HomeVideos

This AI System Will Make You So Smart It’s Almost Unfair

Now Playing

This AI System Will Make You So Smart It’s Almost Unfair

Transcript

347 segments

0:00

I built a fully AI-powered brain that

0:02

runs across every part of my life. It

0:04

helps me run dozens of companies, and it

0:06

gets days of my life back. So, today I'm

0:08

going step-by-step on how you can build

0:11

the same AI brain I use to become part

0:13

of the 0.0001%

0:15

of AI users in the world, starting with

0:18

going pro.

0:20

Simple, get a pro membership for your

0:22

AI. It doesn't matter which AI, invest

0:25

the money. Free versions of AI give you

0:27

the worst version of the tool, slow to

0:29

answer. They're older models. So, as

0:32

soon as you pay, you've now become part

0:34

of the 0.3% of all users in the world

0:37

that have a paid membership. But, we

0:38

want you in the 0.01%.

0:40

So, let's get into how we actually do

0:42

that. Install the brain.

0:44

Every AI conversation forgets the second

0:47

you close it. Sure, they have some tools

0:49

for memory, but it's basic things. The

0:51

0.01%

0:53

give AI a permanent place to read from.

0:56

That place has to be your files, in your

0:58

structure that any AI can read your

1:00

digital brain. So, here are three

1:02

options you can pick from. The first is

1:04

a Google Drive. Most people have this

1:06

already, and it's kind of where it all

1:07

started. But, the problem is is that

1:10

it's pretty limited, and it's not

1:11

process. The second is Notion. This is a

1:14

cloud-based data structure system where

1:17

I believe that the way it's built is AI

1:20

native. Google is a basic digital

1:22

version of a folder system. Notion is a

1:25

modern AI-powered data store. So, yes,

1:28

it has the files, but it also has the

1:30

context, it also has the data. If you

1:32

want to go to the next level, and this

1:34

is what I use, I use Obsidian. The

1:36

challenge with a lot of these AI systems

1:38

is that you can't read what it's saved

1:40

in. You can't read the information. So,

1:42

Obsidian uses what's called markdown

1:44

files. It's free, and it's visual. So,

1:47

not only is it like a big wiki that sits

1:49

on your computer, but you can also see

1:50

how the brain forms and all the

1:53

connections and links to the information

1:55

within it. And the cool part is the

1:56

graph view within Obsidian is the only

1:59

one that actually feels like [music] a

2:00

brain. At the end of the day, it comes

2:01

down to personal preference. There's

2:02

literally hundreds of other tools you

2:03

can use. These are the ones I've used

2:05

and where I'm at today. So, if you're

2:06

not sure, just start with Obsidian,

2:08

download it, and just have the AI create

2:10

a second brain. It takes 5 minutes. Now

2:12

you have a brain that's installed, but

2:14

if AI reads it, there's nothing in it.

2:15

So, how do you make sure that actually

2:16

has the information in it sounds like

2:18

you? Give your brain an identity.

2:21

Without an identity, the AI is just

2:23

going to sound like some robot.

2:25

But to have the agent act like you, you

2:27

have to tell it how to act. That's the

2:29

identity. These three files will

2:31

literally turn your AI from a tool to a

2:34

digital clone that sounds exactly like

2:36

you. Number one, user.md.

2:39

Who are you? What is your role? How do

2:41

you communicate? What are the frameworks

2:43

you live by? How you present problems to

2:46

you? Two is the soul.md file. This is

2:49

for the AI. It's kind of wild, but you

2:51

can define the voice, the tone, the

2:54

values that essentially the AI will

2:57

adopt so that it can interact and

2:59

operate like you. So, the user.md file

3:02

is how you are, the soul is how it acts,

3:06

and then the third is the identity.md

3:08

file. That is who the AI is and what

3:11

they do. For example, I named my Kai. I

3:13

actually didn't even name him. I used

3:15

him for 2 weeks and I had him name

3:16

himself. So, for example, in my user.md

3:19

file, I included who I am. You know, I'm

3:21

a founder, I'm a CEO, I'm an author, I'm

3:23

an investor. My communication style,

3:26

things like being direct, fast, concise,

3:28

frameworks over fluff, and things I care

3:30

about, like freedom, execution, systems,

3:33

leverage, some of my operating

3:35

principles, theory of constraints,

3:37

minimum effective dose, the 1 3 1 rules,

3:39

a lot of different frameworks that I've

3:40

taught you, I put those in there so it

3:42

can operate from those principles by

3:44

default. Then my soul.md includes

3:46

instructions for my AI. So, for example,

3:49

talk like Dick. I'm direct, high

3:50

conviction, founder to founder,

3:52

challenge me with love. No hedgy words.

3:54

I don't like try, I don't like hope, I

3:56

don't like maybe. Get rid of those. And

3:58

then be resourceful. Try to solve the

3:59

problem before you ask me. These are

4:01

things that are in the soul file. And

4:02

then my identity file. Kai is part

4:04

coach, part chief of staff, part

4:06

accountability partner. These are things

4:07

that we define so that he can operate

4:09

from that place. Now, here's a pro tip.

4:12

Don't write these yourself. You can get

4:14

any AI to draft them for you. It's kind

4:16

of the coolest thing in the world. You

4:17

can ask it, "Hey, I need to create this

4:18

new AI, so can you interview me about

4:20

how I work, what I value, how I like to

4:22

communicate, and what I need from an AI

4:24

assistant?" Then have it draft the agent

4:27

files, and it'll know user, soul, and

4:29

identity files. And once you have those,

4:31

now we can just save them and then point

4:33

our AI to it so we can reference back to

4:35

it every single time. Now, look, this is

4:36

great for you, but what about the rest

4:38

of your team? That's why I built a whole

4:40

AI company operating system playbook

4:42

around how to integrate AI into every

4:44

department of your business. It's got

4:46

all the tools, the workflows, the

4:48

templates, everything your team needs to

4:50

use AI to get 10x the results. If you

4:52

want it, just find me on Instagram and

4:53

DM me the words YouTube OS, and I'll

4:55

send it right over. So, the brain knows

4:57

who you are now. Cool. Now, you probably

5:00

want to start feeding it information.

5:02

Not yet. Before you do that, you have to

5:04

make sure it's organized first. Wire the

5:07

brain. AI needs a way to clearly

5:09

identify the context of the brain. No

5:12

structure means AI just drowns in noise

5:14

and eventually hallucinates. I went from

5:16

60% accuracy in asking it questions to

5:19

85% just by setting up the right

5:21

folders. So, create these seven folders

5:24

inside your vault to get your brain

5:26

organized. The first one is people. You

5:28

should have a place where every person

5:30

in your life has a people file and has

5:31

all the context on those people. Two is

5:34

projects. What's one place where all of

5:36

your projects, timestamped, are listed

5:39

and located with all the context that it

5:40

needs. Third is decisions. What are the

5:43

decisions you've made in the past? How

5:45

did you decide? What the alternatives

5:47

are? Four is companies. This is any

5:49

information on the company I'm dealing

5:51

with, the research, the competitors,

5:53

etc. Essentially, it's like another

5:55

index of companies that I might be

5:56

interacting with. Number five is

5:58

meetings, cuz that's where a lot of the

5:59

decisions are made, that's a lot of

6:01

preferences are made, that's where team

6:02

and information about people is made.

6:04

So, we need a meetings folder. Six is

6:06

daily. I use this as a daily dump to

6:09

analyze and lock in every day daily

6:12

decisions that are three to five lines

6:14

that explain what happened throughout

6:15

that day. Then seven is knowledge. This

6:17

is where I'm extracting insights,

6:20

frameworks, quotes, anything that kind

6:22

of makes me me that I could reuse in the

6:24

future, I want to save it in this

6:26

knowledge folder. Now, here's a pro tip.

6:28

I have an eighth folder. It's called an

6:30

MOC folder, a maps of content. It's a

6:33

folder that includes files that is

6:35

summaries of bodies of work. So, for

6:37

example, I could have a youtube.md MOC

6:40

linking all my scripting frameworks, all

6:42

my hook libraries, all my thumbnail

6:44

task, all living in different folders,

6:46

but the MOC pulls it together. It's kind

6:48

of like consolidating information into

6:51

one simple reference file. Now, you

6:53

don't have to build an MOC from day one,

6:55

that's why it's a pro tip, but you build

6:56

them when a topic seems to get messy.

6:58

Think of it like this, your brain has

7:00

different parts of your brain that's

7:01

responsible for different things, motor

7:03

functions, smells, vision. That's kind

7:06

of what we're doing for a AI brain

7:09

that's going to have to do things on our

7:11

behalf. That's why we've created these

7:12

folder structures. Now, you can create a

7:13

bunch more over time, but this is where

7:16

we start. Now, the brain is organized

7:18

and it's ready to be fed information,

7:20

but how do you actually feed it? Then

7:22

the next step, you have to feed the

7:24

brain.

7:25

So, what you do is you connect all your

7:27

system, then you have your AI extract

7:30

all that information into the folder

7:33

structure that your brain is structured

7:35

in, because the brain only needs the

7:36

information it needs to use to make

7:38

decisions. There is a hundred thousand

7:41

data points that are being sent to me

7:43

right now from the light to the

7:44

temperature to everything. And my brain

7:46

is filtering out all the noise, and it's

7:48

only capturing the stuff I need to know.

7:51

This system works the exact same way

7:53

because it doesn't need the raw stuff

7:56

it's pointed back to it, but it does

7:57

need to synthesize and extract the

8:00

people, the decision, the knowledge that

8:02

happened inside those meetings. You just

8:04

have your AI do it. You have your raw

8:06

files where you point it to with your

8:07

connectors, and then you just have your

8:09

AI extract every day all of the

8:12

information based on the folder

8:14

structure within Obsidian. You just

8:16

improve your brain at the speed of

8:18

links. That's where we're at right now.

8:19

Now, I use Grainola to auto transcribe

8:21

all my meetings and then extract

8:23

decisions, commitments, and actions. And

8:25

the cool part is in Grainola settings,

8:26

you can set a custom template for this

8:28

exact prompt to do the extraction. Just

8:30

write this down. Extract from this

8:32

meeting: one, decisions, what did I

8:34

decide today, by whom, and why? Two,

8:36

commitments, who promised what, by when?

8:38

Three, preferences, how people work and

8:41

communicate. And four, key insights,

8:43

frameworks, strategic shifts,

8:45

non-obvious observations. Output as a

8:47

markdown, skip small talk. Now, Grainola

8:50

drops those outputs into the meetings

8:53

folders as, again, the date, meeting

8:55

name.md. Now, this is where things get

8:58

cool. The brain is installed. It knows

9:00

who you are. It's organized and it's

9:02

starting to feed itself. But the real

9:04

brain doesn't just hold things, it

9:06

rewires itself overnight. But how do you

9:08

get the brain to get smarter overnight?

9:11

Compound the brain.

9:14

Now, I'm going to process the brain the

9:16

way my real brain works. See, when we go

9:18

to sleep every night, it takes

9:20

everything that you were exposed to,

9:22

>> [music]

9:22

>> and it organizes, synthesizes, it

9:25

consolidates, it summarizes, it looks at

9:27

things that it needs to connect, and

9:29

then it even prunes things that it needs

9:30

to forget. And an AI brain works the

9:32

exact same way. And this is what

9:35

separates everything else we talked

9:36

about in the past, [music] storage, into

9:39

an actual brain. Here's two ways to

9:41

refine your brain. The first way is

9:44

manual. At the end of the day, if you

9:45

want, you can type in your AI a prompt

9:48

that's going to tell it to go into your

9:50

brain and run this process. Essentially,

9:52

you just say, "Read everything that

9:53

you've added to my vault today, and I

9:55

need you to find orphan notes, mentions

9:57

of people, projects, companies that

9:58

don't have their own files. You got to

10:00

create those files, you have to

10:01

consolidate duplicates, you have to

10:03

update the relevant map of contents with

10:05

new links. You have to flag anything

10:07

strategic for me to review for

10:08

tomorrow." It'll take 60 seconds, but

10:10

you'll wake up tomorrow to a cleaner,

10:12

more connected brain. The second, and

10:14

this is what I do, is automate it. I

10:16

actually use Claude's scheduled task to

10:19

run a cron job that runs the same prompt

10:22

I just gave you every night at 11:00.

10:25

And then, once you set this up, your

10:26

brain will constantly improve itself.

10:29

Every morning, I wake up and I click the

10:30

graph view within my Obsidian brain, and

10:32

I see it morph and evolve every night.

10:36

It's pruning, it's cleaning up, it's

10:37

essentially getting smart because the

10:39

more links that it creates, the higher

10:41

the signal is for context. So, when I

10:43

get my AI to talk to my brain, it comes

10:46

back with better answers. So, let's do a

10:47

quick recap. Number one, we went pro. We

10:50

decided to pay for real power in AI.

10:53

Then, we installed the brain. We figured

10:56

out where we're going to save it, and

10:57

then for most of you, you probably use

10:58

Obsidian. Then, we gave it an identity.

11:00

We created the files it needed to know

11:03

how to react and respond as an AI. Then,

11:05

we wired it to a brain, so it knew the

11:08

folder structure of how we're going to

11:09

save information. Then, we fed the

11:11

brain. And now, it gets smarter

11:13

overnight. You can't have an autonomous

11:16

agent unless it has context. Using the

11:19

brain is how we build that. If it knows

11:22

the meetings you were in, the decisions

11:23

you made, the people in your life, you

11:25

can say things like, "Send the invite to

11:27

John," and it knows who John is because

11:29

John has the most links in the brain for

11:32

the word John, and it knows John's

11:34

email, and it knows his cell number.

11:36

It's crazy how all this information is

11:38

available, but it's never reused in our

11:40

AI prompts. If you build the AI-powered

11:43

brain, you will start to buy back days

11:45

of your life because the AIs can

11:48

actually run your life. And remember, if

11:50

you want the AI company OS playbook for

11:52

your whole team, just DM me YouTube OS

11:54

on Instagram, and I'll send it right

11:55

over. And if you want to learn how to

11:57

build a $10 million solo AI business,

12:00

click here, and I'll see you on the

12:01

other side.

Interactive Summary

The video provides a comprehensive guide on building a 'second brain' powered by AI to automate tasks and manage business operations more efficiently. The process involves four key steps: choosing an AI-native storage tool (like Obsidian), defining an identity through specific files (user, soul, and identity), organizing the data into a structured folder system, and implementing a recurring process to compound and refine the brain's knowledge over time.

Suggested questions

5 ready-made prompts