01What Is Qwen 3.8 Agent OS?

Qwen 3.8 Agent OS is a way to use the open source Qwen 3.8 Max model (2.4 trillion parameters) as the brain inside an agentic operating system. You can build and automate anything you want by plugging it into your existing workflows and systems, like Hermes Agent or Claude Code.

In this episode, Julian Goldie shows how to set up this system. It's not just a chatbot—it's a full operating system for AI agents. You get chat, memory, tickets, builds, and custom workflows in one dashboard.

The model reads your business context from an Obsidian vault, so it doesn't start from zero every time.

The key idea is that you have a giant model (Qwen) as the brain, an agentic harness (like Hermes or Claude Code) as the hands, and a UI to switch between workflows. This setup lets you automate lead outreach, content creation, coding, and more.

02Key Features of Qwen 3.8 Agent OS

Qwen 3.8 Max is powerful and fast, designed for agentic tasks. It's open source and self-improving, so it gets better over time. Here are the main features shown in the episode:

  • 2.4 trillion parameters—one of the most powerful models in the world.
  • Million token context window—handles large amounts of information.
  • Autonomous coding—designed to code for hours alone, without babysitting.
  • Model switching—you can switch between models when doing different tasks, like using Qwen for some things and Claude for others.
  • Memory loop—every exchange auto-saves to an Obsidian vault, providing business context.
  • Work surfaces—includes chat, voice agent, wake word, Kanban boards, and workspaces.
  • Custom workflows—you can plug in open source projects like Open Design for ready-to-go workflows.
  • Group chat—have multiple agents in a group chat, tag them, and ask them to build things.

These features make it a complete system for AI automation, not just a single tool.

03How to Set Up Qwen 3.8 Agent OS

Setting up Qwen 3.8 Agent OS takes about three to four hours per day. If you want a ready-to-go setup, you can get it from the Airprofit Boardroom, which includes a zip file, setup guide, and video tutorials. Here's the general process:

  1. Get the Qwen 3.8 Max model—it's open source, so you can access it freely.
  2. Choose an agentic harness—Hermes Agent or Claude Code are good options.
  3. Plug Qwen 3.8 Max into the harness as the brain.
  4. Set up your dashboard with the UI where you can switch between workflows.
  5. Connect your memory system, like an Obsidian vault, so the model reads your business context.
  6. Add any custom workflows you need, like lead outreach with Hunter API or Open Design for websites.
  7. Set up your work surfaces—chat, voice agent, wake word, Kanban boards.
  8. Test the system with a simple task, like using the /learn command to create a custom guide.

You can do this even if you're not technical—the Airprofit Boardroom provides the exact setup with a zip file to install. Julian says many non-technical people have set it up successfully.

04How to Use Qwen 3.8 Agent OS

Once you have Qwen 3.8 Agent OS running, you can automate many tasks. In the episode, Julian shows several examples:

  • Learning and skills: Use the /learn command to give it a guide, and it creates a custom guide and skill locally.
  • Goal system: Plug it into a goal system, and it codes autonomously for hours. You can walk away and come back to a history of the goals.
  • Lead outreach: Connect the Hunter API key, type in something like 'SEO agencies interested in white labeling backlinks,' and it finds leads, drafts email campaigns, sends them, and manages inboxes.
  • Content creation: Use Hermes Oracle to pull the latest news, draft social media posts, and publish to WordPress with one click.
  • Design: Plug in Open Design (a GitHub project) to create beautiful websites without AI slop.
  • Group chat: Have multiple agents in a group chat, tag them (e.g., Claude), and ask them to build things.

You can also use Qwen 3.8 Max with Claude Code—switch the brain to Qwen, and Claude Code becomes the agentic harness. Then you can view all your builds, like games or websites.

05Results and Comparisons: Qwen 3.8 vs Others

In this episode, Julian shares results and comparisons for Qwen 3.8 Max. It's beating Fable 5 on many benchmarks, and it outperforms in some places Fable 5 and GPT 5.6 Sol (he corrects himself from GPT 6 Sol). There are side-by-side comparisons on goldiebench.com.

Julian also shows builds created with the system—a flight simulator style game, a Minecraft-like game, and full websites. The UI and design look nice, and he says it creates fun, easy-to-use stuff.

The system is open source, so it's self-improving and gets better all the time. This means you don't have to wait for a new version; it learns and improves on its own.

06Tips, Limits, and Mistakes to Avoid

According to Julian, most people make mistakes when using Qwen—like using chat.qwen.ai in a browser tab without a proper agentic system. That's messy because you have to re-explain your business every session, copy-paste answers by hand, and you lose all your work. There are no tickets, no builds, no voice agent—just typing.

And it's a closed-off model.

Here are tips and limits from the episode:

  • Don't start from zero: Use the memory loop so the model reads your business context.
  • Avoid losing work: The agentic OS auto-saves every exchange and makes it searchable forever.
  • Use multiple models: You can run Qwen, ChatGPT, and Claude side by side—pick one or two you actually use.
  • No need for Claude API: You can plug in the CLI instead of paying for the API.
  • It's not a developer's job: You don't need to be technical; describe your vision and use a ready setup.
  • Don't wait for a winner: Models change all the time. The operating system is built for swapping models, so it doesn't matter which model comes out—the systems matter the most.

Building the system yourself can take three to four hours per day, but using a pre-made setup saves time.

07Who Is Qwen 3.8 Agent OS For and Why It Matters

Qwen 3.8 Agent OS is for anyone who wants to automate anything with AI, without babysitting prompts. It's especially useful for business owners, marketers, and creators who want to generate leads, create content, and build websites or apps.

It matters because it solves the problems of using AI in isolation—like losing context, re-explaining your business, and not being able to find your builds. With an agentic operating system, you have a giant brain that reads your Obsidian memory, auto-saves everything, and works alongside other agents.

Even if you're not technical, you can set it up using the Airprofit Boardroom. Julian mentions over 207 pages of wins and testimonials from people who set it up without being technical. The system is open source and self-improving, so you can keep it updated and adapt as new models come out.

08More Ways to Use Qwen 3.8 Agent OS

You can extend Qwen 3.8 Agent OS in many ways. For example, you can plug in free models like Paperclip as sub-agents, with Qwen 3.8 Max as the CEO orchestrating the work. This way, the frontier model handles the thinking, while free or local models do the hard work underneath.

You can also use it for team coordination—have Hermes with Qwen plugged in, and all your other agents in a group chat. Tag them and ask them to build something out. It's a powerful system for orchestrating multiple AI agents.

The system is designed to be flexible. You can add new workflows as you find useful open source projects. For instance, Julian found Open Design on GitHub and plugged it in to create beautiful designs whenever needed.

This makes the system ready to go for any task you want to automate.