---
title: 'Testing Tencent HY3 and Building an Agent Operating System'
source: 'https://youtube.com/watch?v=T6PfRhyUWo0'
video_id: 'T6PfRhyUWo0'
date: 2026-07-14
duration_sec: 0
---

# Testing Tencent HY3 and Building an Agent Operating System

> Source: [Testing Tencent HY3 and Building an Agent Operating System](https://youtube.com/watch?v=T6PfRhyUWo0)

## Summary

This video provides a comprehensive walkthrough of building and using an Agent Operating System (Agent OS) that integrates multiple AI agents like Claude, Hermes, and open-source models. The creator demonstrates how to set up free coding tools like Omniroot and Tencent's HY3 model, automate SEO content creation, and manage workflows from a single dashboard. The focus is on practical, non-technical implementation for AI automation.

### Key Points

- **Introduction to Agent OS** [00:00] — The Agent OS is a system where all agents (Claude, Hermes, OpenClaw) work together from one place, with shared memory and one-click workflows, making automation more powerful and easier than using individual tools.
- **Testing Tencent HY3 Model** [01:30] — Tencent HY3 is a new open-source Chinese model claimed to be the best in China. It is free for a limited time via OpenRouter and News Portal, and can be used inside Hermes Agent for agentic tasks.
- **Omniroot: Free Coding Forever** [03:00] — Omniroot is a free open-source project that routes requests across 237 providers, compresses tokens with RTK and Caveman, and never goes down. It can be plugged into Hermes Agent for free coding.
- **SEO Content Automation with Indexional** [05:00] — Indexional is a tool for quickly indexing content on Google. By integrating its API into the Agent OS, users can automatically create and index SEO content, achieving first-page rankings within days.
- **Agent OS Features Overview** [08:00] — The Agent OS includes a command deck, model layer (free/premium), agent fleet (9 coding agents), content factory, memory system (Obsidian-based), and token minimization tools (RTK, Caveman, Ponytail).
- **HY3 vs GLM 5.2 Comparison** [12:00] — In side-by-side tests, GLM 5.2 outperformed HY3 in 6 out of 7 tasks, offering smoother gameplay and better graphics. HY3 is decent for agentic tasks but not frontier-level coding.
- **Setting Up HY3 with Hermes** [15:00] — HY3 can be used with Hermes Agent via News Portal or OpenRouter. It supports tool use, skill learning (forward/arn), and can build landing pages in about 60 seconds.
- **Community and Support** [18:00] — The AI Profit Boardroom community offers daily video tutorials, weekly coaching calls, and a map to connect with local AI builders. The Agent OS is updated daily with new features.

### Conclusion

The Agent OS is a powerful, non-technical system for integrating multiple AI agents and automating workflows. While GLM 5.2 remains the best open-source model, HY3 offers a free alternative for agentic tasks, and tools like Omniroot enable free coding forever.

## Transcript

Hey, hey hey. Okay. Deep. Heat. Heat. Heat. Heat. Hey, Randall. Good to see you. Yeah, I'm doing great, thanks. How about you? Thank you. Yeah. Today we're going to be testing out HY3 with it, which is 10 cents new model. They claim to be the best in China. So, it'll be interesting to see how this one goes. out. Hey Hey, hey hey. Heat. Heat. Heat. Hey, Heat. What's up? It's okay. Hey, look. Come here. Hey, this is Oops. Heat. Heat. Heat. Heat. Okay. Thank you. Heat. Heat. Heat. Heat. Heat. Heat. Goodbye. Who we got here? Chad, welcome here, sir. M here. What are we learning today? So, today we're setting up uh 10 cent HY3. They're claiming to be the best Chinese AI around. We're doing some more work on the agent OS. Also going to be looking at how to use Hermes for free with Omniroot and testing out 10centent HY3 with news research and doing a Q&A as well. Just setting all up. Maple 5 is working. Running some tests here as you can see. Heat. Heat. Heat. Heat. N. Heat. Heat. N. Yeah, Tencent Hy3 is is free as well. So, it's going to be interesting to test out. Uh, so you can use it for free inside Hermes agent I think for a week or two. So that's the way you can get access to it for free. That's if says wood. There we go. Living the dream, my friend. Heat. Heat. N. Heat. Heat. Hey, hey hey. I'm testing out now. So, let's see. Uh, we'll be updating Goldie Bench based on what we see today and how it performs. So, we'll be testing them out side by side. Um, I'm interested to know. Honestly, I don't expect 10 cent to be better than Claude, but you know, if it is, wow, that would be a big big improvement. Gang, I just love your overall journey from a non-technical expert to one of the best influences on AI. Thank you very much. I appreciate that. That was super kind. What hardware do you suggest to run models locally? I would go over DGX Spark or an RTX uh 5090. Which model could we compare to be as useful as Claude? Probably the closest I've seen is GLM5.2. That was the first time I'd seen any model compared to to Claude Opus in any part way, you know. So, uh, yeah, I'd go with GLM 5.2. It's pretty cheap as well. You can get it on the coding plan. If you look at my tests on Goldiebench, you'll see how they perform. We're just getting 10 cent HY3 now. It's not looking too bad to be fair. We'll see what we get in a sec. So, I'm comparing it, but less than GLM 5.2. I would probably expect that. Yeah, it's kind of like your Opus 4.8, 8. Then you have GLM 5.2. Then you've got something like Kimmy K2.7 or HY3, but it'll be interesting to see what we get. I have a Max Studio coming in October. Welcome to the gang. Me, too. I'm already on a Max Studio. To be fair, one of the best things about it is uh like for example, last week I went down to Phuket in Thailand, took the Max Studio, and it was great. really good just for traveling and getting stuff done and having your full setup with you. Absolutely loved it. So, I think you'll like the Mac Studio, too. I also think it's just so easy. It's like I can come back to my setup anytime. Uh it's all ready to go. Don't need to worry about charging up a laptop or anything like that. Really, really convenient. So, yeah, Max Studio in many ways is is just a fantastic setup. Back in a sec, peeps. Hey hey hey. Heat. Heat. Heat. Heat. Heat. Heat. N. All right. Ever said, Let's Let's have a look here. Oh. Are you okay? Heat. Heat. Heat. Heat. summarize in plain terms what is going on here. So we have um claude coding now for us and the things that we're working on right now. We're going to be testing out Tencent HY3, which is a new open-source Chinese model versus GLM 5.2. We will also be testing out the new Clawude agent operating system. We'll be running through a new case study for AI SEO, how to run Hermes for free with Omniroot. We're currently testing out HY3 with Tencent now. And then we'll also be looking at how to set up 10 cent HY3 with news research. And finally, we'll be doing an agent OS Q&A. Yo yo Heat. Hey. Hey. Hey. Heat. Heat. Not sure on that, but I know there's Ernie's bad that runs locally. Heat. Hey, Heat. Hey, baby. Heat. Heat. Heat. Heat. What do you mean by agent OS? Agent OS is an agent operating system. So we have this system here where we can have all our agents working together. So for example, we have Claude, Open Claw, Hermes all inside one system. Every workflow one click away. And that means that it's way easier to use and way more powerful than just using for example like Claude code or having for example Hermes agent inside the terminal. And so we can pretty much automate anything with this agent operating system which takes all your AI setup to a whole new level. So that's what we mean by agent OS. Heat. Heat. browser only. Yeah, we keep it in browser just so then again it's easier to navigate between things in one click. So if we're working on a guide over here, we can go back to the agent OS over here and everything's efficient inside one place. Whereas if you have to open up an app every time, it gets kind of annoying to manage. Especially if you got like for example CL desktop or other stuff open in the background. Heat. Heat. How do you connect your agents in the agent OS? So when you get the setup from the air profit boardroom, it has an installation zip file and then you just gives that to your agent. It connects all your agents. So that's how you do it. Hey. Hey. Hey. Hey, Hey hey hey. Hey everyone. Hey, didn't Heat. Heat. Heat. Heat. N. Thanks very much, James. I appreciate that. You kind of Good job. Heat. Heat. out. Hey hey hey hey hey hey hey hey hey hey hey. Heat. Heat. is what free what's it what are you talking about there Heat. Hey, heat. Hey, heat. out. Happy. Heat. Heat. Heat. Heat. Perfect. Hey, Have you found for automation Mac is better? Um, it depends. Depends what you're trying to do really. For me, I find like I either is fine cuz really we just use a web app for most automations now with the agent OS. So either way, it doesn't really matter unless you're running local models. And if you're running local models, I wouldn't use a Mac. I'd probably at this point use uh a DGX, something like that. If you prefer local models, I think that's a smarter way to do it. So today we're going to be answering some of the latest questions on the agent operating system. How to build them, how to get the most out of them. We've already built in HY3 that just dropped. This is a new opensource model from China, Hunen 3 from 10 cent and that's creating some pretty nice stuff. So we built that in. We also got some new features for example like we've got omniroot which allows you to code uh for free open source project pretty amazing roots across like 90 different models we have our memory system built in here some people are like what is an agent operating system so it's a place where you can plug all of your agents in one place have everything one click away and the great thing about this is that you can build custom workflows so for example we have a voice agent with Hermes Hermes Apollo over here we have Hermes Oracle we have Hermes Astros And every single powerful use case for AI automation that I personally have, I can just build as a workflow here. We even, for example, have this email outreach tool that can generate leads for us on the spot and can easily send the emails and manage your inbox to for us using Hermes agent. And these are all custom workflows we've designed from scratch based on what we actually think would be useful here and how to get the most out of this stuff. So, what we're going to talk about today is like the best ways to use it, how to troubleshoot some of the issues that some members have with this stuff, and also how to get the most out of this. All right, so we're going to get straight into some of the latest questions we've got here. So the first question that we got here was from M Jordan and I don't know if that's any relation but basically we've got a question here which is has anyone successfully found an agent that can autom automatically edit product listings from Shopify and successfully publish them into Amazon for example. So I think if you want to publish directly to Amazon, I I don't use Amazon that much, but I have found an API for sellers. So this might be useful for sellers and you can just give the API to your agent and then set up a tool to auto post for you. And then for using agent workflows with Shopify, I think you could quite easily set that up using their API. They've actually got a CLI for Shopify so that you can edit postings directly. So, Shopify have tried to make it really easy to use AI agents with their platform and that's why they've got the CLI. And then for editing and posting to Amazon, you could use something like the um Amazon API that we posted above. That's the way that I would do it. So, I would give both of those APIs to your agents, describe exactly what you want, and then you could actually even build that into a custom workflow inside aentic operating systems. Another question from um Jordan was, "What AI agent would you use to convert a Shopify store into WordPress?" So the way that I would do this is I would actually use the Shopify CLI, get the code from Shopify, and then with a combination of either claude or Shopify CLI, you could ask it to rewrite it and export it to your WordPress import. And then what you can do is take the code that you previously had on Shopify, use the CLI to convert that into a WordPress website, and then import it into WordPress. If you're not sure how to do that, it's okay because your agents can handle the 8020 of it. So you either want to use Shopify CLA or Claude. So JW was talking about Fable 5. So Fable 5 is about to expire from the subscription. So that means that in the future if you want to use Fable 5 with Claude, you'll have to pay. And so right now it's free on the subscription. In the future you have to pay. And JW says, "Everyone is scrambling to use Fable 5 before it expires tomorrow. Are you going to help create different prompts or setups for Asian OS that you'll share at a later date with us we can use with a mixture of agents to have similar performance?" So we've already set up fusion sana fugu and mixture of agents inside the agent OS system so that you can get similar outputs in a simplified way using the tools inside there. So for example, if you go to mixture of agents here, you'll see that you'll have access to Hermes mixture of agents inside there which can achieve similar levels to something like Fable 5. Then we also have for example Fusion over here. So we can use panels as well and we can also use Sakana Fugu. And those are three different options for getting like Fable 5 level benchmarks. you know, based on the benchmarks they've done, you can achieve pretty good outputs with this stuff. Now, when it comes to prompting with the prompts, there's no real need for special prompts because most of these models are at a point where they don't need special prompts. They just need you to describe what's on your mind. So for example, the way that I prompt Fable 5 or the way that I prompt mixture of agents is not really any different from any other model out there because what's different is the actual outputs and the quality of intelligence you get from these models. So you don't need to do anything special with the prompting. You just need the right systems which we've already built into the agent OS. There are some skills that will make your agents better, but you can't really use skills with panel models or panel systems like mixture of agents because it's just a oneshot prompt. It's not like anything special. This is interesting. So, Daniel was asking about like Whisper Flow style apps and a few recommendations were for Handy on GitHub. This isn't something I've tested out yet, but it looks really good. You can basically uh use AI so that you don't need to type. You just explain what you want and it types out for you. And it's a free open source project. Now, I've not test it out, but actually that looks really good. I currently use Whisper Flow, so this might be a good upgrade. Jose was asking about how to use Hermes for creating like carousels and that sort of thing. Now, this is something I don't actually automate with Hermes, but I actually saw this skill that looks interesting called open carousel. And that looks like it can create Instagram style content for you using Claude, but I think you could also give it to Hermes as well as a skill. Looks like it's currently designed for Claude, but I think you could probably give it to Hermes as well. The thing that I would do before you install it is look through the skill MD file, check it's legit, run it through Claude, make sure it's legit. Don't install anything that you're not comfortable with, but as a starting point, that looks pretty interesting. Canyu was asking about Omniroot. So, how to use Omnirroot inside an agent operating system. Now, if you're not sure what Omniroot actually is, basically it's a system where you can code for free and it runs through like 90 free providers, which means that it doesn't run out of tokens. You can continue coding with it. It's pretty simple and easy to use, etc. And we can just ask it to build stuff and actually build stuff over here. And we build that into the HON OS. Now, it's a free open source project. Now, you might be wondering, okay, how do you plug that into the HOS? So when you get the zip file from us, it's inside the install folder in section number 28. Or you can go directly to the GitHub which you can see over here and then install it directly to Benjamin was asking about what are the best ways to use Fable 5 right now. So obviously it's only available on the subscription for a couple hours. What are the best ways to use it for us? So, we've actually been building our HY3, Tensson HY3 as a new coder inside the agent OS and then we actually set up new benchmark scores on Goldiebench as well to test out. From what I'm seeing so far, it's okay. It's not like as good as GM 5.2 as you can see right here. Like even when it's coding stuff out, it still feels super buggy. But it's worth testing. I think if you had a choice between open source models now, I would still go with GLM 5.2. Got a question here from Daniel. And Daniel was saying, you know, he wants to get his first client. What are the best ways to do this? How can you make it happen? So for me when I'm trying to get clients I try to look at the inputs to get the outputs. Okay. So for example when it comes to cause and effect what actions can I take today to get more clients and I try and look at the numbers as a formula. So for example that could be how many videos can I create today that would attract clients. How many emails do I need to send to get my first client? how many calls or outbound messages do I need to send? And I would just keep working on that formula and iterating it until you finally get the the number of actions required to get the result you want. So, for example, when I was getting my first clients, I used to use Upwork and we used to send about 10 to 20 proposals per day to get one new client every single day. And so, once we figured that out, we just quickly stuck to that formula and gradually grew it until we could expand the number of clients we got as well. And that's what I'd recommend for you is just look at what actions are you taking today and how many of them are required to get the number of clients you need. We also had some questions about Hermes Asteros. So Hermes Astros, if you're not familiar with it, is a custom workflow we built here, as you can see. Basically, what it can do using the Google Workspace API is it can monitor certain keywords based on your competitors and then recommend ideas for content based on that. And you can customize this to whatever industry or workflow you're trying to do this for. So, you can see some examples right here and it basically monitors certain keywords, gives us ideas of the angles and unique titles we could create around this content. And then from there, we can plug that into our video agent, notebook or SEO content as well. So, it's a pretty powerful system. I've not seen anyone else build this, but it is really, really useful for just figuring out how to reach the right types of topics in a way where you don't have to go off and do the research yourself. Definitely worth using. So, it's like a automated competitor research system to quickly find new content ideas for you. Josia was asking for inspiration when it comes to Fable 5. What are we doing? So for me personally, I was just building out new workflows inside the agent operating system. I think the best place to start if you need inspiration is to look at Goldiebench. You can check out some of the creations we've made there and get some ideas on what we can do. So, Goldiebench basically analyzes all the latest models and we've got Fable 5 on there and it's built some pretty insane stuff, some pretty amazing things right here. So, if you need some inspiration and then also I would just look at what are you spending your time on right now and then how can you automate that with Fable 5 and even if you build yourself one single tool, it's well worth it. So, we got a question from Aaron about Nvidia's Nemo Claw, which I haven't heard many people talk about actually for a long time. So, I remember when that was all the hype and it feels like a long time ago. It was probably only about 2 or 3 months ago. So, when it comes to Nvidia Nemo Claw, yes, you need Nvidia to run it. As far as I'm aware, you wouldn't run it on Apple. I think from what I've seen inside the community, so Aaron's question is, has anyone incorporated this into their AI setup and is it worth it? I haven't seen anyone using Nvidia's Nemo claw, but what I have seen people use is hosting as a VPS, which runs in a similar way. And then you can sandbox your VPS with Hermes or OpenClaw, and that works pretty nicely. People have as actually set up uh VPS with the agent OS as well to sandbox it. And there's some really good tutorials which I'll link to. All right, so that's pretty much it when it comes to agent operating system questions. The thing that I would say for you if you're watching this is like definitely set this up cuz it will change everything for you. It's such a powerful workflow. It's so easy to set up. I mean, you might think, okay, that's great for you, Julian. You do agent operating systems all day. You've built them out, etc. But actually, I'm not technical. I don't code. Don't understand HTML. But I can build a full system like this. And I think learning HTML at this point is useful, but you definitely don't need to do it. So, you can see here, for example, so many people setting this up and getting awesome results of it. And so, if you if you haven't set this up, you don't need to be technical. It doesn't use up new tokens. You can actually have free models like omniroot plugged into this system and it's pretty easy to set up. So if you want to join us inside the profitable and get my setup for the air profit uh for the agent OS, you can get that link in the comments description or just go to the AI profitab.com. This is my AI automation community that helps you save time, grow and learn with AI automation inside the community. You can ask questions and I personally answer them every day with a video tutorial like you've seen today. And so I go through every single question and answer them myself. And then the community is really helpful because we all help each other and there's always people online to help each other whenever we want. Inside the calendar you can jump a weekly coaching calls, get help and support in real time, ask questions, etc. Inside the classroom, you can get all of my best trainings on this stuff. So if you're complete beginner, you can go from beginner to expert with this section. And if you want to get our agent OS, it's over here. We update it daily and you can see when it was last updated. you get a full guide on it plus the latest resources and also every time something useful comes out. For example, Omniroot which is a free way to code and that way you can plug uh a free coder into the agent OS you can get that inside the air profitable we have new tutorials like that all the time video tutorials step-by-step guides etc. And then also you can meet people in your local area who are building with AI agents like you. So, it's just a great community where we're super positive, learning, growing, supporting each other, helping each other, building amazing systems. And if you want to be part of that, feel free to check it out. Link in the comments description or just go to the aiprofitbomb.com. Thanks for watching. Thanks so much. We got some new test back from HY3. Let's see how it is. It's not great. Let's check this one out. It's not bad. But it's a little bit buggy. They all kind of look and feel the same, which is not great. Today I want to show you my favorite indexing tool for AI SEO. Now, one of the things that you'll notice when you're creating content with SEO is like it's pretty difficult to get it indexed on Google because you're creating so much of it across so many different platforms. So, for example, like we'll post on Twitter, we'll post on LinkedIn, etc. And the problem is there's so much content we've created and it's so easy to to create a lot of content with AI that quite often it doesn't show up on Google. Google can't index it. Um, and so what we've set up is really a powerful system using indexional where we can quickly index content, get it ranking on Google, and also particularly when we're targeting like trending terms or we've got new websites being built out. This is super powerful. Now, here's how I'm currently using it. So instead of going directly to index exceptional which you can do and you can like submit your URLs over there we can use this tool inside our SEO workflows with any sort of custom SEO automations we use. So what we actually do is we plug index exceptional directly into our agentic operating system. And then once we found some keywords we want to rank for using our system here for researching Google search console, we'll then plug in the keyword and the case study into our system and deploy it to our websites. Like so the great thing about that is that number one, we don't need to log into WordPress to create the content. We don't need to log into the indexing tool to use it directly. We just have a skill built into this system where we essentially create the content. Our CLIs and our agents can create the content on our websites and then it automatically goes through indexional to get indexed, right? And so this is really useful because then we submit the URLs to index exceptional and we get our content ranking pretty quickly and it's just a fully automated workflow. So, for example, if we have a look at some of the content we've created recently, like this one for example, you can see that that's indexed on Google directly and And then if we search that keyword on Google, you can see it's ranking on the first page for it as well. So a good thing about this is number one, it gets your index ranking uh it gets your content indexed on Google quickly, but then also it helps your content rank on Google quicker as well. So it's just an easy way to to get set up so that you can make sure your content ranks as fast as it possibly can. The other thing that I would say particularly for us is we're creating a lot of content around trending keywords. So, for example, like Hermes desktop installation, that keyword we just showed you is one that is trending right now and so we need to get it ranking right now as soon as we can. And so, the beautiful thing about this is that when we create content for those trending topics, we can get it automatically deployed and index really quickly as well. Now you also might say okay does that help you rank inside AI as well as on the first page of Google. So if we have a look at the content that we're creating over here you can see that we're ranking down here for this keyword and we're also ranking over here for this keyword right so we're ranking in two different places for that particular keyword. Now at the same time because of that and we only created this content last week right so this article and this piece of content was this one was only created two days ago actually it's already indexed and ranking on the first page of Google. So when we're creating this content you can see that Google AI overviews is ranking us number one inside the mentions as well. So the great thing about that is that we can index the content quickly. we can create the content quickly and even though we're creating a lot of it, it's showing up inside Google AI overviews here and here. And then also over here and then additionally with this piece of content, this one and this one. In fact, we've got three pieces of content ranking on the first page of Google for that particular keyword. And so it's pretty powerful. It's pretty easy to get this set up. And because you can get the API, you can just plug that into agents. Now, if you're wondering, okay, how do you uh get the API from indexal? So, they have the full API documentation over here. So you can grab that API documentation and then what I like to do is I will just go into claude and we can plug the API documentation along with my API key into claude and say hey every time I create content save this as a skill and make sure using my API key and the API documentation here make sure that you index the content I create as well and then also when we're doing this we want to save it as a skill right so we make sure that when we give the API documentation and the workflow to claude we save it as a skill now if you want to just get my full setup where you can just plug in your own API key into the system within exceptional then you can get our SEO content pipeline it's available inside the agent operating system inside the air profit board link in the comments description or go to the air profitable.com um with that system. What you can actually do is you can research what keywords you're already ranking for from Google Search Console. It will show you how many clicks you're getting, how many impressions you're getting, the score in terms of the potential of it, and then also the average CTR for that keyword and the average position. So, it's kind of like your own index ranking tracker, except it's based on your data, not data from like 6 months ago on a a tool that everybody has access to. So, that makes it more unique and more useful for you. And then additionally, what you can actually do from here is you can select the keyword that you want to create content around. So you click use topic, plug that in, then plug in a unique case study like this, and then from there you can click generate five articles and deploy those to your websites like so. And once that's done, you can see that we've got that content plugged in on the system. And also, as you saw before, we're already ranking on the first page of Google for that particular keyword as well. And I've shown you proof that every step of this works. So index exceptional is my favorite right now. seems to work really smoothly, seems to actually work. I mean, content that we created two days ago is already ranking on the first page of Google. Plugs beautifully into your agentic systems as well. And I think this is going to be one of the biggest problems. I was actually speaking to James Sulie yesterday about this as well on our SEO podcast. Like in the future, the the easier it becomes to create content, the harder it is to get it indexed because Google's not going to crawl every single piece of content that people create. And so, how do you get around that? Well, I think one of the best ways to do that is to make sure that you have a good indexing system in place that doesn't require you to like click loads of buttons to go in and everything else. You can just use the API and you you're good to go. So, if you want to get our full agent operating system with Claude Hermes, the SEO content pipelines, we have the video agent as well to to create videos and get those ranking as well. Um, we have, for example, like notebook learn, we have a full memory system here so that when you're creating SEO content, it's got context and it understands you and everything unique about your business when it's creating it. And also, for example, every time something new comes out, like for example, H Y3 Koda from Tencent uh HYN just came out today and we've already plugged it into the system so that you can use these powerful systems and test them out for yourself. So you want to get all that feel free to get it inside the AI profitable volume link in the comments description or go to the arrum.com index exceptional it's just indexal.com great tool for indexing content um we have loads of AI SEO workflows inside the AI profitable bottom as well so you can see all our best trainings on the AI SEO automations that I personally use. We also tested out something new recently which is SEO office OS and that was really powerful and fun to use too. We have the Asian operating system over here if you want to grab that from us. And you can see when it was last updated, you get the video tutorial, a full step-by-step guide, and the resources to install it. Then every time something new and useful comes out, we add it right here. Inside the calendar, you can drop a weekly coaching calls, get help and support real time. Inside the map, you can meet people in your local area doing SEO. And there's lots of AI SEOs inside here. And that's all available inside the profit warden. And if you want a free SEO strategy session, feel free to get that at goldie. And on that call, what we're going to do is look at your website with our team of world class experts. We'll give you an SEO domination plan. So you'll get a custom tailored link building plan to generate more leads to your website. You'll discover the secrets to SEO link building. We'll answer any questions you have onetoone. You'll learn the best link building strategies for your website plus how to outrank your competitors with link building based on what's working for us. So feel free to get that link in the comments in description or go to the aiprofit.com. Thanks for watching. Back in a sec. Today I'm going to show you how to run Hermes agent for free forever using something called Omniroot that just dropped. It's a brand new free way to code across like 231 free providers. And so what you can do with Omniroot is you can plug this into like pretty much anything like claw codeex. I've already done tutorials on that today. We've tested out with Hermes agent and I'll show you exactly how this works step by step today. So basically with omniroot this is a way to use coding tools for free right loads of different free APIs inside there. Now what you can do with that is because it's open source you can plug that in to Hermes agent. So you can see for example we have omnirute which is the free profile with Hermes agent and so for example we test out here are you working? Yes it works fine. Okay. So, for example, we say, okay, I don't know, um, code out a beautiful landing page. We can use that like so. And if you're wondering, okay, how does this work? So, essentially, this is an open source project and it means that it's got auto all back across 237 providers and that works in milliseconds. So like if one of these free APIs gets rate limited or you can't use it, no problem. It automatically switches to the next one. It's also got RTK and caveman built in. These are both ways of compressing the number of tokens you use. So you reduce the amount of tokens which is great because if you're using free APIs then quite often you can get rate limited and this reduces the chance of that happening. Also it's free to start. There's 90 providers with a free tier and 11 of them are actually free forever which is pretty cool. Every tool actually works. There's one endpoint and it's production grade. So if we go back into Hermes now, it's beginning to use that system with Omniroot. The way that I like to set this up is I'll have a different agent profile for each Hermes agent I use based on the API. And that means I can just test them out side by side. I have different conversations and different conversation history for each. I can use two Hermes agents at the same time. And every time I want to test something like this, we can go directly into it. So, for example, here you can see that Hermes with Omniroot is selected. But, for example, like Tencent, Hanyan, HY3 just dropped today. We already plugged that into the system as well. If you're wondering like how do you actually use uh Omniroot with Hermes agent? I'll show you that in a second, how to set up. So, basically the way this works is you got Hermes agent, which is the hands. Then you got Omniroot, which is the brain. And then that plugs into Hermes to actually get stuff done. So for example here you can see that we've got the um website fully built with omniroot. So if we just run this terminal command like so we can open up the page we just built with Hermes and omniroot pretty nice pretty clean. So you can actually code out locally you can do a gent task as well using omni route and you can use it for free forever with Hermes agent which is pretty mind-blowing in itself. So let's talk about how this works step by step and how to use it. So there's three things that happen automatically. Number one is compression. So reduce the amount of tokens you have with RTK and Caveman. Number two, roots across 93 providers. And number three, you can switch in one word um between those different models. So Omni Root is basically like a small program that runs on your computer. It's a gateway like a post office that knows how to reach every AI provider in the world. So you wire Hermes to that gateway instead of talking to Open Ianthropic directly and Hermes sends a request. Omnoot reads it, compress it, picks the right provider and roots it. So the model answers and omni root sense the answer back and Hermes doesn't really know the difference. So this is pretty much how it works. Now the cool thing about this is you can prompt it then you can use the Hermes profile for omniroot. Omni root compresses it. It selects a provider automatically. All this is happening in the background automatically without you doing anything. So what you're doing is just chatting to it like you would normally of Hermes but the difference is that it runs through this whole model right here. The other cool thing is you could plug it into custom workflows. So for example, we have Hermes agent with a custom workflow for lead generation and sending our emails and managing our inbox and we could easily just plug the model for Omniroot into this system, give it access to our emails and then it can go off and send emails as well directly, right? Which is pretty cool. Now also what we have over here is a memory system. So with Omniroot we can plug in Hermes agent into our memory system and then it can recall context and knows everything about us straight away. So as soon as we switch the API, no problem. We've also switched the memory and everything else because it works as one powerful system inside this full Hermes agent OS system that we've created. Now you might say, okay, how does this work when it comes to actually setting up Hermes with this model? So the way that we did it, if we have a look at the profile inside our dashboard. So, if we go to manage inside our dashboard here and we'll wait for that to load in the background. So we can go to our profiles. Then we're going to find omniroot which you can see over here. And then that's routed to omniroot directly. So this runs via this gateway and then we set it up inside our profile. The pretty the cool thing about this is like the old way is you have one provider, one API key, one model. It uses up a lot of tokens, the provider goes down sometimes, which means the hermy stops and the agent stops as well. And then you only have a few models to choose from. With this new setup, you got 237 providers, one endpoint locally, RTK and Caveman compressor token, so you don't use too much with your agents. It switches the model automatically. Omniroot is local, so it never goes down, never limits you either. And then you've got free models running in the background with it directly. Now, if you're wondering how to get Omniroot running, it is non-technical. So some people say, you know, you have to be technical to set up. No, no, no, no. So you just run this command and then you run this right and if you want like the full details you can just go to their GitHub to set it up. Now you may also say these models are not free but they actually are. I saw like some people mention it on my tutorial the other day, but no, it's it's free and we've tested it and actually works and there's a bunch of free forever providers as well. So you got Open Code Zen Pollinations QoA, and Cairo AI2. So these are all forever free options. You can also add open router as well. So if you want like 353 more models behind one key then you can set up open router too. And that has a free routter too. And then if you want to set up an API key for Hermes, you can just use a terminal command like this. Now you might also wonder, okay, how does a token suppression work like because that's pretty useful for whatever different API you use. So RTK basically reduces like repeated patterns and the duplication and fluff and caveman reduces the fluff on the way out. So what caveman does is it makes your AI speak like a caveman on the output tokens. So when it replies to you, it replies very blunt and very brief, which reduces the amount of tokens your models use. This is also super useful if you're using like for example Claude or something like that. So that's basically it. That's the whole system. We've shown you how to use Omniroot, how to use it for free forever with Hermes, how it actually works, what it can build, what the outputs here. Uh I wouldn't say this is like Fable 5 level, but at the same time, it can build stuff. I've shown you what it built. It works pretty nicely. It's pretty smooth. It runs for free, and there's a lot of free providers inside there. Plus, it's pretty easy to set up. So if you want to get our full agent operating system, we've got loads of great workflows for Hermes. So if we have a look here, for example, you got the talk mode, you got the voice agent, we have Hermes Oracle for tracking latest news. We have Hermes Astros that looks at your competitors and then gives you content ideas with new angles. We have the outreach store. We even have mixture of agents plugged into the system. And we've got a memory system here to give all of our agents context in like one click. And we have loads of cool workflows for SEO, for video agents, for music, for loop engineering, whatever you want. So, if you want to get that full system, it's inside the AI profitable boardroom link in the comments in description or just go to the AI profitable.com. And if you want the full setup from me, you can go to the classroom, then go to new daily updates and you get the agent OS here. You could build this out yourself. I spend about 3 or four hours a day building this out and improving it. So, if you want my setup, which I think is a lot easier and faster to implement, you can get it right here. We have a video tutorial. You see when it was last updated, you get a full guide on setting up, plus the resources, and we update this with new features all the time. So, for example, Omni Route is a new system we just set up into the agents as well. Inside the community, you can ask questions, get help and support. I personally answer each of these with a video tutorial. Plus, you can get access to all of the other um community here. And there's always people online 24/7. Inside the calendar, you can jump a 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. And that's all inside the AR profitab link in the comments description or just go to the arrum.com. Thanks for watching. or the implementation. I'm not sure cuz I got my agents to set that up um for me. But basically, I just said like we've heard me set up a rule where every time I speak to you, plug that into Obsidian so that it auto updates. So that's just like part of the workflow that it's trained to do. But for the actual implementation, like I I don't really do the coding myself. So I just get the agents to set it up and then I check that it actually works. Nice. Nice. Yeah, you can just ask the agent Philadelphia. That's what I'd recommend. The only good build that I can see on this system is The rest look quite bad and really buggy as well. Not sure why you scored them so highly. Did you actually test playing them? So, there's a brand new free model from China called Tencent Hunion and you can actually use this with news research right now. So HY3 the new 295 billione model from 10centent hun is now free in news portal for the next 2 weeks. Now this is an aentic model. It has 256k context window. If you want to know how it performs I'll show you in a second. And this literally just dropped a few hours ago. So you can basically run it inside Hermes agent for free for the next two weeks. So you can see for example here we have HY3 set up with news portal here. So we ask how you do today did pretty well. And then if we want to test it out let's give it a little task here. So we're just going to go inside Hermes and say okay code out beautiful landing page. And then we can wait for it to load. Now, if you're wondering, okay, how do you set up HY3 with news portal and Herby's agent for free? It's pretty simple. So, what you're going to do is go to your terminal here. Then you just type in Hermes model like you can see. And once you've done that, then you shall go to new support at the top. So, select that one. And once you've done that then you can just grab it from news portal. Now, also what's cool here is you can also get a free API for HY3. So you can see this is going away on July the 21st. But if you don't want to set up Hermes agent via news portal and you want to use Hermes agent via the API instead, you can use open router and then just select 10 cent hy. So that's pretty cool as well. And this is designed for aentic workflows. So you can see here it says specifically designed for agentic workflows, built for reasoning, etc. It's just quite a cool way to get a free API and then plug it into your Hermes agent. Now if you're wondering how it actually performs in reality, let's have a look here. So, for example, we said code out beautiful landing page. You can see the test here. It's given us a link. Let's open it up. Actually looks pretty nice. It's not going to be like Fable 5 level, but it's pretty nice to use. Pretty easy and simple to use. It was fairly responsive as well. It took about 60 seconds to build out the full landing page, which is not too bad at all. And then we can actually use tool use with this as well. So, if you go inside the chat here and we're like, okay, open up this website for us. it actually works. Right? Then what we can also do is try forward slasharn as soon as a task. So we can type for/arn and then give it a link to a guide that I've created here and see if it can actually learn about the guide. So if you're not familiar with for/arn, basically you can teach your Hermes agent a new skill based on the guide and then it will save it for future workflows. So you can see here typically when I use HY3 with Hermes, it takes about 60 seconds for each response. So it's not like super fast, not like chat GP instant or something, but for a free model, not too bad. And bear in mind, like you can just give it task and it work in the background. So that actually took about 30 seconds as you can see right here. And then what it's done is said, "Right, I read the page. Here's a breakdown. Here's what I've learned." And then also it can save that for a skill in the future, which is pretty cool, too. So, it works in three simple steps already. You got Hermes and you can plug that inside your agent OS. Then you can use newsportal or open router as well for 2 weeks and then you can use HY3 as a brain which is good for coding and using tools and aent workflows. This is uh open source model as well and also one thing to note is that it's also free inside kilo code. So you can also get access to it on kilo code as well. So open router Hermes kilo code all good options for using this. If you actually look at this comparison here, so they've compared Opus, Fable, and HY3 in my own tests, it's nowhere near nowhere near the same level as Opus 4.8 or or Fable 5. And you can test out on uh Goldiebench if you want to see the outputs that we've done with 10 cent HY3. But still, if it's free, you can't complain, can you really? And also, you've seen that it's non-technical to set up. So, you don't need to be technical to set this up. It's like a couple of terminal commands. You plug in the model. That's pretty much it. Now some people watch these tutorials and the like it's not free but you can see here that Hermes agent is a free open source model so you can access that for free and then also HY3 itself is free on open router as you can see right here and news portal and also The good thing about this is if you are using an agentic operating system like we are then with these systems you can really reduce the amount of tokens you have. So for example you could delegate subtasks or sub aents to HY3 and then you could have like a more frontier model as a brain. So you could have for example like GT5 as a brain then that delegates subtasks to the free models like for example HY3. One thing that we've tested as well if you like free models um and this is especially useful if you get rate limited or throttled by HY3 if you're using a free model is you can use something like omniroot as well. So the cool thing is here that inside our profiles and on the chat we can have one profile with hy and then one with omniroot and then if we get rate limited on either one we can switch between them which is pretty useful. So that's basically it. The other thing some people say is like I'll try this later, etc. But this is only available for like 2 weeks. So if you're going to use it, just make sure you use it now. If you want to get our full setup for the agent operating system, it comes with the chat with Hermes. We got Hermes Apollo, a voice activated version of Hermes agent, and we can build stuff out. As you can see down here, we have paperclip where you can actually build full teams of AI agents inside um organization chart as you can see. And then we also have for example Hermes Oracle, Hermes Astros studio. We have an outreach tool for sending emails. We have mixture of agents that can actually beat Fable 5 on benchmarks. And then we have for example a full control room and goal mode as well. And we've also plugged in the memory here. Also, if you like HY3, we have the HY3 coder over here. So what this means, you can plug in free API from HY3 and then use HYN with 10centent over here as well. And then preview what you build. So you can use it with Hermes or directly the coder. And this is all available inside our agent operating system. link in the comments description or just go to the aiprofitable.com and you can get our agent OS inside this community. This is my community for just helping you learn, grow and scale with AI automation. Inside our document here, you can see that we have over 196 pages of testimonials and wins. And for example, like Jacob set up the agent OS in like 10 minutes. You can see t David is absolutely loving it. Like loads of people are really enjoying the agent OS right now. And if you want to get our system, I spend about 3 to four hours improving it every day just to help you. And you can get that inside the AI profitable bullet. So if you go to the classroom here and then after that, go to the new daily tutorial section, find the agent OS, you'll see the video tutorial, the last update date, a full guide on how to install it, and the zip file with the resources right there. And then every time something new and useful drops, we add a full video guide and step-by-step tutorial here. Inside the community, we answer your questions and you get help and support in real time. And inside the calendar, you get access to all our latest coaching sessions. Plus, inside the map, you can meet people in your local area who are building with AI agents like you. So, feel free to get that link in the comments description or just go to the apiproforn.com. Thanks for watching. All right, back in a minute. Probably about 5 10 minutes. Take a little break and then I'll come back. Just waiting for this to get done. Thank you so much. I appreciate that. Today I want to break down the Claude agent operating system, exactly how I use it, what I built into it recently, and how it works. Now, if you're wondering what is an agent operating system, this is a system where we've got all of our agents plugged in together. So, they can work and build and learn and grow, and we can have agents running autonomously together to automate whatever we want. So for example here we have Hermes, we have openclaw, we have claude, anti-gravity, codecs all plugged into one system where they all work from the same memory and they automatically log their memories into this system here. So they all have knowledge about what they've done, what they've worked on recently and you can see like every few hours this gets updated with memories and new systems and new workflows. Now why is that important? It's important because then your agents understand your voice, your brand, your goals, your projects, even your clients, and everything's directed into one powerful system. As you can see, now we've been using Claude Fable 5 because it's just about to go away away from the subscription. We might as well make the most of it, and that's why we've created some of the new systems I'm about to show you today. So, for example, we have Hermes Apollo. Hermes Apollo is a voice activated AI system where we can have an agent ready to go on voice whenever we switch it on. We can have it working in real time and we can have it in aentic or auto mode. We can also type and chat to it over here. We can switch between different voices. We can see everything that we've chatted with it directly here. And then we also have a preview of everything that we've created. So you can see all these different things that were built using this powerful system. And so this voice agent is the most powerful way I've seen to use voice agents because usually people create them and never use them again. This way we've got it inside a tab and it's ready to build whenever we need it. We also have Hermes Apollo here. This pulls in the latest trending news. So you can see it was automatically consulted as an oracle uh four hours ago. So this is Hermes Oracle. And what it does, it pulls in the latest trending news. So I can always stay up to date with everything that's just dropped. So for example, today if we look at the latest news, this is the most useful, interesting news that's just dropped. So for example, we've got like information on Meta Watermelon that just leaked. We've got GT5.6, six, which is about to reach wider access soon. We have, for example, Chief Gemini 3.5 Pro and the 2 million context window leak. And we also have, for example, Claude Fable 5 that's just about to disappear from the subscription. And so we have all these pieces of news. And if we want to generate a piece of content about it, like for example, the uh Meta Watermelon Llama 5 leak, we can just click this button and it automatically publishes content directly to my website about that particular topic. And it can do it in the background whilst I go and do something else. which leads me to the next section. So, we got Hermes Astros over here and this analyzes all of my keywords, monitors different uh topics and then gives us new content ideas every single day. So, again, it's working on a schedule and it's just automatically updating and coming up with new ideas for content based on what everyone's creating content around, right? So, it helps us understand, okay, this is working. Here's some unique angles for you. Here's the type of title that it would recommend. So for example like Hermes versus Claude code and then over here it actually has links to quickly create the content. So if we click on the video agent it actually plug that prompt for the topic into our video agent. We can click on write the script and actually generate the video in the background whilst we're doing something else. Pretty powerful system right here. We also have this system where we can organize teams of agents to work on a camb board. We have for example Claude, openclaw, Hermes. If you don't use one of these as well, you can just hide it like so. I think the next step that I probably want to work on for this and already some of the air offer boarding members are giving me some ideas on that is to improve the UI on this and we're going to make it look and feel even nicer. So that's the next step for me in terms of what I'll be working on in the future. We also added new features. So the other thing I like about this is like as soon as something useful comes out. So, for example, like um Hunen HY3 just dropped a few hours ago, and we've already plugged in the free coder into the agent OS. So, the people who use this can code for free, preview what they've built, save it to their workspace, and it's all ready to go inside one place. And that's absolutely amazing when you think about it. So, basically, this is available for free right now. And we plugged it into our agent operating system so that anyone who uses our Asian OS can get it right now. Now, if you're wondering how to get access to this, you can get it inside the air profit boardroom. I mean, you could build your own agent OS if you wanted to, but I spend about 3 to four hours a day improving this daily. So, if you want to get my system from me, you can get it inside the air profit boardroom. You can see when it was last updated, you get the video tutorial and you can get the zip file to install it as well. And it's all ready to go right here. So, it's a pretty powerful system. Now, in terms of the layers, let me talk you through some of the systems we've set up and how we've stacked everything in and how we've organized it. So, we have the command deck, right? This is a mission control. So, we can see every agent's desk and we can see everything inside one place. We can see how it works and we can see um even our to-do list for the day. And I can just tick those things off. So right now, for example, just finished this and you can see here that we've got our to-do list, all of our agents, the status, which version they're on, what endpoints they're using, etc. Pretty powerful. We can even monitor how many tokens we're using and we can see our journal, our memory, our goals, etc. Now, on top of that, we also have the model layer. So we have free models and we have premium models inside here. So for example, if you want to use like frontier models, you could use for example Claude or for example, Hermes. If you want to use free models, then you could actually navigate to Omniroot, which is a free coder, or you could use HY3. And either way, whether you want to use Frontier models or you want to use free models, we've got it built into this system, too. Top of that, we have an agent fleet. So, we have nine different coding agents that you can talk to inside this group chat. And then you can see that every time we ask them to do something, they will come back to us. And additionally, they add that idea into our ideas pipeline so that we can quickly implement and build that. And all I need to do is approve it and then it gets built. And if we want to see everything that we built, we can preview everything that we've built over here and we can see all of our creations which is pretty insane. Never seen anything like that. U but it was something that was like right we need that inside our systems. We need that ready to go etc. So there's so many cool systems inside here. The other thing I like is like you can orchestrate your agents as well. So if you're looking at the ideas pipeline, we plan, we build, we review, and we ship. And it just works in a beautiful system here. And so this is like a really nice camb board where we can go from idea to implementation really quickly. We also have a cambban section here for Hermes agent as well. And it's just about having all these different layers in to make sure that we get the most out of the systems. Now you might say, okay, doesn't this use a lot of tokens? So we actually put our agent OS on a token diet. What I mean by that, for example, is when you look at the systems that we built, we have free models inside there so that you can use free systems if you're worried about tokens and you want to minimize the amount of tokens you use. But also, we have three different systems plugged in as well inside here. So, we have RTK, we have Caveman, and we have Ponytail. These are three different open-source token minimization systems that you can use with your agents to reduce the amount of tokens they use. Now, if you're wondering, how do they work? RTK cuts out the fluff. Caveman makes your agent respond like a caveman. So you use less output tokens. And then ponytail was great for coding because it basically works like a lazy senior developer who doesn't want to code much or work too much. That's basically the three systems that we have logged into here to reduce the amount of tokens you use inside your agent OS. Also we have for example the content factory. So a lot of people ask me like how do you automate content? How do you create so much content etc. So what we do is we have one topic. So we can plug in one keyword and then for example if we're looking at the Hermes Oracle setup here, we can just publish that to WordPress or we draft social media content in one click like so. We can also use our video agent as I showed you before and that can create amazing AI avatar videos as you can see right here and these look pretty awesome. And then also we can automate SEO content. So we can just plug in a keyword, plug in a case study and create content and deploy it to our websites right here. Right? So, however you want to create content, we've got some amazing systems for it. Additionally, you can see here, for example, we can generate videos with the video agent. We can generate SEO content with the SEO agent. We have, for example, the uh notebook LM set up inside here too. And what that means essentially is we have like a full content library where we can generate content in like one single click using Nogramm. And no glamm is free to use. And then we can generate for example short videos. We can generate research. We can generate infographics. So this infographic was created for free with notebook inside our agent OS. This short video as you can see right here is generated with AI using notebook and the new short video update. We have for example this full research report that is very very in-depth super useful and again we just created it with no bookm in like one single click. So, as good as it looks, it was super easy to create. And so, this is really a powerful content factory that you can plug into your systems to get the most out of this as well. On top of that, we have the studios. So, we have Open Design, which is kind of like an alternative that's open source to Claude Design. We've also plugged in a game studio, so we can generate games whenever we want. And we can also use Open Montage. Now, Open Montage, if you're not familiar with it, is another open source project where you can create movies with AI in like one single click. So, all of these cinematic videos you're seeing right now, they're generated with AI. Super easy to build, super easy to create, and it's using a free open source project called Open Montage. So, we plug that into our systems, too. So, anything useful, I just plug it in there as well, just to get the most out of it. Now, what we also have over here is the memory system. Now, for this, this is a free system. It's called Obsidian, and we turned it into a memory galaxy where every star is a memory, and they're all linked together inside one solar system. And then you can see that we can click on any of these memories and learn about them and see what they are. And so, we have today's work and today's lessons that iterate and improve the whole system because today's work and today's lessons go into the vault on the memory files so that tomorrow when we're using the Agent OS, it's even better than it was before. And the great thing about that is that this whole system gets better and improves every single day. If you saw two months ago what we've created with agent OS, it is nowhere near the same system that we have today. Why is that? Is because it's improving. It's getting better every single day. It's always leveling up. And that's exactly what we want, right? We want to make it better, better, and better, and better as much as we possibly can. Now, also, what we have is an amazing community that comes with this system. So we actually have but 196 pages of wins, testimonials, reviews, etc. as you can see from people absolutely loving this agent OS. So for us like you could create this yourself, but if you want to get my system, it's inside the profit boardroom. And the great thing about this is that you can be part of a community that's always learning and leveling up and sharing great tips and sharing knowledge and sharing stuff that they probably couldn't share anywhere else in terms of what they're building. So for example like Jacob and you might say this is technical or timeconuming set up. Jacob actually set it up in like 10 minutes and he was working on something similar as you can see and he was spending 12 hours a day trying to build this out. So with this whole system he installed it in like 10 minutes and saved himself 12 hours a day which is pretty wild but it it is super timeconuming to build this stuff. I mean for me I test AI daily. I test new systems and even for myself like it takes me about 3 to four hours to improve this system every single day. So, loads of people are just loving it. And I think the great thing about this community is like we're all just learning, winning, growing together, people sharing the wins, people enjoying the journey, uh, and people just leveling up every single day. And I think that's really cool when you think about it. So, that's basically it for the agent operating system, how we're step by step. Again, you might say, well, doesn't this use up loads of tokens? But we've got token minimization playbooks and also we can use free models like Omniroot or HY3 inside there. You might also say, "Well, I'm non technical, but I'm non technical, too, and so is everyone else in inside the community." I mean, there's a few technical people in there, but most people are nontechnical, and they can all set it up as you saw before. And also, you might say, "Well, tools change every week. So, won't this be obsolete in like a couple of weeks?" But that's a beautiful thing about having a system, right? So, for example, when Hermes dropped and we couldn't use it again, we took out the system and it took like 10 seconds, but the system was still there. When Fable 5 came back out, we plugged it back into the system and we could use it within 10 seconds again. So, it's very, very flexible. And it doesn't really matter what model comes out. What's going to stay true to you and what's going to last for the next 6 to 12 months is the system, not the model. And so, we can swap the models in, we can swap the models out, but the main thing is we've got systems that stay the same and improve every day. And that's what we're doing with the HOS. So, if you want to get this whole set up and get it inside the air boardroom, everything that you've seen today were plugged in. I also answer tutorials and community questions about this every single day with a video tutorial. So, you can post questions inside the community and I will personally answer them as you can see right here. So, people are absolutely loving GM 5.2 right now and we've shared loads of great stuff about it and also if anyone has any questions then I personally answer them and so does the whole community which is pretty cool as well. So there's always people online 24/7 to get help and support whenever you want to. Inside the calendar, you can jump on weekly coaching calls, get help and support real time, share your screen, meet other cool members, and ask any questions you need. You can also meet people locally near you in your local area and connect with them and meet other people using this agent OS. And then also inside the classroom, you can get access to all of our best trainings on this stuff. So for example, if you want to get the agent OS is right here with video tutorial. You can see the last update date. You get a full guide on how to install it and also the zip file to install it too. We add new daily tutorials based on what's actually useful. And then inside the classroom as well, if you're a complete beginner, you can go from beginner to expert over here. We also have loads of great trainings as you can see. So feel free to get that link in the comments description or go to the aiprofitborn.com. Thanks for watching. No, I haven't had any issues with the the memory galaxy at all. But if I did, I would just fix it straight away. Um, honestly, like I've not, you know, it's really funny you say that cuz I see a lot of people worried about plugins and how to use them and which ones are good, which ones are bad. I haven't really seen any bad plugins. Um, there was only one time with a a plugin that I used that was bad. I think it was Yoast and it literally de-indexed all my content on my website and that was the only time that I saw that. But that was an SEO plugin for WordPress. Uh but for skills and that sort of thing, not seeing anything bad. Uh you do need to be careful having said that. But I always show like how to be careful, how to scan this stuff yourself, how to make sure that you're careful when you're setting things up. So yeah, uh I mean if you're ever unsure, here's the way that I would look at it is if you're unsure, don't use it. So if you're like, I'm not sure about the skill or I'm not sure about setting up, etc., you don't need to. You know, you can create your own version. You can create your own workflow. Might take a little bit longer, but it's 100% worth it, right? That's the way that I would look at this. Today we're going to be testing out 10 cent HY3 which is a brand new open-source project that's available on a free API and we've been building all sorts of stuff with this morning. I'll show you how it performs, how it goes, etc. I've already built it into our agent OS and you can see an example of it working right here. Now, this is open source. It's just dropped from China. Um, I'll show you the benchmarks in a minute. If you want to use it, basically you can get it on the API for free with open router and it's available on Open Router for free as you can see right here. Um, if you just type in HY3 free, you will see it over here and then you can get the API and plug in it. So, that's pretty cool. It's available until July the 21st on the free API. You can also get it on news research with Hermes agent for free and it's also available on kilo code for free as well. So open source free. What's not like to like? Well, I'll show you some examples of what I do like and what I don't like. So you can see actually here like the website that it created is not bad. Not bad at all. We said build out a beautiful landing page. It went off and built it and we can see the preview and we can see our saved creations over here. So not bad, right? Not a bad start. However, let me show you what I do and don't like about it. So, you can see an example of what we built here. This is kind of like an open world game and everything. Every new model that drops, we test it out on Goldie Bench and then we see how it performs and we compare it to everything else. Now, if you have a look at this game, it's not bad. It just it doesn't feel quite right. There's something about it. I would say that's probably one of the better creations from H13. But you see how that light over there, it's just it just looks a bit weird, right? It's just a bit glitchy. And this open world game is not bad, but it it's not really playable. It's not really fun to use. Let me show you another example. So, you can see here, it's kind of like a a Lego man that's been created on this uh example, and it's an absolute giant compared to everything else. But if you look at the graphics and the details of this, it still feels very basic when we're using it. And that's something that I've actually found across pretty much everything that we tested it with. So you can see another example here. We can like drive around a city. I would say this is probably the best creation that we got so far. It looks quite cool. It's quite fun to play. But if you compare that side by side versus, you know, something Frontier uh like Fable, or you compare it versus like GLM 5.2, which we'll do in a second, I would say it's not really on the same level. Like even look at the the city blocks here. They're just super basic and that sort of thing. So it's awesome that it's free. Awesome that it's open source. Is it the best model or one of the best models I've ever used? No. Would I say it's right up there? Probably not. Like, see how it's super glitchy in terms of using this and then we fly in the air. There's no detail. There's nothing going on as well. I'll show you the final example here. So, that is for this parachute game, which is quite fun, like fun background and that sort of thing. If we release a parachute, that's what it looks like. It's it's it's just not quite on the level that I would hope. And we did do a lot of back and forth on this as well. So, if we actually have a look at our um HY3 tests, we've been coding and testing this out. Let's have a look here for like 2 or 3 hours. So, you can see even like 2 hours ago, we were testing um the flight simulator and it still wasn't quite right. So, we've been going back and forth and wrestling with it for like 3 to four hours and still not getting the outputs even when we give a lot of feedback to Fable 5 to test it out more. Now, if you're wondering, okay, how does it perform on benchmarks? Let's pull them up. So, here are the benchmarks and you can see this is versus other models like Claude, you got GM 5.2 and honestly from what I'm testing you can see on MCP Atlas, Browse Comp, etc. claw eval. It's It's kind of well, it's beating GM 5.2 on claw eval, which is like an agentic workflow. When I've tested out myself, it it doesn't seem to feel like that at all. Doesn't seem to be on the same level. So, let me show you an example what we built with our models using the same system. So, this is the test. As you can see, if you compare that to other models, you see the quality of that versus uh and this is with Fusion. The quality of this is way way better versus HY Y3. So, there's a lot more detail on most of the models that we tested out with the same sort of prompts. Let's have a look on the driving game as well. So this is with Fusion and you can see like the quality is so much better than what we saw with HY3 before. This is with Hermes Mix of Agents. This is with Fable 5. So it's nowhere near the same Frontier level and it is open source. So I don't expect that. If we compare it versus GLM 5.2, I would still say that GLM 5.2's 2's output is way nicer, way better in terms of the vibe, the feel, the colors, the detail, etc. versus HY3. Having said that, what I do like about it is it's not bad on Hermes. So, we plugged it into Hermes here, and I think this is really what it's designed for. I don't think it's designed to be like a Frontier Coder or anything like that. But if we test out HY3, which we've set up the profile for here, and we say for example like open up Google, it seems to work pretty well, right, for browsing. If we say forward/arn learn this skill, that works pretty well as well. So it can read um pages pretty quickly. It's pretty fast to use as well. And if we say like code out a beautiful landing page, it's not bad like not bad at coding at all. I've seen worse from paid models. So it's not bad on that perspective either. So it has pros, it has cons. I mean, one of the best things I think you can do with it is use the HY3 coder like this. So you can basically build out a coder using the free API from open router. You can put your prompts on the left hand side, have your previews on the right hand side, and then your save stuff down here. That's really cool. Like that is fantastic. We also tried creating a video with it as well. So if we go back to HY3, you can see the video promo that we created up here, which is not bad. So basically what we did here is we gave HY3 access to Remotion and then it can code out for us. So it can create like full videos, full animated videos using the Remotion skill and that's certainly better than a lot of open- source projects that I've created. So it's pretty nice, right? That that video itself is good. And again, I think this is much better for agentic tasks, not for con frontier level tasks. Right? So for example, if I'm building out the agent OS, I would every day of the week, I would much prefer to use GLM 5.2 to or for example Claude Claude and Fable 5 or Opus 4.8. But if you're using it aentically and you just need like a free quick model that can do tasks and run tools, then I think actually HY3 is pretty good. Now if you're wondering about the technical details of HY3 so it's a 295 billion parameter model. It's got meter of experts. So has 21 billion active parameters. if you were trying to run it locally. Let's have a look. I can't see it on hugging on um LM Studio yet, but I imagine that's going to drop pretty soon. Interestingly, the older models were pretty small. Like only 8 GB. That's pretty small for a local model. So, be interesting to see if you can actually run it locally. Again, if you read the model details, you can see that it's really designed for agent capabilities. I wouldn't say it's the best open-source project from China. I think that I would pick GM 5.2 first and then Quen 3.7. Deep Seek I don't really tend to use so much on the actual benchmarks. It's performing pretty well. But that's why we created Goldie Bench. It's really just to test that stuff out. So if you're not sure, then you can test it yourself or just have a look on Goldiebi Bench and and see what you think. So, we've got everything that we've built with it so far here. We'll run it on all 45 tasks later today. And if it's free, you might as well check it out. You might as well test it for yourself. So, thanks so much for watching. If you want to get our full system with HY3 coder built in, and we've also got a system for Hermes where you can plug in HY3 there as well. And this is our powerful agent operating system for basically building and automating anything and plugging it all together inside one beautiful system that works amazingly well. then you can get that inside the AI profitable link in the comments description or go to the AI profitable.com if you like open source projects and local models as well. We've got trainings on Gan 5.2 and other Chinese models for example like quable that you can test out inside here too. This is our community for learning and growing with AI automation. You can ask questions inside the community. I personally answer them with video tutorials every single day. If you want to get our agent OS system, it's inside the classroom. Just go to the agent OS system right here. And we update this daily with a video tutorial, full guide and zip file how to install it. And then every time something new and useful comes out, we actually test it and then create a video tutorial and a full step-by-step guide on how to implement it. 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 like you. So feel free to get that link in the comments description or just go to the aiprofitborn.com. Thanks for watching. GLM 5.2 versus the new 10 cent HY3. Who wins? I'm going to show you side by side the software we built so far and how they perform and who wins so far. If you're wondering what's the difference between these two models, well, HY3 just dropped. GM 5.2 dropped about two or three weeks ago and HY3 comes from 10-centent Hunette. Whereas, for example, GLM 5.2 comes from Z AI. You can get access to both of them for free. So, if you want free access to 10centent HY3, you can actually go to Open Routter like so. And it's free. It's open source. You can get a free API for it. and it's available until July the 21st for free on Open Router. If you want to use GLM 5.2 for free, you can actually get access via z.AI and then you get access right here. Um, as far as I'm aware, you can't get the API for free with GLM 5.2. But if you're wondering how do they perform side by side, let me show you some of the examples right here. So, the first test that we ran here was with GM 5.2 to and 10 cent hunen. All right, so this is the example from 10 cent hunen. And one of the things I will say here is like it's pretty buggy when you're trying to navigate around, but it doesn't look bad. Now, we did have to go back and forth with this a lot to use it, but again, you can see it's almost unplayable like in terms of the way it's just difficult to navigate anything else. If we have a look at GM 5.2's creation for the same model and test, let's have a look here. So, this is a lot smoother, a lot easier to use. The graphics are not quite as nice, but it's still it can create something decent, and it's much easier to navigate and play. So, I would say GM 512 won that round for sure. And let's check out the next one. By the way, these are both Chinese models. So, it's interesting to see that both models are open source and both have come out of China. Now, this is GLM 5.2's attempt at something called Dragon Realm, which is kind of like a an openw world 3D game, right? And it looks pretty cool, like RPG game. Actually works really smooth, nice graphics, easy to move around, you know, that's that's a pretty good creation. Now, let's have a look at HY3. And you can see here, it's not quite as smooth. It doesn't feel quite as nice. And also, you see how you can't really uh you don't number one, you don't really get many characters inside the game. Number two, it feels a little bit laggy and slow. So, when we move around, for example, that Lank just bugs around a bit. Uh and also, it's just not as enjoyable to play. Now, if you're wondering okay how did you test this? So we got Fable 5 to test it and literally we were going back and forth with it for like 3 hours trying to get good outputs out of HY3 and just couldn't really get there. You know, it just wasn't quite smooth. That's the thing that I found is like GM 5.2 just first time round, it will give you better outputs than using HY3. So if you look at both of these rounds, GLM 5.2 definitely won. Let's have a look at the next one. So we've got a flight simulator here. So, this is the version from GLM 512. Kind of interesting, like first person. Um, terrain is not bad. Easy to navigate, easy to use, to move around, etc. Um, looks pretty cool. Now, let's have a look at the version from HY3, which begins pretty nicely and then kind of goes a bit wild. Like, once you've flown up in the air, you can't really do anything. Um, and the terrain just totally disappears. All the detail disappears. It's It's kind of a weird vibe. So, I'd say again, it just doesn't feel quite as detailed and fun to play as GM F2. Next up, we have this driving game. So, we can drive around a city, a neon city it's called. And this is a version from GM 512. Pretty fun to use. Still basic, but it it works. Um, we can navigate around. It's pretty smooth. There's other cars in there. It comes up with like pop-ups on the text as well. And then we got a bit more detail here as well. All right. Now, let's have a look at that compared to HY3's output. So we can see HY3 here lacks detail is a little bit buggy and just kind of feels a bit more dull. I mean like if you look at the colors and everything else it works but again it's not it just feels more basic. Feels like an older generation versus GLM 5.2. So on every single round that I'm seeing GM 5.2 is beating HY3. Let's test the next one. Now, this one is actually totally buggy from GLM 5.2 to be fair. So, I would say GM 5.2 totally failed this task. Let's have a look at the generation from HY3. And that actually works. Now, it's still a little bit buggy. still lacks detail, which we've seen in pretty much every generation, but it works. So, it beats GM 5.2 on that round. So, if we recap so far, 1 2 3 4 GM 5.2. HY3 is one the other um generations. Then we got two more to test. So, here we have like a a parachute game. We can jump out the plane and then we just have to aim for the section over there. All right. So, that's GM 51.2. And then if we have a look at the generation from HY3, I mean, this it doesn't really look like a parachute. And there's no detail on the ground either. So, I would say HY3 won that one. Uh, sorry, GM52 won that one. And then finally, we've got a cinematic style video advert. This generated with Remotion. So, we've created the same video with GM 5.2 2 and with HY3 and we can see here G 5.2 is not bad. So it's created like this nicely designed video. And if we have a look, we've got the other version from HY3 here, which I mean the colors are not quite as rich. Doesn't look quite as nice, but still works. It's an actual video. It's purely using HY3. So it's pretty amazing itself that the AI can do this and we can use the API which is free as well. So this is generated with free open source skill called reotion. So overall I would say GM 5.2 won that round but HY3 didn't fail completely like it did a decent job. So if we have a look at the rounds in total GM 5.2 two one 1 2 3 4 5 6 rounds and HY3 one one round. So overall and also one thing to note here is Julian 5.2 has a bigger context window at 1 million tokens versus HY3 which is 256k tokens. I would say that the best open source model right now is still GLM 52 from Z AI. However, having said that, there's two key benefits to HY3. Number one, you can go to it for free using the API, which you can't do with GLF upon 2, and then also you can get it for free with Hermes agent if you go to news portal. And then select your model provider as HY3. So, there's two benefits there and two good options. You know, test out yourself. That's why we created Goldiebench cuz we stopped listening to benchmarks and started testing everything ourself. And either way, you can use them together inside an agent operating system like you've seen today. So, if you want to get the model the most out of both models, you can plug them into your agent operating system, test them out for yourself, and see what you think. But overall, if I had a choice, I would definitely go with GLM 5.2. I think that's still absolutely crushing. And I can't wait to see what GLM 5.3 comes out with. So, thanks for watching. If you want to get a full agent operating system with all these agents plugged in, with all our best systems, all our best automations, etc., for the memory, then you can get that inside the AI profitable border link in the comments description or go to the AI profitable.com. Every time something new and useful drops like HY3, we test it out. We try with different projects. We see what we can get out of it and we compare it versus everything else. And if you have any questions, you can post it inside the community, get help and support in real time. If you want to get our agent operating system, you can get that inside the classroom here. And then you can see we have new daily video tutorials. And inside the classroom too, we have a beginner to expert AR automation course here. If you want to ask questions, jump on a call, share screen in real time, etc., you can do that here. Inside the calendar, we have four coaching calls a week. And then inside the map, you can meet people in your local area who are building with AI agents like you. So feel free to get that link in the comments description or just go to the approp.com. Final verdict, GLM 5.2 wins, HY3 not bad, but I think it still needs a lot of improvement. Thanks for watching. Bye-bye. All right, cheers everyone for watching. That is the end of the session. We've tested everything, answered every question, every comment, and every question inside the air profitable bomb community. Hope to see you inside there. Cheers. Ba.