So, we have a new announcement today as we're looking at a new model from Anthropic coming out. So, basically what's happened here is that obviously Fable 5 it got banned and that only got released on June the 9th. So, you see June the 9th, Fable 5, Fable 5 and Mythos 5, they both came out. Obviously, public couldn't access Mythos 5. It was only for Project Glass Wing, but Fable 5 was pretty insane. We were all having a great time vibe coding with it. Then it actually got shut down on June 12th. Uh June the 12th, not the 12th. And then from here, we have announcements of a new model. Lots of rumors out today saying that a new model from Anthropic is about to drop, which is pretty exciting stuff. So, we'll talk through what it could be. There's been some announcements that basically on tests Claude Sonet 5 has appeared on a partner provider. So, you can see the news right here. So, we are looking at Claude Sonet 5. Other people are saying it might be for example something like Mythos 5.1 or Mythos 6 etc. Pretty interesting but it was actually spotted today. So you can see here in the news this from Astropo Claude Sonic 5 is back. A new Claude Sonic 5 slug spotted today. The exact same thing happened in February and we ended up with Sonic 4.6 six which then obviously uh got nerfed later but essentially we are all waiting for this new model from anthropic to drop today you see another announcement is from Andrew Curran so a new more capable version of mythos has emerged from training we don't know whether it's going to be called mythos 5.1 or mythos 6 or if anthropic will keep it internal to accelerate further development but it has arrived And this kind of makes sense, right? So, obviously, Fable 5 got taken down. I think they're going back and forth trying to get it back. It was huge for Anthropic, you know, pretty excited, but at the same time, if they can't get their model back, then why not just create a new one and release that instead? And you can see the tiers here. So, you've got Haiku, you got Sonic 4.6, Opus 4.8, a Mythos class that dropped earlier and then we're looking at the new one that should be coming out very soon. And if we look at this as well, this class level that we're looking at, there's two different types. So there's Fable 5 with the safeguards on that went out for public release then it got taken down and then you have myos 5 with the safeguards lifted but that is only for project glasswing cyber defenders etc. And this was the announcement on the model getting shut down. The stated reason was national security and basically the government believed someone have found a way to basically bypass Fable 5. And then recently a new model has appeared. So this is the part that everyone's talking about right now. This literally just got announced a few hours ago but you know in terms of the rumors. So Andrew Curran posted about it. Qusonic 5 appeared inside the tests as a provider. The interesting thing is it going to be Mythos class or is it going to be Sonet or is it basically Cord Sonet? But it's basically Mythos class. You see what I mean? There's no confirmation on that. And obviously on the official website, there's no Claude Sonic 5, no Mythos 5.1, no Mythos 6 listed yet. So these are rumors. It's not fully confirmed. and bear in mind as well, we're looking at GPT 5.6 coming out soon as well. So, could be a busy week. If we're looking at if we get Sonic 5 and GPT 5.6 in the same week, that is going to be huge. So what's actually confirmed? Let's talk about it. You know, Fable 5, Mythos 5 launched their war model. They got taken down. Everything disabled. What's still just rumor? A new more capable Mythos finished training. Claude Sonic 5 slug spotted on a partner. Could be Mythos 5.1, Mythos 6, or Sonic 5.6. Uh, sorry, Sonic 5. might just stay internal. This is something to note as well. They might have created it already, but just keep it internal. And then we could be looking at a July roll out after safety checks. Bear in mind that Sonet sits way below the Mythos tier. So, if this comes out, it's probably not going to be anywhere near as good as Fable 5. That's the problem. But if it's Mythos 5.1 or 5.6 or six, then we're looking at something super powerful. Main thing I would say that I learned from all of this, you know, if we just take a step back is like, you know, if we step back from the rumor for a second, the real lesson here has nothing to do with the model name. So, you know, we had the most powerful model on Earth. 3 days later, it was gone from the planet. And not because it got worse, but because of a force completely outside of anyone's control. So when you look at this situation, it's like, okay, when Fable Fire was pulled, if you built on a system, you can keep building, right? So for example, we have the Agent OS system and we can keep moving. You can just swap to the next best model. If you build on the model, that's where you get stranded. So whatever happens with these models that come out this year, I would recommend that you build on a system. Don't build on the model. Two totally different things. And I'll show you an example. So if you are building on a system you've got something like an agent OS you can have all your models plugged in whatever happens okay claw gets taken down no problem let's plug in GM 5.2 to if you are building on the model. So for example, if you have everything built on one single model, that's kind of where you can get stranded and that's where for example stuff like Fable 5 can be really annoying. So the people who built their work directly on Fable 5 hardwired to that one model woke up stranded their tools just stopped particularly anyone you know relied on that. The people who built on a system, a model agnostic agent OS holding their memory, the workflows are agents didn't even break stride. You know, the moment Fable 5 went dark, they just pointed the same system at the next best model and keep shipping. So, you have the same memory, same prompts, same agents, but a different engine underneath. And that's the difference here. And that's where it flips from survive to one because a system that always points to the best model available quietly gets better every day. So when a new model like for example Claude Solid 5 comes out, they upgrade by changing one setting whilst everyone else rebuilds from scratch. And that's really the shift I would say here. The model is a rented engine. It changes every few weeks. Uh it could be pulled, priced up, it could be switched off by people who aren't you. And the thing that you actually own, the durable asset is the system built around it. So you want to chase, if you chase the best model, you're a renter. one policy change away from Stranded. If you own the system that swaps any model in, for example, like the Aentic OS, then every release, every model that gets dropped, every price increase, etc., actually works in your favor. That's the way you want to look at it. And if we look at the timeline here, June the 9th, Fable 5 and Mythos 5 get shipped. Then on June the 12th, US takes it down. And then today, 5.1 6 or internal. Nothing is official yet. These are all rumors. So whatever happens next, be ready. The day actually drops. News like this lands every couple of weeks now, sometimes every day. The agent operating system inside the aircraft boardroom turns Claude, Hermes, and the rest into one system with shared memory. So the second a new model is real, you're using it whilst everyone else is still arguing about what it's called. And inside the aircraft, you get the full agent OS zip file. Every model wired into one system that you control. You get everything as you can see right here. We've got claw builtin, we have open claw, Hermes, we have Hermes, Jarvis, the talk mode studio, we have GLM 5.2, we got Fusion builtin, and we have everything that you can see right here to just really make the most of this stuff. So, if you want to get that, it's inside the AI profit boardroom. This is my AI automation community that's designed to help you grow, learn, and scale with AI automation. It's an awesome community. I personally answer the questions inside the community every single day. Inside the calendar, you can drop a we coaching calls. If you want the latest update of the Agent OS, you can get it over here. And you've got our last update date, the video tutorial, and the SI file to install it. Plus, we add new daily tutorials like you can see so you can get the most out of these models. And if you want to meet people locally, we actually have a map where you can meet people near you who are building with this stuff as well. Cheers for watching. >> [music] [music] [music] >> Let's see what questions we got here. Morning. Good to see you. Hello. What's your technical setup? What devices do you use? What subscriptions do you use, etc.? Well, I mean, I would say, you know, I I basically get everything cuz I need to test everything, right, when I'm creating tutorials like this. So, I wouldn't say like you want to copy my stack. If you want to get the full list of tools that I personally use, we've got a list inside the app of one that we keep updated week to week. What I will say is that Claude is is probably the number one one that you want. Hermes is pretty good, too. Uh for technical setup, I just have a Mac Studio. Fable should be back. Let's see. It's been a while, but someone leaked Mythos. So, someone like created a theoretical version of Mythos on GitHub. What is your PC spec? Do are you run local models? Local models are not very good. I run a Mac Studio. I just don't find local models that useful when I've tested them. Build your system on a weaker model such as Miniax M3 in case you lose access to top models. Exactly. Exactly. So, you want to build the system, not build on the models. Totally different framework. >> [music] [music] [music] [music] [music] [music] [music] >> Uh, usually I just go straight to Twitter. Twitter is pretty good. Usually I most [music] of the things that are rumored on there usually come out eventually. So [music] usually that's where we see the the breaking news I would say. [music] [music] >> [music] [music] [music] [music] [music] >> Hey, [music] [music] hey hey. [music] Hey hey hey. [music] [music] [music] Heat. Heat. [music] [music] >> [music] [music] [music] [music] [music] >> Natalie. >> [music] [music] [music] [music] [music] >> for me. No, personally, just because I'm really focused on my own stuff. [music] So, I'm just fully fully focused on what I need to do. Um, and I say no basically 99% of the stuff that comes. But awesome stuff, man. [music] Congrats on setting that up. [music] >> [music] >> Hey, [music] hey. [music] >> [music] [music] >> Heat. Heat. [music] >> [music] [music] >> building out the agent OS. Now we're going to set up some systems with local [music] models [music] potentially. Uh yeah, I would have a look, you know, I would ask people in the group, okay, [music] you you interested in this or not? See what people want to do. [music] [music] >> [music] [music] >> Yes. [music] Usually I get about 10,000 15,000 steps in a day. So you should do that on these sessions. [music] Let's see what we're up to currently. [music] [music] About 6,000 steps so far. Not Not too bad. We're getting there. [music] >> [music] [music] [music] >> Hey. [music] Hey. Hey. [music] Yeah, of course. I've been on his podcast, interviewed him a couple of times as well. >> [music] [music] [music] >> Yeah, building that out with ponytail [music] now. So, we're going to test it, see how it goes. >> [music] >> I've also used head room. Head room is not bad. [music] Nice. [music] [music] >> [music] [music] [music] >> Heat. Heat. [music] >> [music] >> Thank you. [music] Heat. Heat. [music] [music] >> [music] [music] [music] [music] [music] [music] [music] >> You've already got access to Miniax [music] M3. Can you compare that versus the rest two? [music] [music] Let's compare. Let's compare. Claude Code versus Hermes agent GLM 5.2. We're just calling out random names here. [music] Yeah, we set up uh Goldie Bench. [music] Goldie Bench is uh something that we're working on where we're comparing the models here as you can see side by side. So, let's test out Minimax M3 as well. Already got access to that, so we can test it. Then also, we need to fill out the visuals for these. Can you pin the model names in the leaderboard so that when you scroll down you can still see them and made them sticky. half the images don't appear here. So, there's not previews for everything, you need to fix that because otherwise the leaderboard looks empty. So, anything that you haven't generated yet, you need to generate. And you should be doing that now. [music] And if you [music] nice literally everything that we've done has saved so much time, it's crazy. It's pretty wild. As you can see, like everything that we do is proactive and we just build more and new useful stuff every single day, right? [music] Whereas the old stuff I used to work on, we don't do it anymore. So that's the way that I look at it is like you save time and then [music] you can continue leveling up or you can take it easy. And for me like I want to continue leveling up. So >> [music] >> Did you build anything with this? I can't see anything. [music] >> [music] >> You should do more tests on the leaderboard based on everything you can build with it and then add it to Goldie bench as well. Uh we don't really do stuff like that like you know if you believe in it you believe in it. If you don't, you don't. [music] Heat. [music] Heat. [music] Heat. Heat. >> [music] >> Heat. [music] Hey, heat. Hey, heat. >> [music] >> Uh, so you haven't seen the rumors yet, bro. If you rewind to the start of this, you'll see the details on here. We've already covered it. [music] [music] Heat. Heat. N. [music] >> [music] [music] >> Heat. Heat. [music] Hey, hey hey. >> [music] [music] [music] >> Do you always use O Lama Gemma for? No, not usually to be fair, but I want to test this model. It seems interesting. I want to see what we can build with it. So, let's see if it's actually any good as well. >> [music] >> Hey, Heat. Heat. So, you need to do these tests yourself and you need to show proof of the tests. Have you done that or have you not done that? I think I've already done that actually. Uh quite a few videos like that. If you have a look, the main thing I would say is like if you if you want to get the most out of these models, all of them, don't rely on one single model or one single harness. Put it inside the agent operating system and then you can have them all working together. That's the way that I look at it. New model is not released yet, but we're looking at Sonic 5, maybe Mythos 5.1, maybe Mythos six. Those are the the rumors announced today. >> [music] [music] >> Heat. Heat. >> [music] [music] [music] [music] [music] [music] >> Heat. Heat. Heat. [music] [music] Hey, heat. Hey, heat. >> [music] [music] >> I didn't see that. That's interesting though. Hopefully it comes soon. always. If you ask questions, I'll answer them. That's what we do here. Yeah. So, the US already solved the issues. Well, that's good. That's good news. Hopefully, it comes back soon. Back in a sec, peeps. Heat. Heat. Heat. Heat. Testing out some local models. So, it might might just be a bit slow for a sec. Let's test it out. Open claw is personalized but technically weak whereas Hermes agent is less customizable. So you can actually get blank slate mode and then you can customize it a lot if you want. CL code lacks customizations but it's most you can you can customize cord a lot with the CLI if you have an agent operating system. That's what I usually do. Yeah, I would say out of the all of them, you know, Claude is the smoothest he is for sure. That's why we build the agent operating system of it. Nice. >> [music] [music] [music] >> Welcome. Heat. Heat. [music] [music] Heat. Heat. [music] [music] >> [music] [music] >> Heat. Heat. Heat. [music] Heat. Yeah, just fix I'm learning some testing. So [music] just close this when I do that. So we can fix it some stuff in the engine. [music] >> [music] >> Hey, hey, hey. [music] Heat. Heat. >> [music] [music] >> Heat. Heat. [music] >> [music] >> Look at that cool. [music] Yeah, don't worry. All good. We're back on. And it was Claude that did that. Savage for M for fusion. I've topped up open router. Now for Miniax M3, I've already got the coding plan. So you should be using the CLI. If you need me to log in, just use OF. Thank you, sir. When I say a wolf, it thinks I said I [music] wolf. What on earth? Benchmark leaderboard isn't sticky. You should be keeping the title sticky. Where you up to on the ponytail tests? Did you add this to the leaderboard? If the tests don't work, then we probably shouldn't do a guide on it. I can't see any proof of the actual test he ran. Like, there's no proof. You didn't add like the terminal stuff in there. Why not? Okay, we'll deploy it to the website. But actually no. When I scroll down, I can't see the model names on the leadboard. It's not sticky. What's up, Randall in the building? Good to see you here, sir. Always a pleasure. Why not embed what you actually built as well so people can see it? Because then we can check if the outputs are good before and after as well. How does it compare against headroom? You've already tested headroom. Why don't you add that into Ah, there we go. That's looking better. Where you up to minimax? Ah, I see. What's the difference between them? As I say, you'll probably want an RTX or Nvidia Spark for local models. They're the only two things I've seen where people [music] get good results with local. I see. Thank you. Where are the Kim K 2.7 tests? I can't see any of them. You should be using the CLI right? Where are the be rest of the Grock build tests? I can't see them in there. You should be using the CLI. Heat. Heat. Hey, hey, hey. to the Good morning. No, no, no. So, we've already covered this. If you go back to the start, you'll see it. Hey, stop. >> [clears throat] >> So today we're going to be answering some of the latest agent operating operating system questions to help you build your own agent operating system and get the most out of this stuff. Now, if you're not sure what an agent operating system is, this is basically where we've got all of our agents plugged in. So, for example, we have like Hermes Jarvis over here. We have a full studio for Hermes. We have, for example, a pipeline for going from idea to implementation. Um, we also have, for example, an AI agent mastermind group chat where we can have all of our agents working together. So, basically, you get everything in one place. It's easier to organize your agents. is easier to build custom stuff. So, for example, like this morning, we've already built a new research tool into our SEO content pipeline so that we can check any website, see what they're ranking for recently, and then also see if there's any new keyword opportunities we could potentially go for. And then once we create the content and automate it with this system, we can deploy it to our website as you can see right here. So, it's a super powerful way for getting the most out of your AI agents, organize them together. I think this is the future in terms of how things will be built. And inside the air profit boardroom, what I do every single day is answer the questions in a tutorial like this so that I can help you as much as I can cuz I know like for example, if people have these questions inside our community, probably everyone who watches this sort of stuff has the same sort of questions and we can all help learn and grow together. So, we're going to get straight into this. and let's start having a look. So, James asking about the agentic OS here. He said he's moved his primary agent and 25 workers from OpenClaw into Hermes. It was a lot of work, but they all seem to be online, which is pretty amazing. And he's using the camb board feature, which is this section over here, right? Right? And what this means you can like orchestrate teams of agents, multiple different profiles. You can have a swarm of agents working together and then you can build and automate whatever you want. So for example, if we have a look at this cam board, you can see that we automated a blog post and a video fully edited using this whole system and it works together super simple. So it's an easy way to like orchestrate your agents and get them working together. So, let's see what questions we got here. Here's a question. So, He's using an agent profile inside Jav. He pushes uh pushes everything through it and he sets everything up in KBAN, but it seems to like forget about stuff or get stuck in the blocker. So, what you can actually do is for example, we give this a new task here. Let's just test this out. So, we're going to say like create an SEO keyword research report for topics around agent OS and we'll plug that into the system. So, we'll add it in here. So, now that should get triaged and let's see how it performs. So you can see now the goal has been scoped. Everything's been scoped here. The research approach has been organized. Here's a deliverable format etc. And then it's actually found an actionable for our content writer to start creating this. So now it's been put into the running section here and we're good to go. Now if anything gets stuck, there's two ways to do this. Like you could drag the task over to the to-do list or you could actually comment on there as well. The other option that you have if something gets blocked is you can actually just speak to Hermes directly and say, "Hey, we set this up inside Kambban, but it's blocked. How do you fix it?" And then you can get Kamban with Hermes synced together and you can get Hermes to fix it. So, three options on that. Option one is you can comment on the post. Option two is you can actually move the Camban task And option number three is you can actually speak directly to Hermes and ask it why is the camb blocked and can you fix that so it never happens again. And then because Hermes is self-learning it should unblock itself. Option number four is you can actually speak to Claude and get it to fix all of that so it runs smoother in the future. That's another option. But we can now see that the keyword research report is now fully created and that is completed which is awesome. So if we go down here we can see what's completed and we can see the task and everything else. So it tends to work smoothly but you need a good system for this. But those are four options to fix that. Number two, also building his own version of the Agentic OS. How to get the agent OS to actually talk to Hermes and all of the agents are having it. All right. So, if you want your agents to work together, there's multiple different options for orchestration, right? So, if you want all your agents talking together, I would say there's three main options based on what you said. So option number one is you can use them inside the group chat in the agent mastermind section. So you can see over here for example we have this agent mastermind. We can drop in a message and then we have our agents working together inside the group chat. Option number two is you can have them working together inside the pipeline section so they can build and create together. And the third option is inside paperclip. You can get actually get your agents working together as a a team of agents. So you got fixes for both questions right there. This is a good question. So Jay is talking about Hermes workspace. This is actually one of the reasons I stopped using Hermes workspace because it doesn't seem to sync properly. So the reason that we actually built the Agentic OS is because I found like when you were using for example sometimes even if you were using paperclipip or Hermes workspace or these other open source projects quite often they didn't sync properly based on the new updates. So that's why I recommend using the agent OS instead. It's definitely saved me a lot of time. The other thing that I found is like sometimes you would update Hermes workspace and then the new version wouldn't sync properly to the old version. So for me personally, that's why I use an agent operating system instead. And I'd recommend for you to do the same. And we've got that inside the classroom here if you want to get the new version. So you can see when it was last updated, you get the video tutorial and the new zip file to use it. This is a good question. So Mike's asking about open router and fusion. So open router and fusion is like a new way of having a panel of models working together and then they create the outputs. you have like five different models that answer the same question. The judge critiques the answers from all of these different models and then that gets fused into one single answer. And you can see some of the stuff we built here like just for fun just to show you what's possible. And the interesting thing about this is like it's a it's a oneshot model because you don't go back and forth for the API. You just use it once. You build something, you get the answer, and that might take like 5 to 10 minutes each round. Here's another example what we built. So this is like an RPG game that we created using Fusion directly. It's pretty cool. Now for me personally, I only use it on the big stuff where you need an amazing answer and you need to basically guarantee like fable five level intelligence answers, right? Um, the biggest reason for that is like it takes about 5 to 10 minutes for each answer. You have to use the API and also you can't go back and forth with it like you would with a CLI. So, I think it's great for getting a final answer or for a really important decision, but I don't think it's very good for like coding dayto-day. If you actually want to see how it performs, we've got Goldie Bench over here. And it was one of the best performing models that we've used. So, let's open up some examples here. Like, it created some pretty cool stuff as you can see. And when I tested it on benchmarks, it's it tended to outperform other models using the same sort of prompts. So, if we have a look over here, here's something else that we built with this system. And you can see like the the graphics, the way this game plays, how smooth it is, etc. It looks really cool, but again, you would only use it for like big builds or maybe like if you wanted to build a new feature into something, you need the best answer and then you go back to to using Claude, for example. So, a question from France. He's actually building out an AI project as you can see. And he asks, you know, what are you running Hermes with? Is it your laptop? Is it a home server? Is it a VPS? For me personally, I run it locally just to keep it sandboxed away from everything. But there's a lot of people running it with a VPS. I've seen that as well. Now, sometimes people have like technical bugs with building this stuff out as you can imagine. Like it's quite um an interesting system. So, for example, like we've got this game studio here where you can build stuff and Ritz was was struggling to set it up. So, this might be based on the model. I can see for example, you're talking about using OAMA, but if Lama doesn't work for you properly, then you can just use a normal API or even a CLI inside that section. Now, if you want to fix that directly with the agent that set it up for you, you could ask them directly. Okay. Hey, can we change the model inside the game studio to for example claude CLA or to Grock build CLI? That's the way that I would approach it. So if you don't want to use local models or if the local models are not working for you, just switch the model to a CLI instead and speak to your agent that actually set up for you. Andre is asking about sorry Adam London was asking about what's the best way to enforce output paths for paperclip right so basically you know where to put the stuff that you've created here. So you can see all this stuff that we've built with paperclipip which is a way to orchestrate your agents as a team. So you can see for example we have a team of AI agents working together with Hermes and then these all run as a team to build out cool stuff like you can see now for this if you are having issues with the outputs if you use the MD skill MD file and I've spoken to the agent that actually built this for you any path you write in there is a suggestion not a rule. What you can actually do instead is you can change the working directory in its runtime settings and then point it at the workspace you want. So you could use relative paths inside your skills, but you would change the CWD working directory according to the agents. This is a good question from Douglas who's like looking for a custom build of the agent OS that you can give to clients. So the way that I would set this up is you have all your CLIs that your client uses on the left hand side and then you have the agent orchestration section as well. And then what I would do is speak to the client. I'd actually remove these. you know, these custom workflows over here and just simplify it so much that the client only has the automations and the stuff they actually want because really the the goal for the client is like to make it as simple and as easy as possible to use this agent OS without them finding, you know, loads of bugs or or getting distracted or whatever. So, for example, if you're looking for something that can track leads or for example the video content creation section, I would just build those in. The way that I would approach this person is if you're building an agent OS system for your clients, simplify as much as you can. Just remove everything that they don't need and only build in what they do need. So, for example, you could go to Claude with the zip file for this agent OS and be like, "Right, my client wants help with lead tracking and SEO. Can you build those features in?" And then you would test them yourself, make sure they actually work, demo it to the client, make sure they sign off in it, and then from there, you're good to go. The goal is to build something simple enough to be easy to use, but also match what they need on the customizations. Now Sheena's been building out her own version of the agent OS. She's a a member inside the AI profit boardroom and basically what she's added is two interesting things here. So two upgrades and basically what she's done is like taken the agent OS system that we have over here and then customize it exactly how we want which I think is a fantastic way to use this. So the first customization is adding comfy UI inside there too which means like for example you can generate uh videos locally and then the other option is adding in GLM 5.2. So you can actually build the file save it to the workspace and it's good to go. Let's have a look at the video example here. She's running a uh you know, she's got LTX set up there. So that can generate videos and then she's got this skill section here. So every time she creates something with GLM 5.2, she can use the skills inside the workspace to build cool stuff. So for example, you could have a custom skill section here where you use that for landing pages or for example for generating emails or you know creating any tools. And you can see an example here where it actually uses the skills which is pretty cool as well. So this is an example that's absolutely awesome for how you can customize the system. I really like the idea of like having skills for everything you build that are saved later so you can come back to it. It's kind of like a a selfarning system that improves every time you use it. Great idea. Tempted to build something like this myself. Also, what I like is like you're sharing the video demo of how it works, which is really inspiring. Also, if you ever want to jump on a if you're watching this, you want to jump on a call about this sort of stuff, we have a calendar inside the air profitable boardroom where you can jump on live calls, you can ask questions, you can meet the community. It's a great place to connect with people who are building simpler stuff. >> [clears throat] >> Another question we got here is from Ritz. So they were asking like what's the best model setup for the main paperclipip and agent OS operation router. So I would say there's two options for this. The two that I've seen that are the best are Hermes and Claude. So if you don't want to use Claude, then I would go with Hermes and a free API or an oorthth login. So for example, if you're already subscribed to Twitter, then you get Grock. You can plug that in as the main lead on your paperclip projects and that can also delegate tasks to local or cheaper models as well. But if you're already using GPT 5.5, I think that's actually pretty good. So you can see here Mark is already implementing the NA10 classes inside the community and he has three short-term goals. So number one is setting up Obsidian with the shared memory system. Number two is setting up Hermes agent and number three is setting up a dashboard and hierarchy for agents. So this is an awesome way to approach it. I would just focus on one thing at a time. So, for example, this week you can focus on setting up Obsidian and the shared memory system. It's pretty simple, actually. So, what you can do, if we have a look over here, this is our Obsidian Galaxy here. And we only built this like a couple of months ago, but it just gets better and better every single day because our agents are automatically plugging in what they create. You can see, for example, we have a new memory here from 2 hours ago. They plug in what they create inside the system and then our agents can update it and also organize it for us. And if you want to see what that looks like directly inside Obsidian, you've got it over here. So to get started with that, if you want to set up Obsidian ASAP, You can download Obsidian. Then you can go into whatever agent you're using. So for example, OpenClaw and say, "Hey, I've set up Obsidian locally. Here's the documentation for Obsidian in terms of how to use it. Every time I speak to you, please update my Obsidian vault or at least do it once per day." Then you can also tell openclaw use obsidian locally and you can give it the file path and use that as your main memory system so that when I'm asking you questions you draw from the obsidian memory and when I speak to you you also update the obsidian memory and that's a quick system that you could probably set up in like 20 or 30 minutes today and I think it would help you a lot plus make massive progress on your goals right there. And then from there, you can start working on setting up the first Hermes agent on the next steps. Now, Garfield was asking about notebookm and we've done some tutorials on it recently. Basically, where we plugged it into our agent OS system with an MCP. We've got our library of notebooks. Here we've got the research section where we can actually plug in a notebook like so and research it directly. We have the chat. We have the studio where we can generate podcast or videos or slide decks, whatever we want directly inside this section. And then we also have the assets. So this is stuff we've actually generated. So for example, if we look at this, this is an infographic created with notebook. We've got the full research report over here. And everything that we do inside NoBookm can be plugged into our agent OS system. So, we actually have a tutorial on how to set that up over here. And then if you want the agent OS setup, we've got it here with a video tutorial, the last update, and the zip file to use that system. So, it's all inside these sections as you can see. and I think that's it for all the questions. So, if you want to ask me questions like this and I create a video tutorial for you, then you can post your questions inside the AI profit boardroom community link in the comments description or go to the AI profitboard.com. If you want the full agent OS setup that we have as well, you can grab that inside the classroom too over here. And this community is all about learning, growing, scaling with AI automation. You might say, okay, some of this stuff sounds technical. So what I would say to that is number one, I'm not a programmer or a coder or developer at all. I don't understand HTML honestly as a language. But I can easily build out something like the agent OS system because we're at the point now with AI where you can build anything that you have an idea for and then Claude can just go off and implement it for you. If you actually look at Claude for example, Anthropic 80 I think 80% of their code now is generated by Claude. So even their best developers don't code directly. they use cord to help them. And I've also seen like for example, we've got 186 pages of wins, reviews, testimonials of the AI profit boarding from community members. You can see for example like Rick, he joined the school group and within 30 minutes he'd already set up the agent OS system. You can see Jose as well, he set up within one day. Like this is pretty easy to set up and we've tried to make it as simple and easy as possible to install an agent operating system. So, if you want to get my system, I mean, you can set up your own if you want, but if you want to get my system, you can get it all inside here. So, hope to see you inside the next one. Cheers for watching. Let's see what we got in the questions. Impressive Obsidian 3D node graph. Thank you very much, sir. It is a beast. When I look at this, I'm like, just looks fantastic. Thanks very much for joining. Appreciate that. So, there's rumors that either Sonic 5, Mythos 5.1, or Mythos 6 is coming out. So, let's see what happens next. A lot of people saying Claude is down right now. It seems to be working for me fine this morning. Yeah, I mean I I rarely see that sort of stuff happening anymore. Honestly, I think we're at the point now where models are a lot better than they were previously. Interesting take on Hermes Workspace. Don't get me wrong, like Hermes Workspace is pretty cool. So, you don't need to like completely delete it or anything like that. It's still got a lot of value. It's just that I found it was too buggy to use. And so that's why we switched. Anti-gravity seems to be okay, but obviously they've taken away Gemini CLI and swapped it for anti-gravity CLI. So anti-gravity CLI is still working for me. Yeah, me too. I love Obsidian. Fantastic tool. Yeah. So I mean the agent operating system, we have it running locally. You can plug in local models like we have a local engine over here as you can see and you can also use free APIs in it as well. So depends what you want to use and how you want to use it. Are you running loops? Yes. So we have two types of loops running. Let me show you the loop engine we've built. So we have two types. Number one is that we have the loop section for herbies directly. So you can go inside here, go to loop, and then you can see that you can set a goal, change what the name of the builder is. So change which model the builder uses and which model the judge uses. And basically that will loop around until the goal is met. And so what happens is you set the goal, the builder acts and drafts it, a judge grades it out of 100 and until it hits 90 out of 100 or more, it will keep looping round and then eventually we'll get done. So that's method number one. And then also inside there we have a camb board. So with the camb system, what we have here is a similar sort of setup where basically our team of agent profiles with Hermes go round and round until the content judge gives it 9 out of 10. So you can see here for example, this is the judge grading the work of the rest of the team and then we get a published blog post and we get a video. So there's two ways to run it. You can run the loop system over here or you could use this with KBAN. Nice. Sounds good. Yeah, same. My agents know pretty much everything about me, which is super useful. I think we have over 2,000 notes inside our Obsidian system. But it just grows and grows. You know, you can see it right here in terms of how it looks. All nicely colorcoded, all put together beautifully. That's CP useful. Heat. Heat. [music] Heat. Heat. >> [music] [music] [music] >> Hey, heat. Hey. [music] >> [music] >> Hey, hey hey. >> [music] >> Heat. Heat. >> [music] [music] >> Heat. Hey, Heat. Heat. Heat. [music] Heat. [music] Heat. Heat. Heat. [music] >> [music] [music] >> back in a sec. >> [music] [music] >> Heat. Hey Heat. >> [music] [music] [music] >> Heat. Heat. [music] Hey. Hey. Hey. >> [music] [music] >> Heat. Heat. baby. [music] Hey Heat. Heat. [music] [music] Hey, hey, hey. >> [music] [music] >> Heat. Heat. Heat. Heat. [music] >> [music] [music] >> Heat. Heat. >> [music] [music] >> Heat. Heat. N. >> [music] [music] >> Heat. Heat. [music] Heat. [music] Heat. >> [music] [music] >> Heat. Heat. >> [music] [music] >> Heat. Heat. [music] Heat. Heat. >> [music] [music] >> Heat. Heat. [music] Heat. [music] Hey, heat. Hey, heat. >> [music] >> Heat. Heat. [music] Heat. Heat. [music] >> [music] >> Heat. Heat. [music] Heat. [music] Hey Heat. Heat. Heat. [music] Heat. [music] Heat. [music] >> [music] >> Hey, hey, hey. >> [music] >> Heat. Heat. [music] >> [music] >> Heat. Hey, Heat. >> [music] >> GR I've not actually tested out. That's the one for Opal claw, right? The GitHub repo. So you can, but you could basically create your own like just add your own personality to it, I think. Going to test out Sakana. I don't know if anyone's heard of this. Looks fun though. Supposed to be like Fable 5 level. There's all these Fable 5 level benchmarks coming out now. It is indeed. neural index. What does that look like? I'd like to know more about that. Hey, hey hey. basically a file that lists every other file connects to it. Nice. Just setting up Sakana in the background. So, what's that sound in a sec? Well, we got test running on Sakano already. This is going to be fun. Hey, hey hey. Heat. Heat. Max Gio. Today I'm going to show you a powerful new keyword research tool that we've created for SEO. And basically what this can do is research keywords. Then from there we can actually deploy that content. And this is all based on our existing data. So I'm going to show you exactly how it works step by step. Headphones going crazy. Just going to switch them off. So let's take a look at this. Basically what we can do here is we can select one of the websites already own. Then we can have a look over the last 7 days, 28 days or 90 days from our Google search console. And from here we can start finding keywords. So let's say for example, we got the AR profit boardroom over here. So we can plug in the last 28 days and then we can give it an example. So let's say Hermes workspace or Hermes as the keyword that we're trying to filter through. And then you can see from here it gives us all the keywords related to Hermes that we're actually ranking for. And that just gives us ideas for SEO keyword research because the idea here is that number one, we can see all the data. So we can see where we're ranking for. Number two, we can see our average position for that particular keyword based on the website. We can actually see the post that is ranking to as you can see right here. And then we have all the data we need to figure out, okay, what keywords could we go for next? What are opportunities? And then from here we can use the topic. Now this is linked together with Google Workspace API which then has access to my Google search console across the websites that I choose and then we can select those and use this inside our agent operating system right where we've got everything together and this will feed in to the next step of the process which is deploying the content. So the idea is like if you're already ranking for a particular keyword then you're more likely to rank for the next one. And it's a powerful way to use AI and SEO together to come up with keywords. So you can use Google search console and your real data to then find relevant keywords. So if we have a look for example here, Hermes desktop app that could be another keyword that we go for. If we click on use topic, we can now plug that into our generate tool and we can type in the ideas here. Now we might not just want to go for Hermes desktop app. We can actually have a look on Google to suggest and find keywords around that topic. So if we type in for example Hermes desktop, Windows could be a good one, Mac, Linux, etc. And then we can use those to come up with new ideas for keywords. So for example, this is a good one to go for. That means desktop Mac because we're already ranking for it. We already have case studies on that particular topic and we can easily create more content that's new for that particular example. So then we can select the target keyword here and from here we can just make sure that we have a new case study to plug in based on that topic. So if we're talking about Hermes desktop map then we probably want to find a case study that we've created that we can plug into that system. So how do we do that? So I can find a case study like this. There we go. So, we've got this one that we can use and we can use that on our system as you can see. And the reason that we want to have a case study is like you don't just want generic, you know, AI fluff. You actually want to add some information gain on there. That's how you're going to be more likely to rank. And if you're wondering like does this approach actually work, you can see our rankings and our analytics over here. We actually got a free course with 260 SEO lessons inside the comments description if you want to learn more about this stuff. But basically what we can do from here is we can find keywords ranking for tweak them slightly to find something new, add in a case study and then we can click on generate five articles. Now I've also got this option inside the SEO tool which is save and reuse. So what that means essentially is if I save and reuse a case study, if we have similar keywords around that topic, we can use the same case study to come up with more unique content around that topic targeting new keywords. So we can save it inside our workspace, come back to whenever we want to. And also it trains your agents on like here's what we typically talk about. Here's some stuff that we worked on recently. Let's create more content and rank it around that topic. So we've put in the keyword here, we've put in the transcript. Um, then we're going to hit generate five articles. And what that will do is actually auto deploy after generating that, right? So, it's going to deploy a unique article to every single one of our websites that's beautifully formatted. We don't have to log in. We don't have to copy and paste anything. And we don't even have to index it because what's going to happen here as well is the through this workflow and this tool that we've created, it's also going to automatically index the content inside Google so it ranks quicker using indexional. So we have a API from index exceptional. This is a tool really powerful for just getting your content index quickly and then we can use that to start ranking faster for all of our content. The other cool thing about this is it actually helps you rank directly inside AI overviews as well. So not just Google. So if we take a keyword like this for example, you can see that the top recommendation for that particular keyword is based on our content and our case studies. And so we are the the top ranked option for this particular keyword. And you can see us ranking here and also over here. And so when you can create content across multiple different platforms that's all unique based on information gain based on useful content we can rank number one inside the images number two inside the AI overview number three we can rank with videos number four we can rank across multiple different uh positions on Google and we can also rank inside AI mode and Google AI overviews too and this is just a proven case study to show you like this stuff actually works like it's actually useful And I think most SEOs won't reveal what they're doing. They'll just tell you about it. But this way we actually show you what's working and how we're doing this. So it's pretty powerful stuff. And then you can see that's now deploying here. And that's the whole system. Now this was actually inspired by AR profit boarding member who suggested setting this up and they actually took our agent operating system and then tweaked it and customized to build this in. And I was like, "Wow, we need that, too." So, the way that I look at this is I call it the search console engine with the agent operating system. So, your Google search console already knows the keywords you're about to rank for. And this way, it can give you that data live and hand you the exact post to write next based on what Google already understands about your website and what's already working. Because typically, if you're using, for example, hrefs and you're using the keyword list on there, well, number one, it's not customized to you and your business. And number two, it's pretty hard to find good keywords on there. And number three, all your competitors access to that data as well. Whereas Google Search Console is unique to you. It's unique to your business. It's unique to what you're actually ranking for already. And so you get much more relevant keyword suggestions as you've just seen that are super niche, but actually already work. And so you go from Search Console, then the tool will read it live. It will score the opportunities. Then you'll get a ranked to-do list. And in one single click, you can actually generate the posts and get them deployed. Now, if we go to the history section here, you can see the posts are being written and this system is running right now. So, it's just running in the background. I don't need to do anything. I just click a button and we're good to go from there. Pretty powerful stuff. And so, if you look at this system, you know, we've got Google Search Console plugged in. It looks at all the queries, looks at the impressions, the clicks that we get, and then it comes up with opportunities. You see this opportunities bit here. So, it finds the relevant opportunities that we can then use to deploy to our website. So, if you look at my website on this example, it knows, okay, this website ranks for a lot of Hermes and AI community related topics. Let's go for those, too. And also what you can see here is the it will show you how many impressions you're getting for that topic. So you could either reoptimize the content or you can see this as a content gap and if you haven't created a post on it previously well now you can which means you're much more likely to rank. So it's really powerful system and it just helps you find the gaps in your content based on the impressions you're getting. And also bear in mind like if you were going into search console and you were trying to look for all the data there and the keyword research data, it's very hard for a human to look for all that data, especially because you're probably busy as as an SEO. You're probably adapting to all this crazy stuff that's going on in AI right now. And so if you use this system instead, you're going to save time. It's going to help you. By the way, if you want a free SEO strategy session and you want us to look at your website and find your keywords for you, you can book in a free SEO strategy session at goldie. Agency. You'll get a free SEO domination plan. Discover the secrets SEO link building. We'll answer any questions you have. You'll learn the best link building strategy for your website and you'll learn how to outrank your competitors. So, feel free to do that. Goldie. Agency. So, here's an example of what how it works. Um, pretty powerful stuff. You might also say, "Okay, well, these are AI keywords. You know, will it find keywords in my niche?" Because you're looking at my system here. Well, it's all customtailored to you. So, it reads your search console, not mine. So, it services whatever your audience actually searches for. So, whatever industry you're in, the patterns are universal. The keywords are yours. Uh mine happens to be AI tools, but yours could be whatever is your industry. And so, let's talk about the framework here. There's five moving parts that turn your raw Google Search Console data into the next post that you should actually write. And each one saves you time. You know, this would be someone's full-time job previously. Now, you can just automate it with AI. So, first of all, we got the live read. So, it reads your Google Search Console data, not a third party tools guess about it. So, your actual impressions, your actual positions, your actual clicks, straight from the source. And it's read only. So, it's not going to make any changes. is just going to look at your data and pull that in. And then it's going to look at four signals. So four patterns automatically that help you rank. So number one is almost ranking keywords. Number two is page one pages that nobody clicks. Number three is zero click gaps and number four is high volume terms. So it splits them into categories, organizes them for you. You stop hunting for keywords and it surfaces them automatically. And so you can see how those categories work over here. You have all your view queries. They plug into stuff with uh low click-through rates, stuff that's pretty close, but you just need to push it. Stuff that has a content gap, and then stuff with high impressions that's actually working and is worth like doing more content around that. And then they get scored and ranked with the biggest win first. And so you have one pile of queries tagged into four patterns scored and ranked by the clicks potentially on the table here. Now you might say okay I'm not technical you know I'm not an SEO is this going to be complicated and the answer is no because you just pick a site from a drop down and click one button that's the whole job so the four signals the scoring the page each keyword ranks on the engine does all of it for you if you can choose your website from a list you can run this so this is literally it I mean you can see we've already deployed one page here um looking pretty nice and that just worked in the background whilst I was talking to you right now so it looks pretty nicely designed etc. And this is how the whole system works. Pretty amazing. So if you look at my system for example, here's the biggest wins it found ranked by score. You see, for example, Hermes workspace, Hermes desktop, community platforms, Hermes web UI, etc. Um, these are all potential keywords that we could go for or reoptimize for or find variations of. And the cool thing about this is like before you'd be guessing what you can rank for. So you'd be using something like HS or Semrush, sorting by volume, hoping you can rank for it, competing with everyone else, using the exact same tool. Never looking at your own search console. That was a problem I had. The data really matters here. And then you know most people they're writing on a hunch waiting 3 months to find out you are actually wrong. And with the old way, before you did this whole system, the result was generic posts, generic rankings, and a lot of wasted effort. With this new system, you've got 30 seconds. It reads your real search console, your actual ranking pages, tags the four patterns automatically, and categorizes them, ranks every opportunity by clicks, potentially, shows the page each keyword already ranks from, and one tap drops a winner into an article writer. So, so the result is you write what you can already almost rank for and then you win faster with SEO. That's the difference here. By the way, this is all built into our agent OS system. You could build it yourself if you want, but if you want the research system that we've created, the search console engine is one tab inside a full agent operating system that I built. So, it connects Claude, OpenClaw, Hermes into one dashboard with shared memory. So, your agents already know your business, your sites, and your goals. And you get the full agent operating system with the SEO research, the article writer. You can deploy it all to your websites in one place. The 30-day road map, four coaching calls a week, and daily tutorials. And so, this is basically how it works, right? Um, and this can just keep repeating. It's never going to run dry because if you think about the data that you get, it refreshes every single day. Which means the list refills itself which means that every post you create ship and changes your data which services the next opportunity. So the engine just keeps handling you the next move. This how it works, right? So you write, it ranks higher, it reads your data, you get more scores, you get more keywords, it just keeps going round and round and round. Every loop refills the list. It compounds and it improves as you get more and more traffic. Now, some people are going to say, "My site's too small for this to matter." Small sites often have the most lowhanging wins because a lot of your pages are stuck on page two. You just need a lift. Other people say, "Well, I need an expensive SEO tool to know what to write." The one source that knows what you can rank for is free and it's already yours, which is your search console. So, this system just reads it for you. [snorts] And then other people say, "Well, I'll get to keyword research eventually, but right now those impressions are happening." And every month that you don't fix this, you know, you're kind of like leaking potential rankings that you could get. So, if you think about the system, you stopped guessing. You stopped using uh keyword tools that are generic. You stopped hunting for keywords because you save time. You get the keyword research automatically based on your data. You always do the biggest win first because that's how it's categorized. You write in one move with one click as you saw today and you never run out of keyword opportunities because it's always finding new stuff. So this is how the whole system works. And if we have a look now inside the agent OS, we've got the research section here, the generate, the deploy, and the history section, right? And this is all deployed right here. So it's pretty powerful stuff. And this is actually based on a skill that we created, which is like many different steps to create the content itself and make sure that it's ranking. So, what we've actually got plugged into the system is a skill based on what's helping me rank right now. And we've customized it so that your AI agents can take that skill, customize it to your business, and then start implementing it, right? And they can post across multiple different sites at the same time with unique content across everywhere. You could use this for social media as well. And it's a pretty powerful system and it's all inside the AI profit boardroom. So if you go to the AI profit boardroom here, go to the classroom, it's inside the agent OS system which you can get. We update this daily with new updates. You get the zip file, the full guide on how to use it, the video tutorial. And then we also add new updates like you can see right here. And inside the classroom too, we have a full AI SEO automation section with all of my best trainings like you can see inside the community. You can ask questions as well, get help and support. I personally answer every single one of these questions every day with video tutorials. Inside the calendar, you can jump on 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 this stuff. And this is all inside the AI profit boardroom link in the comments description or go to the profitable.com. Nice. [laughter] That's pretty cool. Nice. RTX. I think everyone's moving towards RTXs now, right? Because they want local stuff. Yeah, you can do that, too. Thank you. Yes, it is. Let's see where we're up to on Sakana. Kind of looks similar to Fusion, doesn't it? Multi- aent system, one model, you use it on the API, then you got all these models working together, multi-level intelligence. You got Fugo and Fugu Ultra. Let's have a look how it performs versus Fable five. Pretty insane. Let's see where it scores on Goldie Bench. That's the question. Japanese AI as well. You don't see many Japanese AI models come out. It's mostly China, isn't it? Mostly China and US and Lehaton Fat. Oh, wow. He's pretty good though. Pretty nice. Wow. This is good. Big fan of Japan to be fair. Been twice this year. If you got any questions, by the way, like feel free to ask. Just building some more stuff with Sakana. So, we have a brand new update from a Japanese AI lab called Sakana, and they've created something called Sakana Fugu. I probably totally mispronounced that, but this is designed to be a full multi- aent orchestration system accessible via a single model API. And this has another model inside it called Fugu Ultra that matches the performance of Fable Mythos apparently delivering Frontier capability as you can see right here. Now, basically what this does is it it's quite similar to Fusion. If you've seen Fusion, what it should have is basically a mix of models working together and then they create an answer, right? So instead of just having like one model like Fable 5, you would ask a question and then this has a panel which is a multi- aent panel API that competes headon and then it synthesizes the answer and you get one answer with I mean it literally just dropped an hour ago. So go easy on me, guys. But what I will say is we've already tested out on three different things. And I would always say test out yourself. I'll explain more on that in a second. So we can see an example of a website we built here. Actually looks super nice. I'll show you how this compares on Goldie Bench with everything else that we've created recently. But the website itself looks super nice. So that was one example. We also have, for example, this maze game. As you can see right here, that turned out super nice. And again, I'll show you how this compares with other models because we always use the same tests and the same prompts with every single model, but it's pretty powerful stuff and actually looks really good. And then we've got this one as well, which is pretty mind-blowing. This is like a, you know, a simulation of the galaxy. And the quality of the outputs here is is super nice. Super nice stuff. So, you know, if you're looking for a Fable 5 level model or you're looking for Fable 5 level outputs, this could be it. But again, I would test out yourself. I've already tested out myself as you can see here. And I get the idea of this because we've actually used something else similar recently and it's called Fusion. Right now, if you've never used Fusion before, it's a similar sort of idea. You have multiple models as a panel and then they go for a judge which fuses it and then you get one answer out. Now, if you're wondering how does Sakana perform on benchmarks, let's have a look at this. Let's pull this up right here. So, we got terminal bench and you can see Fable 5 scored 80.4 versus 80.2 and 82.1 versus Fugu. And on all the benchmarks here is pretty much outperforming Fable 5. S SW Bench Pro Fable 5 destroys both Fugu and Fugu Ultra, but on most of the benchmarks I have a pretty even or it's being outperform. So you can see for example Live Code Bench here, Fugu Ultra 93.2, 92.9, and 89.8. So pretty impressive model. Pretty powerful stuff. Really interesting idea. We've already tested it like I've said before. Now if you're wondering, okay, how does this compare against everything else? Let's have a look head on. So let's for example compare this versus GLM 5.2 and we can see the same tasks side by side. So this is GLM 5.2. This is Fugo Ultra. Let's have a look here. So you've already seen this demo and then let's have a look at the version from GLM 5.2 which is still a little bit buggy as you can see right here when you're comparing it on benchmarks. um and pretty difficult to navigate and move around, right? Um now if we have a look at the next example here. So this is the website. So GM 5.2 versus Fugu Fugu looks super nice as you can see. Nice animations, nice colors, nice UI, etc. And then if we compare that versus GLM 5.2's output, which still looks nice, it's just not quite got that same touch. it just doesn't look quite as nice. So, side by side on the outputs here, I would say that Fugu is winning on the benchmarks. Again, test this stuff yourself. See what you think. Let's have a look at the galaxy example here. So, this is the living galaxy, the living spiral galaxy where we can zoom in, we can zoom out, we can move this around, etc. And then if we compare that versus GLM 5.2, looks totally different, right? Totally different. I would say which one is is more interesting. Which one is more beautiful? I would go with this version right here. It's just way more interesting to use on a deeper level and the outputs are pretty amazing, pretty inspiring stuff. I've also got more tests running in the background. So, we're testing this out. Let's compare it versus Opus 4.8 as well. So this is Opus 4.8's output and this is Figus. Like which one looks a lot more interesting, a lot more powerful? For sure it's this one, right? More interesting, better design. I will say just even Fable 5 was not that great at UI, but if you compare them side by side, this one looks a lot nicer. Even like the way you can move the mouse and it has these animations over the boxes compared to this. Super boring. Now, one thing you need to be careful of, and we saw this with Lashhaton FAT that came out earlier this week. Be aware of like, you know, companies scoring their own benchmarks as we saw in this example. This was kind of like a hoax that went viral and a lot of a lot of smart people were tricked by it, right? by Laton fat scoring and breaking the benchmarks versus everything else. So again, this is why I test it. I've created the Goldie bench. This is why I'd recommend you test it out first to see what you think. But in the sideby-side tests that I've done, it looks great. It's created some nice stuff. Now, if you actually want to use it as an API, what we've actually done over here is build it into our agent operating system. And this is a great thing about having an agent operating system like Sakana just dropped an hour ago. We can already build it into the agent OS and then we can use it when we need this stuff. So for example, we've got fusion as well which is the other alternative to this and runs on a similar sort of API standard and what you can see for example is like if you know Fable 5 gets taken down no problem you remove it from the system. If for example Sakana comes out within 1 hour we've already built it into our agent operating system and we've used it we've tested it we can just ask it a question here and then we'll get the the high quality outputs that we were showing you a second ago. So that's the great thing about having a system like this is it's just so flexible and it can change whenever you want. And if you're wondering how to use Sakana, so you can sign up at sakana.ai and then also you would get the API from there. Now there's two different APIs. So there's Fugu Ultra and Fugu two different APIs and obviously Fugu Ultra is the more premium version but obviously a more expensive API as well to use. They also released a technical support so you can find more details on the research and what they did and how it all works etc. But yeah, you can see that it stands shoulderto-shoulder with Fable and Mythos. And I've seen that on our own benchmarks as well. So, it's [snorts] pretty interesting to see and I think this is the future really is like using multiple models together to get the best outputs when you need this. I also like the fact that it uses closed and open models together to test it out and then what it does is it actually manages like the model selection, the delegation and everything for you automatically. So, you don't need to sign up to the individual models. What you do is you just get the API and it handles everything from there. Now, you might be wondering, okay, like when should you use each? So, Fugu, the basic version is low latency. So, it balances strong performance with low latency, which means like you can get answers quicker and faster. And it fits naturally into tools like codecs for coding. And you could even use this for example inside like, you know, a customer face and stuff, which is pretty cool. That's one of the problems with Fusion. When you use Fusion, it's super slow. You know, you might be waiting 5 to 10 minutes for an answer. Whereas with this, you can build out quickly. That's how we built those outputs I showed you a second ago and tested it within 60 minutes and built it into our system. Now, when it comes to Fugal Ultra, this is the flagship model and it's tuned for maximum answers uh maximum answer quality on hard multi-step problems. So, that's really designed for like super deep stuff like AI research or all sorts of interesting deeper tasks that require more complexity. Here's another one that we just had come through. So, this is an orbit task, which is pretty cool. Looks super nice. Like, you can change the time scale here as well. So, you can speed up, slow it down, and it has a simulated date as well, which is pretty crazy. And then you can calculate the days per second as you go around inside the solar system, which is absolutely wild. Now, if you compare this to Claude's output, so this is Opus 4.8, Okay. Which one do you personally prefer? For me, I think the the Fugus looks a lot nicer, a lot more interesting. Now you also might wonder okay how does it perform versus fusion side by side. So let's see and have a look at that. So if you look at the response from Fusion again I would say this is one of the nicest websites that was created in all of our benchmark tests. But I would genuinely say that again, Fugu looks a lot nicer in the way that it's created. Like if you look at the animations, the way that the UI is designed, it looks super nice. It's clean. It's interesting. I would prefer the first website. It's marginal in terms of the output difference, but those are the differences. The other cool thing I like about building it in the agent operating system is like we can save stuff inside our workspace. So everything that we create as it comes through gets plugged into this system as well so that we can check it and come back to it later. We don't lose anything. You know, I think especially if you're using more powerful APIs, like you definitely want to save the outputs and see, okay, what did we create earlier? And then also, you know, as AI models come out and new ones come out every single day pretty much. Then you can see the progress of what you built and how it works, etc., which is super fun. Now also what's interesting here is the cost per build. So fusion itself from open router is a lot more expensive. So if you can land the same prompts is 25% of the cost which is pretty wild. Now, if you're wondering, okay, how does it compare against Fusion? Well, the endpoint works pretty nicely. The panel is actually denser, which is interesting. Bear in mind they're both oneshot. So for example, if you're using Fusion for tests or Fugu, like you wait for the panels to reply to you. So you kind of have this weight between them which means you just use them in one shot like you wouldn't go back and forth for them like a claw CLI or something like that. And also what's really good is like fusion on open router you pay per usage, right? It's API only. Sakana actually has like a a flat rate plan. So if you're doing like high volume agent loops, you probably want the subscription and then you'd go for Sakana instead. So, thanks so much for watching. That is the whole setup from Sakana and Fusion. Seems pretty fun to use, creates some awesome stuff, and we've already built it into our Aentic OS system, which we'll release an update on for today. We actually update this daily with new updates. As you can see, we've got Fusion built in there, too, if you want to compare them side by side. And this is a agent operating system where you can plug in all your agents together. If something new comes out like a new model, you can plug it in. We've got Claude, OpenClaw, Hermes all plugged in together. We have Hermes Jarvis, which is a voice activated version of it that we can talk to and also build with in real time. We have Kimmy code ready to go whenever we need to. And everything that we learn and build and automate, we plug into the system in Goon like a a video agent, an SEO agent, a music agent, a camb board, whatever you need. So, if you want to get that, feel free to get it inside the AR profit boredom community link in the comments description or go to the arprofitborn.com. Inside the community, you can ask questions, get help and support whenever you need to. And I personally answer these questions with video tutorials every single day. Inside the classroom, you can actually get all of our best training. So, we have a complete beginner to expert course here. And we have new daily updates inside this section, including the agent OS system with Sakana. Then, it's updated daily. You get the zip file and a full guide on how to use it. Plus, anything new and interesting that comes out, we actually add a new tutorial for it with a video and a full guide step by step. Right here inside the calendar, you can jump on 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. And this is all inside the AI profitable boardroom. Feel free to get it. Link in the comments description or go to the profitwarm.com. Thanks for watching. All right, that's pretty much it for me. Cheers everyone for joining. I'll see you on the next one. Cheers. is papa.