01What GLM 5.2 Is and How It Works

GLM 5.2 is a powerful open-source AI model from ZAI, a small AI lab in Beijing, China. It can build games, websites, videos, and tools from a single prompt. In this episode, Julian Goldie shows how he used GLM 5.2 to create full open-world games, 3D games, and video content in just a few hours.

GLM 5.2 works by taking a simple text prompt and turning it into code, graphics, and fully functional digital products. It is designed for automation, so it can power AI agents that work on tasks for hours. Because it is open source, you can run it locally for free if you have a good setup.

This makes it a private and cost-effective option compared to closed models.

02Main Features of GLM 5.2

GLM 5.2 has several standout features that make it a strong competitor to top AI models. Here are the main ones shown in the episode:

  • Open source: You can run GLM 5.2 locally on your own hardware for free, which gives you privacy and control.
  • High benchmark scores: It scores 74.4 on Frontier SWE, within one percent of Claude Opus 4.8's 75.1.
  • Design Arena leader: GLM 5.2 ranks number one on Design Arena with an Elo of 1360, ahead of Claude Fable 5.
  • Automation power: It can work with AI agents like Hermes to automate tasks, including SEO content and video editing.
  • Low cost: Building multiple games and websites cost about $80, far less than Claude's $200 monthly coding plan.
  • Fast iteration: You can create impressive results in a single morning without months of work.

These features make GLM 5.2 a versatile tool for developers and creators.

03How to Get Started with GLM 5.2

Getting started with GLM 5.2 is straightforward, even if you are not a coder. Here are the steps Julian shows in the episode:

  1. Choose your access method: You can use GLM 5.2 through Hermes agents, the CLI (command line interface), or both.
  2. Set up your environment: If you want to run it locally, make sure your computer meets the hardware requirements for open-source AI models.
  3. Plug into Hermes: If you have the Hermes agent, connect GLM 5.2 to it. This allows you to control the model through voice or text commands.
  4. Use the CLI for direct builds: Open the command line and enter prompts to generate games, websites, or videos.
  5. Create a team of agents: For larger projects, set up multiple agents, such as a researcher, writer, and judge, to work together on a task.
  6. Test and iterate: Let the agents self-iterate and give feedback until the output meets your standards.

Julian also notes that you can get the full setup, including prompts and memory systems, inside the AI Profit Boardroom community.

04How to Use GLM 5.2 for Building and Automation

In the episode, Julian demonstrates several practical ways to use GLM 5.2. The most impressive is creating a full open-world game with a single prompt. He also shows how to build 3D games and websites, all within a few hours.

For automation, he uses a team of GLM 5.2 agents to create SEO content and videos. The process works like this: you give one keyword, an agent does the research, another writes the post, and a judge checks if it is good enough. If not, it goes back to the team for improvement.

This loop continues until the output is polished.

Julian also shows how to connect GLM 5.2 to Hermes Jarvis, a voice-activated version of Hermes. This lets you control the model by voice and automate daily tasks, like creating songs, images, or videos. He emphasizes that you can build workflows for things you do every day, so you never have to do them manually again.

The key is to give clear tasks and let the agents work together. You can preview everything in one dashboard, making it easy to manage multiple projects.

05Benchmarks and Results in the Episode

Julian shares several benchmark results and real-world tests in the episode. On the Frontier SWE benchmark, GLM 5.2 scores 74.4, close to Claude Opus 4.8's 75.1. It outperforms GPT 5.5 and Gemini 3.1 Pro, which are not even in the race according to the charts.

On Design Arena, GLM 5.2 ranks number one with an Elo of 1360, overtaking Claude Fable 5. Julian notes that Fable 5 was only taken down a few days ago, yet GLM 5.2 already surpasses it.

In side-by-side tests, GLM 5.2 sometimes beats Claude Opus 4.8, especially in design. For example, GLM 5.2 created a more interesting liquid bowl animation and a nicer website layout than Opus 4.8. However, in other tests, Opus 4.8 clearly won.

Julian also compares costs: Claude's coding plan is $200 per month, while building all the examples with GLM 5.2 cost only about $80. This makes GLM 5.2 a more affordable option for heavy use.

06Tips, Limits, and Mistakes to Avoid

Julian offers several tips for getting the most out of GLM 5.2 while being honest about its limits. One important point is that GLM 5.2 is not as good as Claude Opus 4.8 in every area. He agrees with critics who say Claude is still better, but he stresses that GLM 5.2 is close and getting better fast.

Another tip is to use GLM 5.2 with AI agents instead of switching between tabs. This saves time and lets agents work as a team. Julian warns against trying to manage everything manually, as that slows you down.

He also notes that benchmarks don't always reflect real-world performance. That is why he tested GLM 5.2 by building actual products, not just looking at scores.

One mistake to avoid is not setting up feedback loops. Without a judge or review step, your agents may produce lower-quality work. Finally, be aware that running GLM 5.2 locally requires a good setup, which may be costly upfront.

07Who Should Use GLM 5.2 and Why It Matters

GLM 5.2 is for anyone who wants to build digital products quickly and cheaply. Julian shows that even non-coders can use it, like Ethan, who is 61 and not a coder. It is also ideal for content creators who need to automate SEO articles and videos.

Why does this matter? Because GLM 5.2 proves that open-source AI can compete with closed giants like Google and OpenAI. This is a big deal for competition and innovation.

Julian points out that it is shocking to see a tiny Chinese lab beating Google on benchmarks.

For users, this means more choices and lower costs. You can run GLM 5.2 privately, which is a major advantage for businesses with sensitive data. It also shows that the AI landscape is changing fast, and staying ahead requires trying new tools.

Whether you are a developer, marketer, or hobbyist, GLM 5.2 offers a powerful and affordable way to automate and create. It is worth testing to see if it fits your needs.