Claude Code + Obsidian = Best AI-Driven Knowledge Base

Hello and welcome to my video on using Claude Code and Obsidian together to create a an AI assisted knowledge base.

Uh now I am a huge noteaker.

I've been taking notes for a long time. It kind of feels weird to say it, but um because I feel like everybody takes notes, but I've been using it as a productivity tool for like 15 16 years.

I've gone through a bunch of different systems of note takingaking um that you know were the the note-taking system dour the GTD back then if you if you all remember it getting things done I think that's what it was called um and I've used so many different tools like Evernote like bear like uh whatever else really there was and kind of landed on this solution of using obsidian and claude code and both of those two these two tools are both very important in this system and I can't really use one without the other.

Now, Obsidian, uh, Obsidian has a really good reputation.

There's not really much to say about it. It has great plugins.

Um, Cloud Code can understand some of those plugins and be able to like create stuff like queries for uh data view, the data view plugin. Uh, or tables that you can then use advanced tables for.

Um, and uh, Obsidian is just plain Markdown.

Uh, I have the I have like several different vaults that I sync.

And so that means that I can like view the same stuff on my phone, on my iPad on my computer, wherever I want to or need to.

And it's great and and it has a fantastic view for these things.

I mean this this looks great. U being able to read markdown this way. And cloud code does a lot of things for me.

Cloud code is like my I want to say user interface.

This cloud code is how I use my knowledge base u for like if I need to do some research, if I need to do some analysis or whatever else.

and Obsidian is more like the consumption part of the knowledge base.

Uh now I have some views on like how notes how I want to like structure my own notes and what's great is that with cloud code you're able to figure out that system for yourself however you want to do your note takingaking this you can use cloud code for that.

So um just as a heads up we're going to, go, through, uh, my, system, real quick.

Uh I'm going to show you how I use my re the research tool tools for Claude and then I'm going to start a completely new knowledge base uh with Claude.

Um now Claude takes some time to like process things. I'm not going to be speeding it up because I don't just not going to.

So there might be some waiting around but that's kind of like what it is and I think it's kind of important to to notice it. So one thing is I set up my notes for mostly research and for like ephemeral notes notes that um I will be archiving notes that I will go out of date very very fast and I believe that's like all notes for me any note-taking system that I've had after a certain amount of time I basically have to declare like a a note bankruptcy and start over and this system is built with that in mind. Um I I think having like a digital garden is actually or again digital garden or like the evergreen notes I I don't feel like that really exists.

Uh I see a lot of YouTubers talk about these concepts and it sounds great on the surface but then I don't see anybody using it for examples other than for like making a YouTube video and I don't entirely I just I don't really see that happening.

There's so many there's like knowledge and things change so fast that um having an evergreen note just doesn't entirely make sense to me and and even my own perspective on the notes changes.

So that means I feel like I have to like have a new note and editing existing notes I'm not a big fan of uh because I like to have like the the the history of that note. So I I'll show you what I mean. So let's let's go through this.

So I have a image folder here.

This is where Cloud Code downloads any relevant images.

I have an inbox uh folder.

This is where I put my uh what I call the human written notes. Uh so for example, I put in this quick note for rip grip.

Um but I was going through the f ofmpeg uh uh ASM lessons and I put down a couple of notes here uh that don't really mean anything to be honest but uh I wanted to showcase this.

I don't use the handwritten or the human written notes as uh that often, but it is an option here. And then I have a series of Claude Code commands and agents that help me process this data.

And I'll show show that off.

We'll skip over notes real quick. I have an output folder.

This is where I ask Claude Code or Gemini or whoever else to uh like put in an output. Uh for example, if I want to run an analysis of my notes to repository, that's where the output goes.

But I have a really cool command called feed for cloud code.

And that will create uh that will look at the 20 most recent notes and compile them together with like some summaries and whatever.

And uh I call this this is like my RSS feed for my own notes.

I pull this up on my iPad. I pull this up on my phone and I look at okay what have I been researching? Let me go read through it.

Everything's linked together.

Um, and you you know because of Obsidian you get even like previews and everything.

This is great.

Like this is this is just like my favorite feature of this and that is how I consume majority of my notes is by having this feed uh because I typically just need to look up stuff that I looked up in the past week or two. If I need to look up anything past that before that I always have command O to find whatever uh file I want to or I can ask cloud code to find it for me. Um, and that again that's like my favorite way of dealing of of, of, interacting, with, uh, with, my notes.

This is a research knowledge base that is constantly updating and that I can update however I want to.

Now, cloud code because of the way it works lets you structure things however you want to and if you're not unhappy with the structure, you just tell it a different structure and it'll move everything for you.

It's fantastic for that. So, then I have a research inbox and I'll go ahead and show you uh an example of this one.

So yeah, I'll look at that.

I was looking up kids sandals uh for my kids.

I was looking up Lord of the Rings Blu-ray denoising, uh revision history music tagging, uh terminal user interfaces, all these different things.

And I should have one for getting started with Rust. So my research inbox is essentially notes like this stuff that I want to like look up and research um and uh and and learn more about.

But my problem has been recently over the past few years is that whenever you look stuff up on Google, whenever you look stuff up on duck.go, the results are subpar.

They're not very usable.

You get a a Reddit post from six years ago that's irrelevant.

You get a Medium post from like three years ago that's irrelevant.

And then you get like AI slop, which I know ironic saying that because I use cloud code to generate my own slop, but it's my own slop in my own way that works for me. Um, and so I whenever I want to research something, I just create a new note and I tell it to research something.

So I had the getting started with Rust. How about we do getting started with Golang?

Um, here's what I know. And this is almost like a prompt for my cloud code setup. I have a cloud code uh command for doing research.

So it'll like take this note do the research, create a note, link it correctly, save it, and do all the stuff that I needed to do and then I can go through it.

So here's an example of how I work in in this space. So um uh I'm a software engineer engineer.

Uh I have knowledge of let's see we have Python, JavaScript.

This is what I've worked with.

uh a little bit of Scola. I've done some of that.

I've done some Go as well, but like um and then I'll just ask it here's what I want to know.

And you can make this as comprehensive as you want to. You can create a single sentence like I like I did for the Rust uh for the Rust note where I just said I want to learn it and this is what I know.

Um or you can make it like super comprehensive.

So here's what I want to know.

I want to know how to set up setting up a new web project.

Uh basics of syntax plus gotas uh best practices and uh what every new dev in golang should know.

Fantastic. Okay. Um so this is this is this is how I use it. Uh I then fire up claude in my uh in my notes repository.

Like I said, I have a bunch of like commands and stuff. So, I can just tell it to research and it'll do its job in the background.

So, while it's doing that, because I am not skipping around I feel like it's kind of important to see the the workflow here. Um, I'm going to go into my dotcloud folder.

I have that in my notes folder. It doesn't show up, for, Obsidian., It's, kind, of a, downer.

I think there's a plugin for being able to sync uh hidden folders and stuff.

Uh I haven't tried it out. I haven't used it,, but, I, can., Um,, and, we're, going to, go ahead and look at this research command.

So, Claude has these commands right here.

And it allows you to create commands for anything that you really want to do. Um, let's say that you like to uh I don't know, occasionally just archive all of your notes that are a month or older. So, you can create a command for that. Um, let's say you want to have a pro a special project set up and you need to create a new project in that project set up. Here, let's look through that notes folder I skipped over.

Um, you know, I have the the maps and OC's.

I don't really use those.

So I just archived it literally this morning, but I have this like project section where I have manga recommendations, how to do Roblox games and a game that I want to work on and and whatever.

So, I can create a new command that creates a new project that links it correctly, that will do some initial research, some initial notes.

It might even find notes. I can tell it to look at look for notes in my current notes repository and move them into this project so they're all together whatever you want.

And so that's what the command is.

It's a think of it like programming or bash scripting but in natural language.

You can kind of see that um we will be using the research synthesizer agent.

That's an agent I set up.

Um, we have this comprehensive long prompt for this command.

Um, that we wanted to leverage both Claude and Gemini, that we wanted to prioritize authoritative sources, um, that we wanted to like cross-link stuff and do all this all this stuff and it's written down here and Claude can do it.

Um, what's important about this is that this prompt is generated by Claude. So, I typically tell Claude what I want it to do.

Um, we might even do it once and then I tell it create a prompt from this or create a command from this, create an agent out of this, what whatever it is, and it'll create this comprehensive list then I would recommend like manually uh reviewing, but it does a really good job for creating prompts for itself to redo what it already did. And let's say it doesn't do it right, and next time you run it, you're like, "Yeah, do some research.

" and you're like, "Hey, uh why are you messing up this part?

" You can just tell it to fix it and update all of the relevant prompts with that fix.

So, one thing that I did this morning, actually, again, uh, was that uh, Claude was having a hard time uh linking together images for Obsidian because Obsidian doesn't use like the regular markdown image uh, reference tag.

Um, and I knew there would be an issue.

Now claude because tools change over time used to be able to do this correctly but then it started messing up.

So we should have it here.

Yeah there it is. Critical. So uh I asked it to fix it. It fixed it and I said, "Hey I need you to update the prompts to make sure that it saves it correctly." Um and uh you can kind of see that it actually I did this during the Rust logo uh or the Rust research is when this occurred.

Sorry to use that as an example.

Um, but it made sure to let itself know that this is how you're supposed to link it.

Don't link it any other way. Um, and if you look through like prompt uh guides and prompt uh tips, you'll notice that it will tell you mark things as critical, do like a correct and wrong and just be very explicit about it.

Great. So, Claude was able to figure that out for itself. So, yeah, we have this research uh command. Now they're in commands.

I have these agents. So I have a research synthesizer. So let's say I wanted to do some research, but I don't actually want to write down this markdown file.

I can just talk to CL claude and uh tell it to use the appropriate agent and it'll use this agent.

Now an agent is basically like a pro.

Wow, that's everything's prompts with AI, right? So it's basically a prompt for uh claw to use when it's doing some kind of research. So here you can see that it's critical for it to use Gemini CLI.

It actually has a quick description of how to use the Gemini CLI.

So, Claude is calling into Gemini CLI and then using those results.

Um and you might want to have these agents or personas.

I feel like personas is a better way of putting it like calling it or what to call it. Um, because you will have this uh research synthesizer that you want to be skeptical that you want to be comprehensive that you want to do this and this and this and this.

Um, but for example, if you have a uh here I have a MOC maintainer uh agent that I no longer use uh and a maintenance specialist that I do use. I want my maintenance specialist to not be skeptical because all it does is look through my notes. I want it to be precise.

I don't want it to think outside the box. I want it to follow instructions.

So, you can create these like agents and personas and uh hopefully I'll be able to show you here.

Yeah, there you go. um that Colin can pick up on and use the correct uh and use it correctly. So here we're gonna One thing I want to mention is that I have all the permissions turned on so that it asks me for every single link that it will fetch. Um you can turn this off so that it just like I think people call it the yolo mode basically that it just does whatever it does. Um but yeah look at this. So it picked up that I'm doing research.

is decided to use the research synthesizer um uh researcher that means that it'll uh prefer like documentation over uh secondary articles or Reddit or forums for its research and uh wow it is having a hard time here but that's what it can do and so whatever you're trying to do in your notes repository you can create a command and an agent for it you or one or the other it doesn't really matter which way you want to do Um, if you're doing uh programming, one thing you might want to do is have an agent that knows how to run code, right?

Like you can have a node runner agent that knows how to use the node CLI.

It knows what to run, where to run it.

Uh you might even ask it to, hey, uh, I'm taking notes.

Can you ext extract all of the code blocks and run those in node for me?

And it'll be able to do that.

So, whatever it is that works for you you can make happen. Um, one thing that I kind of it's kind of hard is uh people coming up with their own systems, right?

What do you do? What how do you want to how do you structure it so that it works for you?

Um, and I think that's a great place for like being able to set up a brand new repository for nodes and say, "Hey, make it zealcasten uh use the zealcasten method.

" and then you notice that you don't use certain things or you use certain things a little bit differently or use some things more and you're able to talk to Claude and kind of figure that out and uh um yeah fix fix things um fix things and change the flow here.

Okay, so we've gone over my current repository, how I use it, what I do with it.

Uh one thing we can do is run the inbox command and we can while all of this is running, we can look at our demo vault and I'll show you how to set set all this up.

So, we'll do inbox. I have an inbox command.

So, all the commands that you put in here, look at this. You have you can even list out all the agents.

You can do all these different things.

I have this inbox command. It should just process that inbox how I wanted to.

And I'll show you how to how this works.

Look at this. So, the re research synthesizer is already creating a new folder in my reference folder here um for Golang.

And it'll try to save the not save the note there. Um, now you've noticing that you might notice that I'm using this for software engineering.

Like I said, I've even used this to look up the best uh like kids sandals for my kids for for school. Um, and honestly it works a lot better than just going through Amazon and reading fake reviews or going through Google and seeing uh people on Reddit argue with each other.

It just kind of creates a list and and and it makes it a lot more like doable to compare like let's say five different types of sandals. uh rather than like the million that exists and you know whatever it is. Okay, we're we're going fast here.

It's fantastic.

Um,, cool., Yeah., Yeah., Yeah., All right.

This is the inbox processing.

Look at that. It's using the inbox processor.

I'm going to switch over here.

Research synthesizer. We're doing so well.

All right., Okay., We're, going, to, actually wait for these and then we're going to go ahead and start on the new new setup.

Um, yeah, with my notes structure here I want to mention this. We have the notes folder.

Um, I had things like that for fleeting nodes, learning baths, and all this other stuff. And I've noticed that I never went to those notes.

I never checked it out. I always used my RSS feed because that's what worked way better for me. So, I just got rid of it.

And then I told uh Claude to get rid of any reference to MOC's from all the readmes and uh prompts so that it wouldn't deal with them anymore.

and it doesn't and that's great and I just move everything into the archive. Then I have these projects.

These are like longer running research projects where it's not just a one-off note. If I wanted to really learn go, I would just create a new project as I call it. I have a few active ones here. Uh let's look at manga recommendations.

It's kind of crazy crazy.

It's kind of weird, but um I created a list of manga that I uh wanted to research for my kids to be able to read.

Um so I went through a bunch of lists.

I asked it to go through a bunch of lists to fetch the you can ask it to fetch all the age ratings and it'll figure out how to get them and then it compiled a list for me.

I selected manually from that list and u in these archived notes and then I told it hey collect all the selected manga from all the notes that I selected or that I read through and create a the selected manga collection as you can see. So that the witch had a talier. We see what the age rating is.

The format it's black and white because my kids wanted to read uh coloriz like color comic books and that's more of like manua like Korean uh comics rather than manga.

So that's important information for me. Uh has a status to read. I don't really care about that.

I can actually just ask it to remove it. We have Little Witch Academia.

Couldn't find a cover image for it.

Couldn't find one here. Here.

Here we have Sailor Moon. Fantastic.

that's fantastic for kids.

Really cool manga that I've never heard of in my life, but it's fantastic for kids.

So I'm I was able to do these kinds of this kind of research with it with like an active project.

I have a few archive projects or completed projects.

This one called human manual that I was kind of working on.

And then I have these like reference notes.

This is where like majority of my stuff goes is these reference notes.

Um, and there's just like different folders and different things that I'm trying to learn about or one-off research tasks like the Rust one, getting started with Rust.

Let's look at the getting started with Go.

So one thing it did was actually get the formatting incorrect.

So, this is where we can apply this uh this this system of like um back and forth with Claude.

So uh the formatting in the Fmatter is incorrect.

Uh fix it. and explain what how what uh what is wrong. Um that way that I know exactly what's wrong, but I can also tell it to update its prompt.

Did it get the logo again incorrect?

No it just referenced the logo that doesn't exist.

That's okay.

Uh what are you trying to do here?

I think that we don't want to have these dependencies.

Okay, remove the remove the dependencies front matter al together remove that key from uh any um any prompt prompt so that uh no new notes will uh contain this property. Uh, I I don't have the best way of communicating, but Claude just seems to understand me well enough.

So, we'll see what happens.

Um hopefully it'll work. I think it's because it's trying to like put links into the front matter, and that's just like not working correctly. Um, so we we can fix that. Um, and I left that in there on purpose. This was something that I noticed was wrong this morning and I figured let me not fix it so I can show you.

Cool. We got all that fixed.

We have the source generated correctly.

We're in the correct category.

incorrect tags. We have the type, the status, it has the incorrect date.

That's another thing we can fix. Um, again, left a bunch of mistakes in on purpose so we could go through it, but let's go look at that research. Um, so we know where to download and install it. We want to set up our first project.

Initialize Go module. I actually don't remember ever seeing this when I worked on Go because I haven't worked in Go in a long time.

And it'll tell me how it's set up. Cool.

Dependency checks on. Wow, I didn't know that was a thing either. Uh, we have the code syntax uh fundamentals over here.

Cool. Um, here also, this is this is like all over the place, but I like this.

Uh, so I'm going to go back actually here.

Uh, you also have the date wrong.

Uh, what happened and why?

Please make sure to update uh any prompts to reference the correct date next time a uh note is created.

Cool. So, while it's doing that, we're going to go back here. We have the syntax fundamentals variable types.

This all looks correct.

Yeah. Yeah. Fixed it. Cool.

We have functions. Cool. This is how you declare functions.

I actually didn't know that you could declare return values this way.

Again, been out of Golang for a while.

And it's at the right level for me.

I already know multiple uh programming languages.

I don't need to know what a string is. I know what a string is.

And so it it can just do this.

And I don't need to have an explanation of what an int is.

I don't need uh the short declaration I'm actually I'm familiar with, but I'm glad that it has like a hey, this is the long one and this is what people actually use.

Um, this is how you can do multiple variable declarations.

This is great.

So, I want to set up a web project with Jyn.

So, I've used Jyn before.

I hope it's not pronounced Gin. It look it is Jin, right?

Um, oh, look at this update.

It's updating the prompt.

It's updating the prompt. Cool. So, now we have the we have the prompt updated. But yeah, look at this.

We have everything we need for Golang.

Setting up the first routes.

I'm familiar with Express for Node.

And this very much looks like express for node and how you can set it up. Uh, okay.

So you pass in the context this way.

Oh no. I don't know what the star means.

Great. So, that that could be another thing that I ask it like, hey, I read my getting started with Go.

Actually, let's just do it. Um, I was I read the getting started with Goes. Um, I'm not familiar with the star syntax. BG C star gender context.

Add a new section. Uh add a new section explaining what this does and what it means.

Cool.

And I just kind of communicate with Claude throughout my day.

Not throughout my day, but as I do this research.

This is great. Like having the standard project layout with the error handling.

And this stuff may not be I can't believe it downloaded the mascot.

It's kind of funny. This might not This stuff might not be exactly correct, but I can recognize all the stuff that is correct.

So, I'll go through it and then on the bottom if I Oh, this is great.

I didn't even know that they would create this like based on your background.

Um, it has all the references to it. So, if I start going through it and something doesn't compile right or I'm not sure about something, I can go to the source material.

Now, it added the section for the context.

This is a fairly long note.

So I would maybe ask it to separate it.

But let's go ahead and see if we can find this um what did they call it?

Their pointer syntax. Okay, pointer syntax.

Okay, cool. So now I can read about the pointers and how they work and what they do.

This is fairly good explanation some examples of uh what it does.

Hey here it copies here.

Here we're modifying the original. Great.

I'm I'm glad to know these things.

So that's that's that's my research here.

Now for handwritten notes that finally finished up.

We have two new notes for ASM lessons and RIP GRP. I don't remember if I showed the these offs but we have these the ASM lessons here and RIP GRP.

And again found a mistake here.

Hey uh the original notes in inbox never got archived.

Make sure to archive them and add to your to the relevant prompts to do so in the future when uh processing a box.

Cool. So, we'll check that out what it does and we'll let's see.

Let's do command O. We're going to do ASM lessons.

Uh is this it? Assembly language.

This is it. Yep. So, it will the way I set this command up is for it to do a summary of my notes and to actually preserve the original content.

So, it this this note here looks as bad as I wrote it.

This is great. And then I have a summary.

And in fact, it already did a bunch of research for me for some basics to learn about it. So, it worked on top of my notes. And this is all stuff that you can uh set up. Um so how about we do that?

Um hey, I noticed you did your own research when processing inbox.

I actually don't want uh any additional research to be done when uh processing my inbox notes.

They should be formatted, moved, etc. but no new research.

Update update all relevant prompts to follow these rules.

Cool. So, it'll just do that for us.

All right. So, I think we're pretty much good here.

Why don't I talk about how we can how you can set this up for yourself?

Yeah. Look at this.

Look at this. Only no research. No research.

No web rearches. No content expansion.

It is being very adamant that this is the way you're supposed to do it only based on users's original content.

Fantastic. And there's a memorize.

I don't need to memorize it.

We're already updating all these things for us for research inbox only. What does it say here?

Let's just make sure that it didn't mess something up. Yeah. Yeah.

Okay. Okay. So, it's letting us know that there's in all of our prompts to do things correctly.

We do want all of that extra stuff for the research inbox.

We don't want it for the regular inbox.

So whatever your flow is, again, you can customize these things however you want to.

But isn't this cool? I mean, it did this.

It tagged it correctly. It put it into the right folder and everything.

So, imagine that you can just write notes and you don't have to worry about doing that 45 minute at the end of the week uh post-processing of all of your notes.

Where do they go? How do you tag them?

How do you search them? Are all my MOC's updated?

just throw that all away and just ask the ask uh claw to do this.

And uh some of these tags I might not be a fan of. So I can just ask it to retag my entire system like hey come up with a new tagging system. This this is kind of useless, right?

If if I felt that way.

So yeah. Cool. Now let's go look at setting up our own uh system.

So we're going to go from notes to a demo vault that I already set up. And when I say setup, I mean I just created it and I didn't want to create it uh separately.

So the first thing we want to do, so you're creating your own uh note-taking system.

Maybe you want to create a note-taking system that's based off of an existing system, existing method.

Um maybe not. That's up to you. But what I usually start with is creating a readme file.

Um because Claude can read this and be able to infer what you want.

So, what do I want my system to do?

So uh what we need to do is just describe um as if you're describing to somebody else what your note-taking system looks like.

So, I'll say this is a knowledge base uh knowledge base uh for all of my note takingaking and research. I uh let's see topics uh topics in this knowledge base.

Um let's let's start with something that um let's start with I don't even know what topics.

Let's just do the software engineering.

That's something I know well engineering.

I also want to have photography in here. Um and I also want to learn about uh gardening and plants. Uh we'll do format.

This is going to be very rough because we're going to ask Claude code to rewrite even the readme. So we'll do format uh utilizes ztle caston method. I think that's how it's spelled.

Even if it's not, that's fine.

Cloud code is can figure this out and we want to use zle caston method.

Uh obsidian first markdown.

Um uses a lot of tables.

utilizes Mermaid for graphs. If you haven't checked out Mermaid Mermaid is a fantastic system.

Mermaid MD, let's let's do this. Let's let's just do Yeah, there we go. Mermaid uh mermaid is markdown converted into these kinds of graphs.

So, this graph gets converted into this.

And guess what? Clot code can read these extremely well. All right.

So we have this um and then you want to know let it know that it is AI assisted.

Uh this knowledge base knowledge base will be AI assisted utilizing cloud for research and note maintenance and uh note organization.

Cool. Now I'm going to go ahead and fire CL fire up claude. It'll be like, hey can I initialize here? Yeah. Yeah.

Just go ahead and create your claw.md.

So it's going to read the readme file.

It'll create a claw.md file that helps us do exactly what we need to do here.

Um, and uh kind of get us a good starting point.

This might take a little while.

So, in the meantime we can look at a few different things here.

Um yeah, let's uh So, I'm going to go ahead and actually go here. This is my old system.

I'm jumping back and forth because cloud just takes a while sometimes, and that's okay.

Um, I typically need this stuff async, not right away.

If I needed it right away it would work a little bit differently.

But I'm going to use my feed command here to create that RSS feed that I was talking about.

And I want to create a brand new one. So you can see a few of the things that I've been researching lately.

So that will take a little while.

Uh we'll get rid of this because we already did our inbox. Oh, by the way it does like Claude does cost money.

Um so, uh, it's kind of important to realize that while you're not paying for like a subscription for Readwise or Evernote or Bear or whatever else, you will be paying for cloud code.

And some of that research will cost you like more money, but I found it to be extremely worth it.

And if you kind of give yourself a an amount that you're willing to pay for a subscription for a service um I would just translate that into claude and use that money there.

Okay cool. So we have the we have the claude set MD setup. So we have okay settle casting approach obsidian compatible.

So it's kind of kind of cool. I'm going to go ahead and ask it uh update the readme to look nicer uh nicer and be more comprehensive.

So, we're going to do that so that we have a nicer readme. Again, like claude is my user interface for this note takingaking system, which is fantastic.

I I really like it. Um it'll do that update.

And next, after all of this runs, again it's kind of hard to do it in the video uh real time because sometimes it just takes a little bit longer we have the zetle casted method.

All right., And, yeah,, we, have, this.

, I hate emojis. That's going to be like one thing um that I'm going to change here.

Working with notes.

Important.

Do not use emojis in uh file names or uh our heading headings.

Uh headings. Yeah, I really don't like it. I really don't like it.

I'm, going to, do, that., Um, and, we're, going to actually Yeah, that cost 37 cents.

So, it's like I said, it costs money.

Um and if you just press up, I'm going to go ahead and uh try to reprocess that same note because we realize something that we don't like about it. We can just update it.

Edit.

Yeah, look at this. Remove emo emojis.

It already is doing that. Cool.

Um, now while it's doing that, you can actually fire up even more of these.

Um, we're gonna fire up Claude and I say I'll tell it uh read over the read me and create a folder and file structure uh following best practices for knowledge bases the zettle cast and method etc. So see the initial notes uh seed initial notes. So, I'm just going to say that um notes with examples on how to example examples of how to use use this notch.

So, we're going to do that.

Did it update the read me? Now, it did.

Yeah. Nice. Okay. So, it updated the readme.

, Now,, we, have a, more comprehensive read me. Look at that.

No emojis. Finally. Yay. I I just I do not like emojis.

Let's make this Let's make this bigger.

We're gonna have it take up half the screen. Um, so here's the overview.

It's NI AI the knowledge base.

You all saw saw that I wasn't that great at like writing down all this stuff, but this is um yeah, let's go ahead and create all those f folders. Um, we have the overview.

We're going to be using AI assistance for research enhancement.

All this intelligent organization, that's a great way to put it. We have the knowledge domains that we want to uh that we care about. We have the zillcaster method formats pre-orquisites.

Hey, you have to have um you have to have Obsidian. You have to be familiar with Markdown.

Here's the setup. Uh look, we're using data view tag wrangler.

I've never used that, but that's a really good one. So, we have these recommended plugins that we can install that Claude is going to be able to use uh for us. When I say it's going to be able to use it, it can't.

It doesn't connect with Obsidian, but because data view runs off of markdown files.

That's the beauty of Obsidian right?

Claude can update a data view query and a markdown file and Obsidian can run it against the data view plugin.

That's, fantastic., All right,, we, have, the notes how it's going to look like.

We have attachments, templates, there's a notes folder with all the different topics.

Uh we have the usage, creating these notes, linking strategies, AI assistance, best practices, and again you can update all this stuff.

And this is so overly comprehensive.

That's kind of like what people call AI slop, right?

Where AI takes one sentence and makes it into a paragraph, but it still communicates the, you know, what one sentence said.

So we can actually ask it to like pairs down and make it smaller to compress it.

Cool. It's creating a bunch of stuff here and we can even like look at it.

We have an index. So this is the MOC's maps of content.

That's what it stands for.

Great. I totally forgot. So if I want to use the Zcast method, I know there are different ways of using it, but this is like one thing I can do here. I have the knowledge base how to use this vault.

Cool. Does that go into my article?

Yeah. Yeah, there's a how to use this vault meta article. Cool. We can look at the zetle casten uh principles how it works.

Um it has a node structure.

Uh look at this. We got a mermaid graph over here.

Initial node find connections create links.

This this kind of like loop.

So I can read about that if I want to.

The tagging system uh what the tonomy is fantastic.

And yeah we have the knowledge grow growth how it's going to work and all kinds of cool stuff. it is not using uh the front matter formatting. So I can actually tell this to do that.

I think it's doing a lot of stuff here because it's uh it's it's seeding with those like basic notes and that feels kind of weird but for the purposes of this demo I think that makes sense. We have a photograph photography.

Looks like it already created uh all these notes here. Let me delete that because it's still going through its system and I'm not letting it finish.

But it's creating, look at this exposure triangle uh markdown file.

It's creating all these like uh demo notes for me.

And I didn't even tell it to do that.

I just told it, hey, create some see it with some initial notes. Um it's doing some basic equip uh basic knowledge stuff that you might want to know about like what the difference is between DSLR mirrorless and using a start smart smartphone.

uh the focus systems stuff that maybe not doesn't uh doesn't interest me right now because I told it to use the initial demo notes but it's um yeah I it's it's pretty useful.

So you know do you like this setup?

Do you not like this setup?

Um if you don't like it which you know I kind of don't like this some some things here.

Uh I'm going to let it finish and once it once it finishes through all these different tasks I'll be able to tell it hey utilize uh front matter.

Um I can ask it to uh double check on its research of the zetocasten method because I remember that there is the zettocasten what is it like the evergreen notes the fleeting notes the daily notes all that stuff and I wanted to use this use that format. Um so I'll let it finish up here. I just honestly I don't feel like waiting for it.

But let's let's try it again. Let's let's do this.

Um and this is I I I think this is like the the positive part of doing notes this way. Um that like we can change the structure. Um, I'd like you to research obsidian zeal caston setups and reorganize all of my notes to utilize that uh to utilize uh that structure.

Make sure to uh use obsidian markdown front matter for tagging in your metadata.

And once you're once you're done with restructuring and um restructuring and uh editing files update collab.

md with uh the correct prompt uh what do I directions so we can do that.

Um, and I'm moving a little bit too fast for it for the video, uh, here. But the other thing that we can do here, so it's going to re do some research.

It'll analyze the current structure, all this cool stuff.

We do want those commands, right?

We do want to to have those commands but it works the same exact way.

Oh look at this. It's using uh one of my uh I have a research analyst uh uh agent that actually works on my entire machine, not just in this repository.

So, it's utilizing that automatically.

I didn't realize it would do that, but it does, and that's cool. Um, and so we can what we can do is actually do the same thing for our demo vault and say, uh hey, I like you to add cloud commands and agents that would be would be relevant to this uh repository.

Now, here's a trick.

Look at this. So I have some research in my vault about this sub agents guide and I remember that there is a community agents examples uh page and there are so many different examples here right um check the community agent reposi repository for ideas.

So, let's see how that works.

So, we have our research analyst is still researching.

Uh, we have this going on over here.

I think I'm going to actually cut off the video and come back to see how it all ends up.

Look at this. So, it's doing a it is doing that web search for us.

It'll read through it and then set up all of this stuff for us. Okay, we'll be back in a sec.

Okay, so all of our cloud kod instances are done doing everything that they were doing.

Let's go ahead and check out the results.

So first for my existing notes repository, I created that latest feed uh notes feed project here.

We have uh well I created a lot of these files.

So I updated the readme, I updated the search setup, all of this stuff.

We have two new notes on learning Rust and Go.

We have the created ad. We have even the tags quick quick uh summary developer CLI tools and stuff that I was working on like assembly and language learning for me.

Um some active projects that I'm working on.

Uh some archival stuff that I was working on. So we have a lot of stuff here that we worked on.

And uh this is my little RSS feed of all the different changes and all the different research.

And this is kind of the way I would be going through it. So I load this up on my iPad. I would probably scroll through this. Maybe set look at search setup that I set up.

Uh maybe looking at the um front matter template that I created. And um we then have a getting started with go getting started with Rust.

Uh the articles that that uh this is the article that we created in this video.

This was for a my first attempt at this video was getting started with Rust. Great. So, we have everything we need here. Let's go look at what our new project looks like.

I think we can be done here with this one.

So, the folder structure looks way different.

So, what happened?

So, first we created sub aents. We have a knowledge architect, research synthesizer, and a vault maintainer.

Um so we have all these different things um that um all these different sub aents that are made specifically for what we need to do.

Uh we have a mermaid diagrammer a note connector. We have so many different things that we can do here already.

And then we have that cloud code instance that removed all of our notes and made this new structure.

So this should look more like a little casted structure.

We have the fleeting notes, literature permanent notes.

So we have fleeting notes, nothing there literature notes, nothing there.

Then we have permanent notes and this is where what we had uh we have previously a notes folder with everything in here and this is where all of those notes went to.

Now we have this great front matter.

Now earlier if you remember we had the dependencies problem on the in the front matter on my original repo.

We don't have that issue here. We actually have it all connected correctly. We have the aliases even.

Um great we have all the knowledge here.

Exposure triangle for the um photography stuff.

We have version control with git. We have rule of thirds and beyond. Great.

We have everything we need. We have MOC's.

We have a, gardening, MOC., Um, if, I, remember correctly, this is the one that I kind of stopped early uh so that we could try this out and and move everything where it needs to be. So we have all the MOC's we need here. I can go to my software engineering MOC and have um all the different fundamentals and all the different notes that we have here.

Some of these are not even created. So I can ask Claude to make sure that any note that's being linked gets created and researched.

So we have everything put together.

Um we have active projects.

I don't have any. And we have attachments.

So this this all looks good to me.

Um and this is how we can get started.

Now we have we have the sub agents.

Let's go look at what they actually look like.

So, we have the the hidden folder here.

Uh let's look at a mermaid diagrammer.

So, it will tell us what the different diagram types are that it supports.

Uh these are the the domains that we want to uh research over and diagram.

We have the different uh what is it?

The visual princip visual design principles.

We have best practices. Great. So, we have that.

Um, I don't have any commands.

That's fine. I don't need any commands yet.

But this is kind of it.

It doesn't really go beyond that. Um, if you want to do something special with your system, you can set it up here.

You can set up a command. Let's let's do that as an example.

uh set up a command to transform um let's see fleeting notes into permanent notes.

Um let's let's make sure that it says cla command and it can create that command for us.

Um but yeah this is essentially it.

Um you can go as deep as you want to.

You can go you can redo this entire thing.

Maybe you're not using literature notes.

So, you can just get rid of that folder and tell cloud to do things differently.

Uh, you can uh change how the MOC's are registered.

You can change anything that you want and you can just tell Cloud to do it and it will um it will move everything where it needs to be and tag it properly.

Here we go.

It's creating a template for fleeting notes.

Oh, for all the notes. Cool.

That's actually something I wanted to have.

Anyway so we have a templates folder now.

We have a fleeting note. What that looks like.

We have a permanent note what that looks like.

And now it's going to create a specialized agent for transforming those.

That's not exactly what I wanted but I'll be able to just go back and tell it, hey, make sure you create a clot command, not just an agent.

You're a note transformer. That's cool.

Oh, now it's going to create a fleeting node to demonstrate the transformation process.

So, it created a test fleeting notes.

Is it about test-driven development?

Yeah, it looks like it is about test driven development.

Couldn't find the Couldn't find the agent.

Huh.

I do have the note transformer.

I'm guessing that I need to actually reload Claude for that. Um, so there I mean you know, there is a little bit of this de debugging,, but, that's, kind, of, part, of being able to have exactly what you want out of it.

Okay.

And yep.

And and at some point soon, this will be create this will be converted into a permanent note. So we can we can wait for it. Oh, there we go. I just saw something change here.

Where was it?

Testdriven development shapes design. Cool. So it actually did all the research for us. It took that fleeting note, expanded it, explained everything.

Kind of like that research command that I had earlier um that I had uh that I had in my own repository.

But yeah, this is this is what you can do with it. This is the power of it. Um, and again I can go into uh another cloud code instance and do some other research or I can ask it to move things around, connect everything correctly, and it'll be it'll be exactly what I want it to have. So that's that's that's it for my video. Thank you for watching.

If you have any questions or anything any comments, obviously leave them and let me know.

Um but this is uh this is one of what I wanted, to, show, off, kind, of the, the, power of having obsidian for consuming the notes and writing uh like fleeting notes or like a what I called it the research inbox notes and using cloud code for everything else.

Um it's great because again I can change this into uh a GTD system if I want to go that way or some other system and it's uh it's fairly straightforward.

You just talk to clot code and it'll do what you need it to do.

Sometimes not perfectly, but that's totally fine because uh even if it doesn't do it perfectly, we can um iterate on that process. Thank you for watching.