GLM 5.2 + Hermes Agent: Build a Multi‑Agent Content Machine!

Today, I'm going to show you the power of GLM 5.2 plugged into Hermes Agent. And this is one of the best models that I've tried recently. It really does compare with something like Claude Opus 4.8. You can see what we've built with it here.

Here's an example of like a open world game that we've created. And you can get your agents to do this and automate anything with this. You can also see, for example, we've created some other cool stuff here and this plugs straight into your Hermes AI agents. So if we go over to our agent operating system over here and we check out Hermes, we've created a separate profile for GLM 5.2.

We can plug it into our Hermes Javis system, which is a voice activated version of Hermes that we can control and build anything with. And also the great thing about this is you can have multiple different profiles with GLM 5.2 plugged in. You could even use the CLI directly with this as well. And you can see all the cool stuff that we've built right here as an example.

So if you're wondering how can you use GLM 5.2 and Hermes together, that's exactly what we're gonna cover today. And this is something that we've actually built out today, which is the content machine. So for example, we've actually got multiple different agent profiles working together to create content for us. So if we have a look at this example, this is an example of a video we created using Hermes and GLM 5.2.

And it actually created the video, edited it, wrote it, et cetera. And it's pretty nicely edited as you can see. We also created a blog post and deployed it to our website using a team of AI agents. So we have like a content writer, a content judge, we have a content editor, a video builder, a video director, et cetera.

And this is on a Kanban board where we can get teams of agents working together, which is pretty amazing in itself. We also did this, for example, with this game that we built out. So we had a full team of AI agents working together to build and automate anything like you can see. And then they just went off and created this amazing stuff, right?

And the great thing about this is when you've got it plugged into an agent operating system with Hermes and GLM 5.2, you can see everything you've built, you can visualize it and you can come back to it later, which is mind blowing in itself. So you might be wondering, okay, like, how do you do this? How do you get your models plugged in? So basically what you can do is you can go to your terminal and from your terminal, you can just type in, for example, Hermes model, and then you would make sure that you go to ZAI and then you can add the coding plan with that directly.

Another way that you can do it is you can go over to your Hermes agents like so, go to manage, and then you go to models and you would just change the model, right? So you can manage all the models that you have plugged into your Hermes agent over here, which is super useful as well. You can also see, for example, what the profiles you've got and everything else set up as well, which is really, really powerful. And so one thing that I've actually done is I've set up this chat dashboard, like you can see, where I can switch between the models for different APIs because there's so much good stuff that we've actually got, for example, GLM with Olama plugged in over here, and we have GLM 5.2 over here too.

And in terms of how this got built and everything else, let me show you some examples. So essentially the way this works, and this is a content engine that was set up with Hermes agent and GLM 5.2, is that you can have a small team of GLM 5.2 agents living inside your Hermes Kanban board. So you can drop in a topic, a keyword researcher finds the angle, a writer drafts a post, an editor sharpens it, a judge scores it, and they send it back if it's not actually good enough. So only when it passes, does it move on, which means, for example, when you get a clean blog post, or for example, a powerful video and it's ready to publish, great, that's when it's gonna move on.

Otherwise, it's gonna keep iterating. And so you can drop in a topic in one line, the keyword editor finds the angle, the writer drafts it, the editor sharpens it, the judge scores it, and actually loops back to the writer if it's not good enough, and then it just keeps going until the blog and the video are published, which is basically creating a relay of GLM 5.2 Hermes agents with a judge that won't pass weak work. So all the quality control is fully automated inside this Kanban board. So if we take a look at this example, right, this is exactly how we do it.

So you can drop in a task over here, and then you get the finished work, as you can see. We've got, for example, a content editor, and we can see exactly what they've rated the work at and where they're up to. So if we wanna see, for example, a keyword brief, we can see over here in the judges, judge the quality of it, right? If we have a look, for example, at the content outline, we can view it over here.

If we wanna see the post itself, we can view that over here too. And so this just keeps going round and round until eventually it is fully published, which is pretty insane. And the same, for example, for the video. So if we have a look here, this is a video that was eventually created, did it look like this first time round?

No, Hermes agents just kept iterating and going round and round on the Kanban board until eventually it was ready to go. And again, didn't write this, didn't create it, you can use a free skill, which is Hermes plus hyper frames to create videos, and then you can get GLM 5.2 to just implement and go off and do that for you, right? And so it's so powerful, it's absolutely crazy how cool this is. And so we can also see the live blog posts that we created using this content engine system.

So this is the live blog post, and it's kind of meta because it talks about how to run GLM 5.2 inside Hermes agent, which is exactly what we got our agents to do. So they wrote a blog post, published it to our website, right, aoprofitborn.com, and fully created and designed this blog with a full tutorial of what they did together, which is amazing itself. And then we also, for example, created this game as we talked about earlier, using a similar sort of setup where your agents can work together and they just keep self iterating until eventually it's ready to go, right? Pretty amazing.

We also created a full agent operating system. So if we have a look at this, we've got our apps down here, we've got a terminal, we have our music player, we have paint. So this is literally an agent operating system, but it is created with the agents that we're working together as well. So when you look at this, you're like, wow, this can actually build stuff.

It can build operating systems. It can create amazing stuff. It can have it all working together. And it's super simple and easy to build.

Even like the graphics themselves look super nice. And this was with a full team of Hermes agents. What else can it build? Here's another example.

So this was like another game that we created, a bunch of other stuff here. Really amazing. Now you might say as well, like, okay, GLM 5.2, Chinese model, cheap on the coding plan. Is it actually good?

And you've seen the builds that we've created. So that helps as well. But if we actually have a look here and we compare this versus the Frontier, GLM 5.2 is a step above. So it's not really a toy model, it's Frontier class.

And you can see how it performs on Long Horizon TAS here, Terminal Bench. And it's not far off achieving the similar sort of benchmarks to Opus 4.8. So for example, GLM 5.2 scored 74.4 on Frontier SWE. Opus 4.8 scored 75.1, right?

So they were within like one and a half percent, which is pretty wild. Terminal Bench, MC Atlas, it's really good for agentic tasks. But the biggest difference here is like, if you look at the API costs, right? And bear in mind, like, you can just use the coding plan with GLM 5.2, but if you look at the API costs, like huge difference, huge difference right there.

So with this whole system that we've built inside the Kanban board, the old way was like, you'd research a keyword in one tab, write in another, lose your place, draft the whole post yourself, edit it tired. No one checks it and then mistakes go out. And then if you wanted to make another video, well, that would take a separate day. With this whole system that we've got here, we have a team of agents working together.

They work on a pipeline inside the Kanban board. We have all of our agents working, building amazing stuff, as you can see here. We can see what we've created. And literally all I need to do is just drop in a task and everything gets done.

So you can drop one topic onto the board, the agents take it from there. You got keyword to a writer, to an editor. Each is a separate agent profile with Hermes and you can hand them off cleanly. A judge scores every draft and sends weak work back until it actually passes.

It refused to publish fabricated stuff as well. So one of the other problems with AI agents is like they can hallucinate. So with this system, you've got the judge who actually checks for accuracy on the facts and sends it back. And then you can create a video at the same time.

So we've created the blog post and the video itself as well. And that was super high quality. And the end result is like you get a finished, checked post and video whilst you go off and do something else. And so this system just works beautifully together, especially if you've got an agent operating system like we do.

By the way, if you like this sort of stuff and you like, you know, you need content, et cetera, we've got an SEO agency. You can book in a free one-to-one SEO strategy session, link in the comments description, or go to goldie.agency to book them. And so the four parts of machine are the crew, which is a team of GLM 5.2 agents, the judge, which is the quality control agent. We have the board where everything runs on a Kanban board so we can see where the project is and where we're up to.

And then finally we have the shipping, right? So it can actually publish and deploy directly to our website as well. Now, some people want to say AI content is generic and boring and not very good, but the judge is the whole point here. It blocks generic, it blocks fabrication.

It sends weak drafts back until they're good. So you never publish anything bad. Other people say, well, running a team of agents, that must cost a lot of tokens. I see that all the time.

Actually, no, because you're using GLM's coding plan, which is pretty cheap. And so far I've not run out of tokens. We've built out an insane amount of stuff and we've not run out of tokens at all. Other people say, well, I need to be technical to wire up a team of agents.

It's literally a Kanban board. It's just a Trello board, right? You drop in a topic, the agents pick up the cards. You don't even need to move things around.

And the setup is inside the agent OS, which, you know, the whole thing is pre-built inside the airport for boarding. So this is basically how it works. You've got the posts in the video building, as you can see here. You want the live content, like so, which is the full article.

It's pretty amazing itself. And people will say like, what's the actual use case? So the use case here is like for SEO, or for example, for creating content for social media. And if we have a look at, for example, that website that I just showed you, and we look at the traffic of that website, you can see here that it's performing very well since we implemented processes like this.

So, you know, if you're thinking about setting up a new blog or a new website, now is a great time to do it, as you can see, right? SEO is working better than ever before because you've got better quality control. You've got better agents, you've got better outputs, right? And this all works together beautifully as a nice machine.

So these are real use cases. The same, for example, for SEO or for social media content, you can see an example of this video, super nicely edited. That would probably do pretty well on, for example, like Twitter or something. And that was just fully automated with our team of AI agents working together.

So that's the whole system. Now, if you want the full content machine inside your setup, the whole thing, the GLM 5.2 agent crew, the judge, the Kanban board, the auto-publish system, it comes pre-built inside the agent OS in the AirPuffer boardroom. So you get the agent OS with GLM 5.2 wired in, with the agents, the Kanban, the live previews, you get the content machine board ready to drop a topic into. You get four coaching calls a week with people running this in production, daily tutorials and a prompt library for every workflow and a room of 3,600 operators.

So someone's always online 24 seven. You can see the amazing results people are getting with this stuff. So for example, you can see here, we've got 183 pages of testimonials and wins from happy AI profit boarding members who are building out stuff like the agent OS themselves from the system that we give them, right? So so many people winning with this stuff.

So if you want to get the full system from me, you can get that inside here, just go inside the classroom and then you can get the new daily updates. As you can see, we've got the agent OS here with the last updated date. You can download the zip file for that. We've got new tutorials drop in every single day to help you as much as we can.

I answer questions with video tutorials daily inside the classroom as well so that we can help you. And you can see, for example, like Amanda building out these amazing agent operating systems themselves. If they can do it and I can do it, you can do it too. You don't need to be a coder to get this stuff done.

We also have four weekly coaching calls where you can get help and support in real time on a call. Inside the classroom, we've got all of my best trainings, including six week beginner to expert AI automation masterclass. You can also connect and DM with me personally. And then also inside the map, you can meet people in your local area who are using AI agents like you.

So you can meet other people winning with AI automation and learn and grow together with us. It's a great place to be and an awesome time to do it. So cheers for watching. Appreciate it as always.

I'll see you on the next one. Bye bye.