Build a Second Brain in Obsidian and Codex

In today's video, I'm going to show you how to create an entire second brain just like this one that you can inquire inside of cold decks, for example, or cold cords, or whatever it is that you

use.

In today's video, I'm going to show you how to create a second brain in Obsidian using cold decks. It literally will take you like 2 minutes to set it up, and then you're good to go. So, over here on

the left, this is like Obsidian, what you would actually see. Um and don't worry, you don't have to actually use it like this. This is just so that it looks

like this. This is just so that it looks cool, basically. You can see all the

cool, basically. You can see all the connections and how it all works together. If we select on like a video,

together. If we select on like a video, cuz this is all of my YouTube videos and how they interlink. If we select on something like hallucination checks, if I select on that, we can see that it

creates tags, it has sources, uh it has like a little blurb, when it was created, etc. You can go actually through all your pages and see everything that it's created. And so,

this comes from all of my like YouTube transcripts that it's created. So, go in like here, let me see.

In today's video, I'm going to show you how to create a second brain in Obsidian just like this one using cold decks.

It's super simple. Literally here, you just set it up, ask it to work, and it creates these like connections between videos and between topics, which it inferred on its own. And it means that

then you can ask questions over here that are specific, and it will have this information to go and get. And it works so much better than rag.

[snorts] And so, what I'm going to show you is basically called LLM Wiki. It's a

pattern for building personal knowledge bases using LLMs, and it was introduced by Karpathy here. Um and he is like one of the main guys for OpenAI, Andre

Karpathy.

Um so, he knows what he's talking about.

And basically this here, you just copy and paste it in, and it's good to go.

So, let me cover how this actually kind of works super briefly, and then show you how to get this all set up. So, the

first question you may have here is, "Isn't this just basically rag?" And the answer is no, because for rag, what happens is like you give it some text,

"I like strawberries," for example, and it will like take chunks of this text.

So, I maybe like maybe straw berries. And then it will try and like

berries. And then it will try and like put these in some sort of rag database.

So, it's normally like three things, and and it plops it somewhere, right? And it

tries to then tokenize and use these tokens to relate them to each other. So,

that's how the example here is it's like trying to find relevant transcript chunks. That would be more like rag,

chunks. That would be more like rag, because then it's try to find things that are related like the actual contents of the what. Whereas here, um

in this LLM Wiki, let's check this. LLM

Wiki instead, this is more like it pre-reads the transcripts, and then builds an encyclopedia, and answers from that. So, it doesn't do this type of

that. So, it doesn't do this type of tokenization thing. It like researches

tokenization thing. It like researches into what you've given it, and then turns that into an encyclopedia, which is amazing. And again for this, you

is amazing. And again for this, you literally don't have to do anything. You

can just let AI do the whole thing. So,

if you look up here for this, what I've done is simply give all of the YouTube transcripts to um Codex by using

Scripty. So, my SaaS Scripty, we've

Scripty. So, my SaaS Scripty, we've introduced ScriptyKit. Um and it's

introduced ScriptyKit. Um and it's completely free to go and use for the transcription stuff. Um and basically,

transcription stuff. Um and basically, you can just say like pull the transcripts for all of my videos, and give it a channel, and it will just go and do it. And so, that's what it's done, and it pulls it in. And it pulls

it into a place called raw, which I'll go and show you in a moment. And so then you have all your actual transcripts in raw. That will then get ingested into

raw. That will then get ingested into your actual wiki when you ask Codex to do it. And then that's where it goes

do it. And then that's where it goes through and like creates all these nodes that it can then compare together and then that's how you can ask basically a question and it can go and look that up.

So if we come over to like the actual thing here, you can see AI Sas stack.

Now this isn't a transcript, this is just something that it's in filled from these sources here. So these are different videos.

And it say it says AI Sas stack is a set of services and agent blah blah blah current stack from the corpus working synthesis and then related contents. And

so these related things are other nodes that it's connected to. And we can imagine each of these is a node. You can

see then how they're interconnected. And

so you you can ask a question about AI coding agents and it will just know that these are the AI coding agents that you've talked about and this is the kind of patterns that

that are inside of this encyclopedia.

So, how do you actually set up yourself?

Simply go to obsidian.md

and then download it. That's it. Just

download it and set it up. Once you've

opened it up, you will have like let me see if I do a file.

Um we'll do a a new window. Let me see.

File new window.

And then here I have a new window and you go to let me see.

So once you get once you download it, you'll see something like this. If we go to Obsidian vault down here, you can see you have like different vaults. So this

is my YT vids. This is just like the default one that it makes. If you can't see anything like that here, you'll probably see a pop-up that's more like this one here.

Um where you can then create a new vault. You give it a name

vault. You give it a name and then you click on browse and you can go and select where you want this vault to live cuz it's like files, right? It

lives somewhere. So, here's in documents. This one's in YTVids, as you

documents. This one's in YTVids, as you can see here.

If I open up the side, you can see everything that's in YTVids. Now, inside

of raw here, this is where you're going to store all your stuff. And then once it ingests it, it's going to then store it, you know, and for example, pages here. You can see all of these different

here. You can see all of these different pages that it's created. So, how do you get that now that you've created your Obsidian vault? Well, it's actually real

Obsidian vault? Well, it's actually real simple. All you have to do is come to

simple. All you have to do is come to this other link. This will just be in the description. Select that link, copy

the description. Select that link, copy it, come over here, or you can copy all the text if you'd rather that.

Come over here, paste it in, and then at the bottom here, we've added a little bit from Nate Hacks for the vault actually. Um you are now my LLM Wiki

actually. Um you are now my LLM Wiki agent. Again, this will be in the

agent. Again, this will be in the description. Implement this exact idea

description. Implement this exact idea file as my complete second brain. Create

the agent.md.

If you're using Claude, this will be a claude.md.

claude.md.

Um blah blah blah. And then it'll go and set up this scaffolding. Now, this here, you don't really need to know what it does. Kind of like loads of things in

does. Kind of like loads of things in this kind of AI world we live in now.

You don't really need to understand what's happening under the hood. You

just have to understand kind of how it works, which is basically it creates nodes that relate to one another. And

then that's how it How do we get back?

There we go. Relate to one another, and that's how it searches. Now, once it does this, this is where then you want to like give it its specific goal. So,

if you're just creating a generic second brain, this is where you would say something like, you know, say here inside of Codex, you would maybe want it to be instead of I've got it in my

YTVids here, but I maybe want it just in documents because that's like my top-level folder. And so, inside then of

top-level folder. And so, inside then of documents, you could say create a second brain based on everything inside of my documents." That'll probably take like

documents." That'll probably take like an hour to run, maybe longer depending on how you're doing. But, for me here, we'll just create a YouTube videos one because it's kind of It's one of these things where

I like to have things compartmentalized because I notice when I try and do like a one-shot one brain for everything, you end up getting stuck in this problem of like you just have chaos. And you don't

know what like it just there's no way to figure out how stuff's working if that something's going wrong, and it just becomes an absolute mess. So, keep it compart- compartmentalized is, you know,

my general advice. Um but, then if you like have a query where you want to query say all your different vaults, you could just have like a documents uh chat here. So, let me let me show you

what I mean.

And say then if you actually want to like talk to multiple vaults at once, like all you'd have to do is say, for example, documents is my like top-level

folder. And then inside of here, I'll

folder. And then inside of here, I'll have, you know, whatever, one vault one, vault two, vault three. And if you have a query for all of these vaults, then

just open up a new chat inside of documents and then ask a question, and it would just know to come and use these basically.

So, for you to actually get this kicked off, for me, I used YouTube transcripts, and I'll show you how to do that. I'll

also show you how to get some YouTube pages or general web pages. So, I just threw all transcripts into here, your my second YouTube brain, and then it just went and got it all. And within 11

minutes, it takes it all in, creates all these pages, adds all the links. And

then when you ask it questions like, "What are my, you know, recurring arguments? Like, what kind of themes and

arguments? Like, what kind of themes and stuff?"

stuff?" It gives these um examples here. So, how

do how would you actually set this up?

Well, if you want to use YouTube videos, I'll show you that in a moment. But, if

you want to just get like web pages, use Obsidian Web Clipper.

Um you can see here obsidian.md/clipper.

And basically, all you have to do is go to a web page that you want to add in.

So, say this one here, right? Once

you've added the Chrome extension, you just select up here. And then you can just say add to Obsidian. Um or you can save to file, or copy to clipboard. But,

you can see that it's creating a page here with a title, source, description, etc., etc. And then if you just click add to Obsidian, it'll then add it to um I think to

clippings. If you go to settings, you

clippings. If you go to settings, you can change the location that it goes to.

So, select on this template here.

And then inside of here, da da da da da, location. So, what do you want it to do?

location. So, what do you want it to do?

Create a new note. What's the the name, location, and um which folder you want it to go into. And then also, you can change the content and stuff. But, I

mean, you're probably just going to do all of them in one shorter. So, for

YouTube videos, if you come to scripty.ai and connect your Scripty Cloud, basically, what it'll do is once you sign up, you can get an API key. You

can use that API key. Um

and then whenever you ask stuff like transcripts, write scripts, title generation, etc., it will use Scripty.

So, for this one, we're just using pull transcripts, basically. So, once you

transcripts, basically. So, once you click connect to Scripty Cloud, and then connect it like this here, it'll give you your API key. You can

just copy that, and then paste that directly inside of here, and then you'll be able to go and get your YouTube transcripts. Otherwise, it's kind of

transcripts. Otherwise, it's kind of problematic to see cuz to get YouTube transcripts, you need to like go to, for example, a video. So, I'll go here, open me up. Um

me up. Um There we go. So, open this one.

And then, of course, you'd have to go down here, get transcripts. Or you could have if your scripts are somewhere on your computer, you could just get that by default. Then you can copy all of

by default. Then you can copy all of that. But, that's a bit of a nightmare.

that. But, that's a bit of a nightmare.

So, once you have all of your scripts saved, so I have this other chat here.

Move that. I have this other chat here where I just said I initially got it to price it up, but it thought I was trying to get a quote.

[clears throat] [snorts] So, here I have my other like chat where I was working on actually pulling in these YouTube transcripts. And basically

just gave it my URLs and said, you know, pull them in. And it only pulled in a few just for testing.

And so, here's my my other chat where basically I just said download all my transcripts and then move them into raw.

So, that's if we open this sidebar here, you can see raw. This is

where you just want to have your raw stuff in here. So, I have transcripts, AI look. So, I've kind of split them up.

AI look. So, I've kind of split them up.

But, basically you put your raw stuff in here. And then once you have all your

here. And then once you have all your raw stuff, that's where you want to go back then to your LLM Wiki idea. And

this is where you can just say ingest them. So, ingest the raw

ingest them. So, ingest the raw transcripts into here. You are my second YouTube brain. And then that's it. Boom,

YouTube brain. And then that's it. Boom,

done. Ready to go. And then you can just ask it questions and you're sorted. So,

to recap, literally download Obsidian, then just grab this link, paste it in, and then just tell it what you want it to do. Give it the files in raw that you

to do. Give it the files in raw that you want to add in. Tell it to ingest them and it becomes your second brain. It's

that simple. So, go check out the links below if you want to get started with AI coding and you want to kind of community of people and ask me questions, come and join my Skill Community, skill.com/lukem

build with Luke where you can get my Codex Masterclass, Claude Code Masterclass for SaaS, everything you need to get started and actually start building for real with AI. Anyway,

thanks for watching and I'll see you in the next video.