Original video in Chinese.
Key Takeaway
- Core components of the AI OS: OpenCode (open-source Claude Code) + newtype Profile plugin (multi-Agent orchestration + 5 Skills + methodology/workflow) + personal content repository (notes/scripts/Newsletter), fully covering daily thinking and creation.
- Mobile remote access: a Discord bot connects to the Mac server at home, so when I’m out I can invoke the system like chatting (look up docs/search/edit articles), with personal memory/methodology always on hand.
- Implementation and value: I used AI to help write the Discord bot and startup script, and it runs with a simple double-click; this is real infrastructure, not an app, and it doubles my content production efficiency.
Now, I can connect to my AI operating system from my phone too.
This system runs on the Mac in my home. When I’m out and about, as long as I open Discord, I can talk to it just like chatting with ChatGPT. Looking up docs, searching the web, editing articles—it quietly handles all of that in the background.
Let me first introduce this system, since many newcomers may not be very clear on it. It consists of two parts:
First, OpenCode plus the newtype Profile plugin.
You can think of OpenCode as the open-source version of Claude Code. And the newtype Profile plugin is a multi-Agent orchestration system I customized specifically for content creation and daily thinking. It contains my methodology, my workflow, and also has five Skills I created specifically, so this whole thing is the framework of my AI operating system.
Second, on top of that framework, I connected my content repository that I’ve built up over several years.
All of my output is in this repository, including notes, video scripts, and Newsletter. For AI, these things are both references and constraints. Only then can it become my system.
This entire AI operating system has already fully taken over all of my daily output. But there’s always been one limitation: it could only run on desktop. When I went out, I could only use Gemini on my iPhone or iPad.
This split left me dealing with two systems—OpenCode and Gemini—which stored two separate sets of data and two separate memories. That goes against my overall principle.
I believe that, for users, the most valuable thing in the AI era is your memory assets. You have to hold onto that tightly and accumulate it over the long term.
I specifically added a memory system to the newtype Profile plugin before, splitting it into two mechanisms, long-term memory and short-term memory, all saved as Markdown documents, precisely for this reason.
Fortunately, a couple of days ago, this problem got solved: Discord on my phone and OpenCode on my Mac were fully connected. And all of this was made possible by two things:
First, OpenCode server mode.
Once this mode is enabled, it wraps OpenCode’s core functions into a backend HTTP server. You can call OpenCode directly through a port.
Simply put, it turns your OpenCode into a server, with no need for front-end interface operations; everything is handled in the background.
Second, the Discord bot.
This bot is basically a mouthpiece. On one hand, it talks with me on the front end, meaning inside the Discord channel; on the other hand, it transfers data on the back end, meaning on the OpenCode server side.
In other words, the bot receives my request, hands it off to OpenCode to process, and then brings the result back to me. Just like many people in the workplace—they pretend all the credit is theirs, when in fact it’s all from the team behind them; they’re just the interface facing the boss.
Creating a bot like this is very simple. I had AI write all of it for me. I only needed to go to the Discord developer backend, create an Application, check all the necessary permissions, copy the Token, and that was it. I let AI handle everything else.
To make it easier to use, I also had AI write a script for me. After double-clicking it to run, it automatically starts OpenCode server mode, and then starts the bot.
Before, when I went out, I could only use generic AI, which was disconnected from my system. Now it’s different—my methodology, my memory, and my workflow can be invoked wherever I am.
This is the AI operating system I want: not some app, but infrastructure that follows me around.
With the help of this system, my content production efficiency has at least doubled. I use this every day, and I recommend it to everyone.
OK, that’s it for this episode. If you want to learn about AI, want to become a super individual, and want to find like-minded people, come join our newtype community. See you next time!