Original video in Chinese.
Key Takeaway
- AI is changing the way we learn and take notes. Tools like NotebookLM are changing the way we get knowledge from documents.
- The Basic Memory project aims to solve AI’s long-term memory problem by saving notes locally in Markdown and connecting AI through MCP for natural-language interaction.
- My AI note-taking system uses a “funnel structure,” made up of an AI note library (created and managed by Basic Memory) and a personal management note library (Obsidian).
- The AI note library lets AI automatically create folders and tags, while I only give instructions and filter out valuable notes.
- It emphasizes using AI’s intelligence and efficiency to bring advanced productivity into my personal system and widen the gap between people.
AI is completely changing the way we learn.
For example, chatbots that can search the web, and the more powerful Deep Research, have changed the way we get knowledge from the internet.
NotebookLM, which stands out on its own, has changed the way we get knowledge from documents.
Now, the way we record our thoughts, or in other words, take notes, is also about to change. Let me show you.
On the left is Cursor, on the right is Obsidian. I asked it to help me record this note:
The value of open-source models lies in localized scenarios. But in localized scenarios, the most important things are data and business-flow transformation, not the model itself.
After receiving the request, Cursor called an MCP and completed the note creation. On the right, you can see that based on the meaning of the note, it created a folder and also added the corresponding tags to the note.
Since it can record notes, of course it can also improve them. For example, when I have some immature ideas, I can do this.
I told Cursor to help me record and improve this note about how to improve AI product performance. You can see that it not only filled in what I was thinking, but also added two points based on its own knowledge.
Let me show another very common scenario. In daily life, we often encounter concepts, viewpoints, and the like that we think are valuable, but we don’t know their exact meaning. At that point, we can have AI help explain them and then record them.
For example, “the data center is the new compute unit” is Old Huang’s view. I asked Cursor to explain it and record it.
At that moment, Cursor first called the Sequential Thinking MCP to carry out multi-step reasoning and ensure the completeness of the explanation. When it thought the logic was complete, it began recording. In Obsidian on the right, you can see that it broke this sentence down in great detail.
What I just showed were only very simple demonstrations. In actual use, I also added a search MCP to make AI’s information-gathering ability even stronger. And the MCP that helps me take notes is called Basic Memory, which is the project I want to recommend to everyone today.
Hello everyone, welcome to my channel. Modestly speaking, I’m one of the few bloggers in China who can clearly explain the Why and How of AI. What I provide is more valuable than tutorials. Remember to follow. If you want to connect with me, come to the newtype community. More than 1,000 friends have already paid to join!
Back to today’s topic: Basic Memory.
This project mainly aims to solve AI’s long-term memory problem. When we chat with AI, a lot of things are talked about and then passed over and lost, which is a pity. So Basic Memory’s idea is to save them in the form of notes.
First, all notes are saved in Markdown format. This format is understandable to humans and also to AI, so it couldn’t be more suitable.
Second, all notes are stored locally, so there’s no need to worry about privacy leaks. And if needed, they can be handled with other tools, such as Obsidian and Cursor.
Third, the note repository and AI are connected through MCP. AI can automatically create notes and read notes. And because it uses MCP, it fully supports natural-language interaction, so you can just give commands directly.
Honestly, I didn’t think that much about it. I don’t care at all about things like long-term memory. My needs are very simple: when I’m talking with AI about something valuable, or when a new idea suddenly pops into my head, I want AI to act like a recorder or a little secretary, helping me record and helping me expand on it.
Installing Basic Memory is very simple. It only takes two steps, so anyone can do it.
First, run this command in the terminal: uv tool install basic-memory
Let me explain. uv is a fast, lightweight Python package management tool. When everyone installs open-source projects, you must have used pip before, such as commands like pip install and pip list. The advantage of uv is that it’s faster and more feature-complete. For example, it includes package management, virtual environments, and tool management. It has everything, so you don’t need to install a bunch of other tools.
There’s something similar called npm, except it’s for JavaScript, so I won’t go into that.
After we finish installing with the uv command I just mentioned, we can move on to the second step and paste these lines into the configuration file.
This uvx is part of uv, introduced just now, and is a subcommand. You can use it to run commands in Python packages. And it has the advantage of automatically creating a temporary virtual environment, which is very convenient.
I use it in Cursor, so I just pasted these lines into Cursor’s MCP configuration file and saved it. Then you can see this project in the MCP server list.
Basic Memory is very easy to use, just like in the demo at the beginning. Let me focus on my positioning and thinking about it.
Now I’ll have two note libraries: one is the original Obsidian Vault, completely created and managed by me personally. The other is the new AI note library, created and managed by Basic Memory.
For this AI note library, I let go completely and absolutely do not interfere. For example, the folder structure inside it and the tags for each note are all handled by AI. I only do two things:
First, give instructions. Whenever there’s something that needs to be recorded, I tell AI right away.
Second, filter the notes. The AI note library takes in valuable, worth-recording content from my daily conversations with AI. I regularly filter notes from it, or let AI help me choose, and then move them into the note repository I manage.
This is actually a funnel structure. The AI note library is the top layer. I want to leverage AI’s intelligence and AI’s efficiency to help me screen things once first.
These are my current thoughts, for your reference. AI is developing very fast right now, and both models and various tools are iterating at high speed. So we definitely need to use our brains more and think about how to bring these advanced productive forces into our own systems.
Believe me: this is how the gap between people gets widened.
OK, that’s all for this episode. If you want to understand AI, if you want to become a super individual, if you want to find like-minded people, come join our newtype community. See you next time!