is an agent operating system. So basically what you have here is all of our systems for every single tool that we use inside one place all one click away. So if we need to do SEO, it's over here. If we need to create a video over here, right? If we need to, for example, generate something with Heromy's agent, it's over here, right? And so we can organize everything in a very efficient way and everything's just one click away. Plus, we can save our work and save our workflows, save lots of time. One final thing that you might say as well is like, do you quality control this stuff, Julian? Like, how do you make sure that it's actually good content? So, if you look at that page that we just showed a second ago, like it reads better than a human would write it. So, how do we achieve that? We have a very detailed skill built into the system that allows the agent to quality control the everything that we're doing here. A human would not be able to do this. There's like over a 100 different rules inside here to make sure that everything is formatted and set up properly. I mean, if we check out this guide, this is a 5,000word skill. No human would be able to implement that. But because AI, you know, you got Fable 5, you got Claude Opus 4.8 okay, are so powerful. We can quality control our work by having a fantastic system in place that make sure the content we create is the best it can possibly be, far better than any human could produce. So, if you want to get my full system for this, feel free to get it inside the AI profit boardroom link in the comments description or go to the AI profit.com. Inside the classroom, we have the full agent operating system that you can use to automate and build anything. Like you see, you can see the last update there. You can see a full guide on how to use this. We've got a zip file to install it, a new daily tutorials like you can see right here. Inside the classroom as well, we've got a full section on AI SEO. So, if you want to learn exactly how we automate AI SEO and all the systems we use personally, you can get that inside here. If you want to connect with other SEOs, we've got loads of SEOs inside the community and you can post questions. I answer them with daily tutorials. The reason we actually built out the SEO system in the first place was because so many people were asking for it. I was like, right, let's just make sure we automate it and share with the community. And then you also get four week of coaching calls where you can ask questions, get help and support in real time, share your screen, etc. Inside the map, you can connect with other SEOs locally near near you using AI agents just like you. And that's all inside the prof, which is not just for SEOs, but it has a lot of SEO and stuff inside there because so many people involved in SEO joining this community. Also, if you want to get a free SEO strategy session, if you're like Julian, this sounds fantastic. I have no time to implement it. Then you can book in a free SEO strategy session at goldie. Agency and on that call onetoone you'll get a free SEO link building acceleration session where you'll get a free SEO domination plan. Discover the secrets SEO link building. This is a onetoone with a person where you can ask one of my team of experts I've personally trained any questions you have. You can get your questions answered in real time. You'll learn the best link building strategies for your website. How to outrank your competitors based on what's working for us. And that's all inside the air. One final thing I want to say is I've shown you proof that we've ranked on Google, that we ranked inside AI, that we've ranked with feature snippets and also inside Google's overviews and also the images. So, however you're trying to rank, this system works and you can implement it today. Feel free to check out the agent OS in the profit boardroom or get a free SEO strategy session. Thanks for watching. What is your take on um yeah absolutely absolutely does how would you do keyword research differently today I've already shown you today how we do keyword research differently using the trending topics method do you have a discord we have the AI profit boardroom link in the comments description or go to the profit bomb.com is your agent OS free no it's only for AI profitable boarding members. It's the only way you can get access to the aircraft uh to the agent OS is inside the aircraft boardroom. Thank you very much, sir. Awesome. All right, let's see where we're up to with Laguna 2.1. Looks like it might be ready to go. Today we're going to be testing out a new local model called Laguna XS 2.1. And this is a model that seems to do pretty well. It's pretty fast and easy to use as well. We've already installed it and plugged it into our Asian OS. show you how it's working in a second. And it just got updated today. So, this literally just dropped. It's available on hugging face and it's pretty powerful from what I've seen so far. Now, it is quite a lightweight model. It's pretty chill to use. You can see how it performs right here on the benchmarks. So, you can see, for example, Laguna XS 2.1. There was a previous version called XS uh.2 and you can see how it performs versus Quen 3.6. Now, if you're familiar with local models, you're probably familiar with Quen 3.6. six which is very very popular and north mini code which I've already tested and done a showcase on that already. So if we have a look at them side by side for S sw bench verified bear in mind this is a coding model is a model designed for coding uh you can see here that it's not far off quen 3.6 so it's a little step up from xs2 but it's pretty close to quen 3.6 six. And then you can see for example, North Mini Code, which was actually pretty decent when I tested it. It was quite fast as well, is out is being outperformed by Laguna XS 2.1. Now, you might say at this point, Julian, I don't care about benchmarks. I don't pay attention to benchmarks. I want to see what you've tested. No problem. We're going to come on to that in a second. So, don't worry, my friend. But I just want to show you how it looks side by side. You also might say, "Okay, well, how does it compare against like, you know, other models?" So Claude Haiku 4.5, which is not that relevant now, but I can see why they've added it. It's obviously, you know, local models are not Frontier models, right? So, you know, if you're if you're hoping that local outperforms Frontier, I would probably wait a long time for that. But if you want to see, okay, how does it perform versus something slightly more comparable like Claude Haiku 4.5, you can see that it does outperform Haiku on SWE Pro. It also outperforms GBT OSS by a long long way. So this is GTOSS. Um this is um Laguna XS 2.1. So overall it's an exciting model. I'm excited to test it and let me show you what we've built with it so far. So just to recap here, it's from a company called Poolside and they've released Laguna Access 2.1. This is a new agentic coding model. So it's designed for agentic coding and terminal work. Now it has a 256k context window which is enough for a local model to be honest. And if you're wondering, okay, what did we build with it? Let me show you what we got from it side by side. So we created three different um different pages with it. So this is the first one which is kind of like a simple landing page. Honestly, I've done the same test with something like Gemma 4 and this actually turned out better than Gemma 4. So I would say, you know, for aentic coding, for stuff like building landing pages, for creating, for example, simple mini apps, this is not a bad alternative at all. I mean, it looks quite nice, quite clean on the page. Again, it's not going to be like Fable 5 level, but hopefully, if you're into local models, you don't expect that either. Now, I've also got another app here. This is like a to-do list app. So, we can type in our to-do, and then we can add it inside here. Again, like if you look at this UI is not that nice. So you either want to lower the expectations on UI or train it on a really good skill on how to use AI because straight out of the box it's not going to create something amazing. So if we have a look at this and we're like okay um check out local models add it. It actually works which is great. We can delete it. We can add it. We can see what's active, what's done, cleared, etc. So it actually works quite nicely. Um and then we have this page as well. So this is another page that we built out uh with the local model. And again, if you look at this, it's kind of like, okay, this is something that I could create with Claude last year. But again, it's it's usually about a year behind when you're using local models. So is it fast? Can it code? Does everything that built work um actually work? Yes, absolutely. So it's not bad at all. And again, I'm running this. You might be saying, "Okay, Julian, what setup have you got here?" So, I'm running this on a Mac Studio M4 Max, 36 GB memory, and it seems to run pretty smoothly, which is nice as well. Uh, also something to note here, not good at all for 3D models. Now, you also might say, okay, well, how does it perform versus all the other local models? So, this is brand brand new. It just dropped less than 24 hours ago. We're going to plug it into our benchmarks on Goldiebench along with the newer version of Gemma 4 and MLX. It runs 90% faster. I've already done another tutorial on that, but basically we'll be testing out with 42 different builds and seeing where it performs. You can see who's top of the leaderboards here. And if you want to check out what we're building with local models and how it works, etc. You can check out everything that we've created inside Goldiebench and you can see side by side what we've created, how it performs, how good it looks, etc. And then just test out for yourself as well. always test this stuff out. Don't believe the benchmarks. U that's why we created Goldie Venge because we want you to see what we've built with it instead of just relying on some data that came out from the actual company. So yeah, for for building mini apps, for building landing pages, this does the job. You might also wonder, okay, how can you actually use it? How does it look like in practice? So there's a couple of ways you can use this. We've got it built inside our local engine inside the agent OS and then we can also use it inside free clawed code. So free claude code can use this with Gemma 4 or you could plug this into for example uh Laguna XS2.1 and then everything that we create is plugged into the workspace over here. So we can see everything that we built with the workspace. Um and yeah, it's pretty easy and simple to use. So, however you want to use it, you can plug it into the local setup here or you could use it with free claw code or you could even for example have a separate Hermes agent profile with this local model and then just use it that way as well. So, you could use it aently with Hermes, you could run it with free claw code or you could run it as a local engine inside the agent OS. You have three different options for using local models and how to build with it. So, it's pretty cool. And also bear in mind that like if you're using Laguna XS 2.1, you can run it locally for free, run it offline, you can get uh you can get it set up on a Mac and also if you yeah, if you're on a plane or you don't have Wi-Fi, you can still use it and everything that you create with it as well is not going to the cloud. It's your own data on your own machine as well, which is super useful. The other thing that you can do here is like you could use it inside agent operating system and then you could have it plugged into your memory galaxy so that it can use that context to create and build better stuff as well. Uh it's also available on open router and you can get a free API with it which is pretty interesting too. So if you have a lo if you have a setup that's not great for local or you just don't want to run local models because it can slow things down a little bit then you can use the free API that literally just dropped today on open router. So if you're like ah I w you know I'd love to run local models or I'd love to run free models but I can't because of my setup. No problem. You can use open router instead. And again you could plug that into free clawed code. You could plug that directly into Hermes agents and it would be pretty simple and easy, right? If you wanted to plug it into Hermes agent, uh the way that you would do that is you would just go into your terminal like so. Type in Hermes model, select from the drop-own list, for example, something like open router and then from here you would just give it the name of the API key that you've plugged in right there. Right? And then you can get a free API key from open router and you can use Laguna for free as well. So, I think that's pretty good. Like, you get the free API or you get the free local model. Either way, you're living the dream, my friend. So, thanks so much for watching. That is how to use Laguna XS 2.1, how it performs. Again, I'm going to test it on the local leaderboards, see how it goes. Um, and if you want my full setup with the agent OS, you can get that inside the AI profit boardroom. This comes with all your agents plugged together. So like Hermes, Claude, we got for example a voice activated version of Hermes over here with the full transcript and all the builds that we've created with this. We have a talk system where you can actually have a live conversation with Hermes. We've got different profiles inside the chat here. So if you have, for example, Lagoon Access 21, you could create a new profile for it. With Hermes Oracle, you can get the latest news, create content for SEO with it. Check out the latest news on X and this automatically refreshes every 24 hours and then you can draft social media content with it too. Also got a studio where you can generate images, videos and voice notes with Hermes agent. Uh we have the mixture of agent setup here that can beat Fable 5 benchmarks too and an outreach tool as well. Right. And then we have the local engine. We have the memory setup and everything else you need to win with AI. So if you want to get that feel free to get it inside the air profitable boardroom link in the comments description or just go to the airruffboard.com and this is my community for learning growing and scaling with AI automation. If you like local stuff we actually have some amazing people in there building awesome stuff. So you can see for example like Daniel he's sharing what he's creating with AI avatars locally. So he's generating videos locally using AI models as well and he breaks it down into like a super detailed tutorial. So we've got some absolute legends sharing the best stuff inside there. And then also you can see for example we have Robbie who actually created this whole system which is like a a local outreach system. So it's like a lead generation system for job opportunities and then it actually finds the leads and reaches out to them using local models as well. So if you like this sort of stuff feel free to get the profit boarding with the agent OS plugged in there too. Inside the classroom you can get access to our new daily tutorials. Plus, if you want the agent OS is right here with video tutorial, the last update date and a zip file. And then also, if you like stuff like local models, we've got new tutorials on all the latest local models dropping in. Inside the community, you can post questions and I personally answer them with um with my thoughts. Plus, our community helps each other 24/7 as well. So, there's always people online ready to help you whenever you need. Inside the class, you get access to all my best training. If you're a complete beginner, you can go from beginners to expert over here. And also if you want to get coaching calls you can check out the week the calendar. We have four weekly coaching calls where you can share your screen ask questions in real time inside the map. You can meet people in your local area who are building with AI agents like you. And this is all available inside the air profit boardroom link in the comments description or just go to the aiprofit.com. Thanks for watching. Let's see what we got on the final questions here. How can I use GLM 5 for free? Uh you can just go to Z AI my friend. So just go straight to Z AI and then boom Shakalaka you have Z AI and GM 5.2 for free right you get the flagship model right there you can start using it obviously tokens would be limited but you can use it up there. How can we build something which is really good? Uh well you could just go into whatever you want to build you just go into claude and you ask it to build it right that's it's as simple as that. That's what we did with the agent OS. Like you have to start where you are. So if you want to build something amazing like we've got with this system right here, you know, he started one step at a time, but basically I don't code. I don't have any experience with this, but I can easily build something like this out. And the way that we did it is just by describing what we want and the the vision and the mission that we have in Cyclude and and he built out step by step for us. So if you want to build something great, just start one step at a time and iterate on it daily. And I spend about three or four hours a day on this just improving every day. So that's what it takes. But yeah, 100% worth it. All right, peeps. We've answered every single question, every single comment, every single tutorial we've been through as well. Appreciate you watching. I will see you on the next one, peeps. Cheers. Bye.