China's NEW GLM 5.2 AI Super Agent is INSANE!

**Julian Goldie** (0:00)
Today, I'm going to show you how to build and automate anything with Glm 5.2. Now, this is a super powerful way to use a frontier level AI agent. I mean, you can see all the stuff that we've built right here. I mean, this is something that we literally built just in the morning for fun. And it's actually super impressive in terms of you look at the graphics, you look at the way it's been built, you saw the other game that we created a second ago, and you can use this to build and automate anything. I mean, look at the quality of this stuff. Here's another example. Let's have a look at some other stuff that we built here. So this was pretty interesting. This is like a full 3D game, as you can see, and it's absolutely amazing what you can do. I've just been playing around with it this morning, super fun and easy to build. Here's another one. All sorts of interesting things. The other thing is, it's very visual. So if you look at the quality of it, it's like this stuff wouldn't have been possible a year ago. And also the other thing that blows my mind here is like, you can get it on the coding plan and we've not run out of tokens, even though we've been using it as much as we have for building all the stuff. Also, we've already plugged it into our AI agents, like for example, Hermes and it can build websites, it can build games, apps, tools, whatever we want using this full process. And I want to show you exactly how to do that in three simple ways today and what you can build with this. The other thing that I like is like, all this stuff that we built, we can always come back to it later and make it even better, right? So for example, this is fun to build, but we can make it better and better. We could build out, we could improve it, we could create something even more interesting with this whole process. We even built out a full operating system with many apps, like you can see. So this is literally an OS built inside an agent OS. And you can see the potential and the power of this. So how do you use it? How do you automate anything with it? What is GLM 5.2? Let's get straight into this. So GLM 5.2 is a Chinese AI model coding plant, and it's comparable with frontier level intelligence, like, for example, Opus 4.8. It's actually quite a big step up from GLM 5.1, and this comes from Zed AI. So this MIT licensed and open source is available on Hugging Face, which means people can run it locally if they have an amazing setup. And also it's comparable on benchmarks with other models like Opus 4.8, Gpt 5.5, and it's actually outperforming on pretty much everything. Gemini 3.1 Pro, which is wild when you think about it. This is, you know, the latest version, the most powerful version of Gemini's model, Gemini 3.1 Pro from Google and Glem 5.2 is outperforming it by a long way on every single benchmark. How mind blowing is that? So if we look at the benchmarks here, you got SWE. And for example, Glem 5.2 is right up there with Opus on SWE. It's outperforming Gpt 5.5 on there. On Post Train Bench is outperforming Gpt 5.5. On Humanities last exam is right up there with Opus 4.8. And again, outperforming Gpt 5.5. So if you're thinking, oh, this is not as good as chat Gpt or it's not on the same level as Opus 4.8, I think today is the day that all changed. And when I saw the benchmarks, it shocked me. But I'm kind of like, well, you know, I wouldn't believe it until I see it. But then I built all this stuff of it, as you saw earlier, and it is pretty amazing what you can build with it and how it works, right? Benchmarks are one kind of proof, but here's the other. And here's all the stuff that we built with it. And it's super fun and easy to build as well. So it's so cool. It's really, really interesting as a time to just go off and build things. And I think with this coding plan, it's even better. So how do you actually start building stuff with it? That's probably the question you're asking right now. I'm going to show you several ways and what you can do with this and how it works. So there's three ways that I look at this in terms of building an AdFim. Number one, you've got GLM 5.2 as a dashboard. So you can use, for example, the CLI in the same way you would use CLI from Claude. And then you can start building that. You can type in what you want to play in English. It can then start using the code and it lands inside your workspace. And I think that's really the most powerful and best way to use this. So for example, if we go to GLM 5.2 inside our agent operating system, this is what it looks like. This is how you can use it. And it's really easy to just chat with it inside here and then see everything that you've built inside the workspace. So you can come back to everything you built. It's easy to see. And it's just like the same as using Claude. You just talk to it. You don't need to be technical as well. That's a great thing because you can just type what you want in plain English. It goes off and it builds it.