Your AI Has Amnesia? Fix It With Obsidian + Claude!

Full video transcript

What you’re looking at here is I’ve just told Claude to plan my week for me. It’s gone out and it’s gathered my calendar and looked at my to-do items and it’s asked me a few clarifying questions, and I’ve given the answers to those questions. And you watch as it lays out a complete plan for me, building it into my weekly note here in Obsidian on the left. And just like that, Claude has written me an excellent weekly plan: what went well last week, goals for this week, priorities, projects, must-do tasks, scheduled work blocks that has helped me arrange around my schedule, and a perfect plan for the week to be highly productive.

In this video, I’m going to teach you how to add memory to your AI tool of choice, in this case Claude, and give it all the memory it needs about you, about your business, and about your life so that it can really help you maximize your productivity and supercharge your AI and what it can do for you. If you don’t know me, I’m Bruce Fleck, and I teach business professionals how to maximize their productivity using AI tools and not only improve their productivity, but maximize their impact as well.

What we’re going to cover in this video is how to add a memory system to your Claude program and put it in the same workspace, what I call an intelligent work environment. And by the end of this video, you’ll be able to learn how to set up the Obsidian folder structure that I use to really supercharge my Claude Code and be able to do all my projects throughout the day, all in one workspace. So let’s jump right into it.

The problem that I see, and very few people are talking about it — I initially called this the second brain dilemma, but it’s also called AI amnesia. The challenge is that with generative AI, you’re starting from zero memory at every beginning prompt. Every new conversation begins with no memory. Now, when Claude introduced projects, ChatGPT also developed projects later. You’re able to put in a few of your project files and some system instructions. It was a good start, but it only took you about halfway there, because AI has to interpret what’s in those project files — it’s not AI friendly. Secondly, anytime anything changed in those documents, you had to go delete it and upload a new one. It just wasn’t AI friendly, and the AI didn’t know what was important and what wasn’t unless you told it to. So it really only took care of about half the problem.

We should also look at our note-taking apps in the past. If you’ve heard of the concept of a second brain, using Notion, Obsidian, Evernote, or one of those note-taking tools — it really wasn’t a second brain at all, because it had no thinking capability. It only acted like a repository, a great big filing cabinet. And it felt like the more stuff you put into it, the harder it was to find stuff. So while it was a really good place to store a bunch of things, the bigger it got, the more complicated it became to use.

So myself and many other people have paired Claude with Obsidian as a memory and knowledge system. By adding Claude to it, we get this synergistic effect of not only giving Claude memory, but giving intelligence to our Obsidian note-taking file structure and second brain. I call this an intelligent work environment. It’s where you’ve got the power of AI and you add the memory of your note-taking system, so it knows everything about you and what you’re working on, about your business, and it combines it and magnifies your productivity in a synergistic way. You’re combining thinking and knowing all in the same place.

If you’re familiar with programmers, they use tools called an IDE, an integrated development environment — tools like Cursor or VS Code. They have an environment where they’re developing software, writing code, and their output is working software. For us who are business professionals, we’re knowledge workers, and our output is finished knowledge work. So we need to combine all of our knowledge work, put it in an integrated environment, and add our knowledge and combine the knowledge we can gather using AI to produce our finished work. This is just analogous to what developers have been doing for a long time, but we haven’t really had that in place for knowledge workers, and this is what I believe is an outstanding solution for knowledge workers to be able to integrate intelligence with their knowledge.

What does an intelligent work environment do? It does what a vault could never do alone. First, it’s got persistent context. It gives your Claude all the context information it needs. We had projects in Claude when using it from the chat app, but with Obsidian and Claude Code, you have files called CLAUDE.md files. You’ll be able to build in custom instructions and tell it your knowledge structure and everything you need it to do. Then it makes a very navigable structure — the vault is organized both for you and for Claude to be able to easily connect and synthesize, not just retrieve, information. And then you’ve got embedded AI: as you’re thinking about doing something, you just work with the AI to help you do it. This saves so much manual work. The key is making your vault both human friendly and AI friendly.

When you set this up, there are some mindset shifts you need to make. You’re going from a filing system to a workspace — not just a drawer where you store all your stuff, but a desk where you do work. It’s not only organized for you, but also navigable by the AI, set up in a structure that AI can work with. AI likes examples, and it likes structure. The third shift is that this is not a one-time briefing for a project, but an always-on briefing for your AI, so that everything is available and you have a permanent briefing. The AI always has all the context it needs to do its job.

I’m going to jump over to my Obsidian vault and show you how I’ve laid it out, just to give you an example of how this might work for you. Over in my vault, I’ve organized things from most dynamic to most static at the bottom. Things that are just coming in — ideas or other news feeds I’ve got running that are feeding in information — that’s my basic inbox. Then reflections and reviews is where I do all of my daily notes. I keep a daily note of planning the day and, at the end of the day, reflecting on it. It acts as my journal. Then I have an area for goals and tasks — laying out monthly or quarterly goals and then all the specific tasks I’m trying to accomplish. Then I have active projects, things I’m working on. If I’m producing content for this channel, that goes into content and communications.

Then I get into something that’s a little less dynamic — knowledge and learning. This is where I have all of my folders of knowledge I’m collecting. I’ve got a lot of different folders and subfolders on AI, but also business and coaching, personal development, health. I enjoy learning some history and some other things. PKM stands for personal knowledge management, psychology, research, etc. Those are some of the things I collect data on, and that becomes my knowledge area.

Next, I have my business memory. This is where I’ve got stuff stored like my brand guides, the clients I’m working with, what my offers are, what my products are. I have a product registry that is the single source of truth for the products I sell. Then I’ve got a personal memory, and you can put as much as you feel comfortable in there, whether that’s personal information or just basic stuff.

One of the new things I’ve added is the people network. I didn’t used to do this — in a note, I would just put the person’s name in the front matter since it’s for me personally on that particular note. But I find it really useful, especially for the AI, as I’m trying to link all this information together, to keep track of the people in my life. These are colleagues and friends, people I work with, as well as clients, and also the people I follow online and get regular content from. I follow a variety of AI experts, business experts, and personal development experts, and those are people who help feed me to get better over time. I want to be able to track what pieces of information they’ve given me — if they’re colleagues and I’ve got follow-up items with them, or they’ve got follow-up items with me, I can put it all in their people note, and it all links to the other notes throughout the vault.

I’m sure everyone is familiar with Wikipedia. The real power of Wikipedia is not the individual pieces of content — it’s that it’s linked to so many different things that it becomes a web of knowledge, not just individual pieces of knowledge. That’s one of the best things about Obsidian: it allows that wiki-type linking within Obsidian to link all these pieces of information together. So any piece of content I have from one of the people in my people network — let’s say this is Rich Schefren, someone I follow and take some of his classes — it tells me who he is, and then these are a bunch of the different vault mentions I have from various notes in my vault. It automatically links all that stuff together, and it’s super helpful.

That’s one of the key superpowers of Obsidian, being able to link that knowledge together, because that’s how our minds work — our brains are associative. That’s why sometimes when you’re doing something, you remember something else. If you get a smell, you remember an old time in your life, or you hear a song and it takes you back to some time in your life, or you see something and it reminds you of something else you forgot to do. Our brain’s memory is all associative, so that’s why wiki-linking is so powerful — it mirrors how our brains actually work.

Now, I know this looks a little bit ugly because I’m running Claude Code and not the desktop app. I’m actually able to open up a terminal window and just start Claude right away. I’ll open up even a new terminal window down here, and you can see it’s going to start up Claude Code, and then I can work on another project. I actually will run two or three projects at the same time and let Claude do the thinking and work, and it asks me questions and I approve things and give answers and let it do stuff. So I can be really productive doing a couple things at once. And I’m always able to see all the outputs over here in Obsidian — that’s the beauty and the power of this integrated work environment.

I’m showing you it on one kind of small screen right now. That’s okay — I have a much bigger screen I do my daily work on. If you have multiple screens or a bigger screen, you’ll find this really powerful. It just allows you to lay out your work and get a lot of work done without flipping back and forth between several different apps. I spend probably 80% of my day with Obsidian and Claude Code up in one big window, and then I’ve got my browser open to check on other things I need to on the web. Between those two things, I can get almost everything I need done in a day.

So I showed you my vault. Now I want to help you build your own Obsidian vault. You should be able to just go to obsidian.md, download it, and put it on your desktop. One of the things about how Obsidian works is it lays on top of all your local files. You just pick a parent folder, say “build me a vault,” and it builds out all the subfolders within that parent folder. Obsidian just lays on top and is able to view and manage all of those files and folders.

Obsidian natively works with markdown files, but you can also put PDF files, Word documents, PNGs and JPEG images, even JSON files in there. You may need to go to a community plugin to get a reader to work within Obsidian — I’ll go more into plugins, especially community plugins, in a separate video. But right now, I just want to show you some basics of how to think about setting up your vault.

One of the things that was really interesting is when I first tried to set up my Obsidian vault, I struggled a lot with trying to figure out how to organize it, and getting it organized is a really important step. This was before AI, so I just thought about how I wanted to organize things, watched a bunch of YouTube videos of people using Obsidian, took some of their stuff and adapted it. Every once in a while I’d want to change things, but it was such a pain to change stuff manually in Obsidian. Now, with AI, it’s so easy to make changes to your vault, because if you’ve got AI sitting there that can do all the work, you just tell it what you want done and it goes and does it — it makes all those tiny little changes. Just before this video, yesterday, I did a bunch of renaming and organizing of my vault to make it cleaner, because I realized I was going to teach you how to make your vault and I hadn’t done some housekeeping in a while. I took the time and had AI reorganize that vault for me. I didn’t make huge changes, but it was enough that it would have taken hours and hours to do manually — I was able to do it in about an hour back and forth with Claude.

So when you’re thinking about setting up your vault, think about a few different things. What do you want to capture? Do you watch a lot of YouTube? Do you read a lot of blogs? Do you listen to podcasts? Also, if you’re in meetings and capture meeting transcripts, or you’re in classes — whatever it is you want to capture, think about how you want to capture it and where it’s going to land. For me, that was up in the ideas and inbox.

Then you’re going to have tasks and to-dos. If you’re using some other system for managing your tasks and to-dos and don’t want to put that in Obsidian, that’s okay — keep it separate. These are just optional things to put in there. Also, journals and reflection — if you like paper journals and don’t want to put it electronically, that’s fine, but if you do want to add that, you can just say add a section for journals and reflection. All of us are working on projects of some kind, so you set up a projects folder, active projects, and when they’re not active, you move them to the archive, organized the way you want.

Then, for your business, job, or career, I highly suggest you have a business memory — this is where you tell it everything about your business and/or your job, everything Claude needs to know so it can instantly reference that for everything you’re doing. If you want to include some of your personal information, that’s a good thing to do too. I highly suggest you set up a knowledge and learning section, because knowledge is changing so fast — this is the whole concept of having a second brain. If you’ve been learning and collecting information and you put it into your knowledge vault, you always have easy access to it and you’re not starting from scratch. I highly recommend that in this AI age, when things are moving so fast, you set up your own knowledge vault — that’s one of the most important things about using a note-taking app like Obsidian. And lastly, if you’ve got content and communication, if you’re a creator, influencer, or expert of any kind, you’re probably going to be creating content and communicating with others, so that’s where this might live.

All of these are just optional. You can just tell Claude, “I want to make a vault in my Obsidian vault, I want to put it in this parent folder, and I want you to ask me questions about how I do all of these things and let me give you answers,” and then Claude will build the vault.

One last thing about Obsidian and vaults: it automatically sorts alphabetically. You can switch this to ascending or descending, but how I get around that is by putting numbers in front of each of the areas — if you just put a number in front of it, it sorts lowest to highest. That way you can put the most important, or how I’ve done it, the most dynamic things up at the top, and the most static things down at the bottom. That’s just a little trick in Obsidian for ordering folders the way you want.

Now, one of the most important things for Claude to work well is Claude looks for what it calls a CLAUDE.md. This is how it starts every session, by checking that CLAUDE.md. It doesn’t just check one CLAUDE.md — it checks the root CLAUDE.md, and then if you’re going into an Obsidian project folder, it will check the various CLAUDE.mds in each of these project folders. So for each folder, you can have its own CLAUDE.md — it will take some parent instructions, some overall instructions, and then some very specific folder instructions. That’s what makes it really powerful, being able to have overall instructions and then very specific instructions. It will really supercharge your Claude sessions and make them very accurate, which is what we want. You’re not going to spend lots of time giving it tons and tons of context, because you’ve already built it into the system.

Here’s what we’ve gone over today. We’ve got a structured vault so Claude can navigate your work — it’s also set up the way you want it, human navigable and AI navigable, so it’s easy for you to find what you want and easy and structured for AI to work with. With each of the notes, there’s a meta layer — a section above at the very top of a note. It doesn’t have to be there, but it’s very helpful if you have it. It’s in a format called YAML, yet another markup language. There’s nothing complicated about it, but it allows you to put various fields up there and gives the note metadata. AI really likes metadata to know what that note is about, what type of note it is, what kind of tags there are. This is where I put people information — if it’s a video from Alex Hormozi, then I’ve got his name in the metadata, and that automatically links it to the people file for Alex’s meta information. This used to be painful when done by hand, because I’d have to type it in for everyone. But now, if AI is helping you create the notes, you just tell it to add the metadata and it’s already there. This provides that permanent briefing — the CLAUDE.md files we talked about mean your Claude is never starting from scratch. It’s got all the context it needs, which makes this much more powerful for you.

If you like what I’ve shared today about how to build out an intelligent work environment using both Obsidian and Claude Code, I’m going to give you a prompt that will help you build out your structure, your vault, the way you want to — whether you’re brand new and have an empty vault, or you’ve got an existing vault and want to go through that process I showed you of having it interview you to organize your vault a little better so the AI can work with it more effectively. Either way, this prompt will work for that. Look for the prompt below, and I hope you’ll enjoy this.

Also, if you like the kind of teaching I’m sharing, look out for my book on Amazon called The AI-Powered Professional, where I teach you a pathway for growing your AI skills so you get more and more productivity, and more importantly, more impact in the work that you do. Thank you.