Best Skill to Learn in 2026 (AI Can't Replace This)

Most people think writing is about expressing themselves.

You have a thought, an idea, a feeling and they want to put it into words.

In my opinion, this is the starting point for almost all bad writing.

In this video, I'm not going to cover syntax or which word comes after which.

You can figure that out yourself. But

I'm going to cover the substance, not the style. This video is about the

the style. This video is about the secrets of good writing. It's about the underlying structure of how you can communicate and persuade people. And

it's also the unspoken rules that govern how ideas are accepted or rejected.

There's a pattern to it. From this point on, we're not going to care so much about what you want to say, but we want to care about what people hear.

I think the red pill truth about writing is that it has almost nothing to do with you the writer. Effective writing is not an act of self-expression. It has a little bit in it but it's not primarily

that. It is an act of applied

that. It is an act of applied psychology. It's a kind of specific

psychology. It's a kind of specific desired experience in the mind of another person that you are engineering.

The words are not the point. The effect

the words have on the reader is the point. Most people are focusing on the

point. Most people are focusing on the syntax and the words because of the transmission. But the masters of this

transmission. But the masters of this craft focus on the reception. This is

why a software engineer named Som Parik was able to get hundreds of interviews while most people can't even get one.

Recently, a scandal erupted in the startup world. There was a software

startup world. There was a software engineer named Som Parik who was accused of moonlighting by working for three or four different startups at the same time, deceiving them all. As founders

compared notes, they discovered he had used nearly identical cold emails just to get started. It wasn't because he was necessarily the best engineer, but he was pretty good. It was because he wrote

a damn good cold email. And that email wasn't good because it expressed who he was. It was good because it perfectly

was. It was good because it perfectly understood who he was writing to. He was

not really just looking to send a message. He's kind of like executing a

message. He's kind of like executing a program in the mind of his target audience. While the ethics of his

audience. While the ethics of his employment were debated, I'm sure everyone has their own viewpoints. A

different conversation started bubbling up, right? Which is the email itself was

up, right? Which is the email itself was a work of genius. It's not genius in the sense that it had some crazy words and writing. It was genius in a

writing. It was genius in a psychological sense and I want to deconstruct it.

So the emails that Parik sent to various founders had a very precise template.

Here's the core of it. Hey, I love everything about what company name is doing. I don't have many hobbies outside

doing. I don't have many hobbies outside coding. I'm not athletic, bad at

coding. I'm not athletic, bad at singing, don't drink, can't dance.

Building is the only thing I'm good at.

He would then list his technical skills, mentioning his experience in super lean teams and his ability to work across the stack. See, on the surface, this looks

stack. See, on the surface, this looks like a decent, you know, application.

But that's the blue pill reading. The

red pill reading shows that every line is carefully calibrated. It's like a psychological program that is running in their brains, especially in the brain of a startup founder. Think about it. I

love everything about what company name is doing. That's not flattery. Flattery

is doing. That's not flattery. Flattery

is generic. He's taken the time to mention the company's name for starters.

Most people don't even mention company name. In a world of generic copypasted

name. In a world of generic copypasted applications, the minute this line exists, it signals that he's done his homework. And it immediately shifts the

homework. And it immediately shifts the conversation from what he wants to what the owner of the company wants. The next

line is brilliant. I don't have many hobbies outside coding. Building is the only thing I'm good at. This is a master stroke. It's not saying that oh I'm

stroke. It's not saying that oh I'm awkward I don't know how to hang out with other people. It's just it's not any of that. It's a sign that look at me I'm so focused. Startup culture

particularly the you know the Y cominator guys fetishizes the relentless resourceful founder and the engineer who lives and breathes code. It's kind of like a meme now right and this sentence

is a very quick way of so I'm saying I am the ideal engineer you dream of hiring. It's the cracked engineer you

hiring. It's the cracked engineer you read all about on Twitter right? I have

no distractions. I will not be looking for work life balance. I only build.

It's a form of you know kind of putting him making himself look bad by saying I don't have hobbies this that. But same

time he's boasting about his traits the audience values the most. And next he's saying being part of super lean teams one of my strongest suites has been the ability to work across the stack. Here

he's using what we like to call in-group language. Words like super lean teams

language. Words like super lean teams and work across the stack are not like normal words that most people would use in their application. He's signaling

that I belong. I know your words. All

these YC companies, they they use a bunch of lingo, right? They have a language in their world. And he's saying that look, I'm part of the group. I know

the language. I'm a member of the tribe.

Now, to understand why this worked, you must first understand the audience. Who

is a startup founder? A startup founder is not a corporate hiring manager.

They're not optimizing for any process or something. They're optimizing for a

or something. They're optimizing for a word called conviction. How much do I believe in this? Their world is defined by, you know, there's a very famous line, right? Missionaries, not

line, right? Missionaries, not mercenaries. Missionaries are people who

mercenaries. Missionaries are people who really believe in the mission.

Mercenaries are people who go out, you know, collect a bounty for everything that they do, right? Their greatest hope as a startup founder is to find someone as obsessed with their company as they are. This parex email was not, you know,

are. This parex email was not, you know, just some skills that he was listing like most of you do. It was a mirror to show a founder.

It's their deepest desire of the founder reflected back at them. And I think this is crazy audience analysis, right? Who

cares about technical needs? In the real world, you'll realize all those things are backseat. Psychological needs are

are backseat. Psychological needs are more important. That email was so

more important. That email was so effective. It didn't just ask for a job.

effective. It didn't just ask for a job.

It's saying founder, your worldview is correct. These are the kind of employees

correct. These are the kind of employees you should be looking for. I am one of them. So the core belief of a founder is

them. So the core belief of a founder is their mission is so compelling, so good, such a grand vision that should attract exceptional obsessive talent. And Parx

email comes in and says, "You are right.

You are exactly right. You amazing,

right? People like me do exist and your company is one of the few ones that I have chosen to notice." And this changes the interaction. It's no longer some

the interaction. It's no longer some subordinate asking for a chance. It's

this transaction between equals confirmation that they both belong in the same world, right? The world

building has been created here. The

founder in reading that email has their own identity as a good leader of a worthy mission, it's been reinforced.

Okay? And this is called confirmation bias. You're favoring information that

bias. You're favoring information that whatever existing belief you have that you're looking for that kind of information. Forget labor and all that.

information. Forget labor and all that.

He was selling an identity that the founder was already desperate to buy.

And that is god tier persuasion. So what

can we learn from this? Good writing is about building a world for your audience. So email works because it fits

audience. So email works because it fits in a world that already exists. That Y

cominator, that startup lingo, the kind of world they exist, the kind of language they use, it's already there.

It's all part of their head. And the

secret to all successful storytelling is this thing called worldb buildinging.

And we're going to learn very deeply what worldbuing is.

There are two types of world building.

Either you build a world from scratch or you fit into a world the audience is already familiar with, one that they're already in. Let's talk about how to fit

already in. Let's talk about how to fit in a world first. What So did this is the easier part compared to building a world from scratch. You can take any community you are already part of and

you'll know what they agree on and it's easier to communicate with them. You can

convince your friends on some idea. For

example, Sohham understood the founder world. But you can't convince

world. But you can't convince communities you don't know because you can't fit in because you have no idea what their world is. Remember, every

world, no matter how complex, is built from simple atomic units. In writing,

these are the foundational points that require the least amount of new information for your reader to accept.

Think of them as points of common ground. They're often so obvious that

ground. They're often so obvious that they feel unnecessary to state because they're the foundation of that world.

For example, for a business proposal, if you say something like, "Hey, we all want the company to grow, right?" Or for a health article, "Everyone wants to feel healthy and have more energy, right?" For a political argument, "We

right?" For a political argument, "We all want our children to live in a safe community, right? You start with these

community, right? You start with these simple lowinformation points because they're non-controversial. They

they're non-controversial. They establish a shared foundation. Once the

reader has nodded along to your atomic units, you can begin to construct more complex structures on top of them. For

example, once you've established that, hey, we all want the company to grow, you can then introduce them to a more specific idea. One of the biggest

specific idea. One of the biggest obstacles to our growth right now is our outdated sales process. You are leading the reader from a broad agreement to a more specific problem within the world

you are building. It's hard to know what atomic units are there in a world and what already exists. You can fit into someone else's world in two ways. One is

you can zoom in on the problem. Like you

describe their universe first, then zoom into a continent, then zoom into a city, then zoom into a location. You can also start the other way around. You can

start with a location, then the city, then the continent, then the universe.

Let me show you both zoom in and zoom out. I'll give you an example. If you're

out. I'll give you an example. If you're

not in startups in India, like if you've never actually raised venture capital, you'll have no idea what big investors or founders in India want because they don't actually want to invest in slightly profitable startups. You all on

the internet are like, "Oh, this startup is a little bit profitable." Actually,

they don't care about that. Everyone

sitting and telling big investors to invest in profitable startups are making a mistake because that's not what the investor's job is. For big investor mathematics to work, the company has to IPO or sell at a billion dollars or

above. That's because big investors

above. That's because big investors usually end up owning 20% of 20% of a company at max. Making a small company with a small profit is usually useless to them because the company will never

sell at huge valuations. So they will not make money for the people who have invested in the investors. If you don't understand the big investor mathematics, you'll never produce a piece of writing that they like because it doesn't fit in

their worldview. Now this makes it seem

their worldview. Now this makes it seem like no outsider can change a worldview of an ecosystem. But if you start with their worldview and then slowly as they nod along with your whatever piece of

writing and then in agreement you flip it around and put in your own worldview that's how you change things. I like to call it a frame shift. I'll give you AOS's example. Like the big investor

AOS's example. Like the big investor world in India didn't care much about content creation 10 years ago. But we

start with the idea that hey well you need to build billion dollar companies and you need a good product and good marketing for that. By the way the word for marketing is GTM or go to market. So

we'll use that word instead of marketing and then say by the way you know meta and Google ads are expensive and saturated and companies have to keep spending on it so there's cost of acquisition. Remember they're all

acquisition. Remember they're all nodding along here because this is their world. They already agree with you and

world. They already agree with you and then you say well there's a section of society that doesn't have a cost of acquisition for products and services they make and they're called content creators. So what if you stole the

creators. So what if you stole the content creator game plan and built YouTube channels so you don't have a cost of acquisition over the long term and hence and zooming out of the world

again you have a good alternative to the decline of Google and meta ads which means GTM which is go to market which is their word for marketing is solved which means as long as you have that and a

good product you can build a billion dollar company plus because you're not spending on acquisition long-term the margins are better so better valuations Now next part of this is to give

examples of your modified proposed world. You've given an example of their

world. You've given an example of their world with this content creation modification. Now you have to give

modification. Now you have to give supporting evidence. So then you point

supporting evidence. So then you point your finger and say look at physics walla. Physics walla is a content

walla. Physics walla is a content creator with low cost of acquisition and a billion dollar company. Then you show Mr. Beast and Fastables. Then you show Logan Paul and Prime. Then Kylie Jenner.

The more examples the better. This takes

this modified world that they don't fully agree with and solidifies it with examples. It roots the world. It grounds

examples. It roots the world. It grounds

your story. Otherwise, you might be talking nonsense or some idea that hasn't happened yet. That's why it's so hard to do something without previous examples. The hardest thing for me was

examples. The hardest thing for me was to convince other people when we started using AI avatars that people watch it before I had any proof. I did the same world building. People watch content,

world building. People watch content, presenter gets tired, AI getting better, avatars getting better. Look, here's a low resolution version. That's why I run experiments. Sometimes when you don't

experiments. Sometimes when you don't have concrete examples, you have to resort to a prototype. I had an avatar prototype back then. That's why when startups raise money, but they don't have a full product, it's better to just

go with a prototype. Just root it in some reality that people can touch and feel. So your world building feels real.

feel. So your world building feels real.

But with avatars for YouTube and Instagram, we had a very early prototype and no examples of success. That's why

my avatars at that point at least were such a big phenomenon to the point where even YouTube published that we were the world's largest avatar in the Davos book because once a case study is formed the

example allows hundreds of other creators to root their stories. Hundreds

of people can now point at me and say hey it worked for him therefore I can do it. It leads credibility to their

it. It leads credibility to their stories. After that worked, we got to so

stories. After that worked, we got to so much scale and made so much money, the rest of our storytelling got stronger because my ability to root modifications to other people's worlds started working

better.

I'm in the same place with game development now. There's no proof that

development now. There's no proof that India can make a top class 3D game. But

if I prove it, it becomes a case study and other founders will point to it and be able to raise capital. But because we are going first, it'll be unlikely that we are able to raise capital because I have no previous examples of Indian

success to show investors and root it in any world. I already knew this going in.

any world. I already knew this going in.

That's why I had to wait so many years to do it. So that first I can generate enough cash to self-fund it. Trust me,

unless there's some crazy investor, if you have no external examples of success, you get almost no funding. Or

rather, it's rare. That's why that game Claire obscure was so important. It's

the game of the year. And what was important was that it was made with Unreal Engine and with 30 people. So

it's a good case study for other developers to root their story of hey we can do this with a small team. And this

world building is not just important for investors, it's also important for your team. So as I showed my team that

team. So as I showed my team that another 30-man team could do this, more confidence was built. Right now there's a massive demand for video editors. But

it hasn't hit everyone's head yet. It

will when people start going out and saying look I'm making this much LPA.

Like in software engineering, a lot of the 2021 boom of software engineering was people making YouTube videos saying, "See, I'm making 40 LPA, 50 LPA." That's

examples. That's world building. But in

video editing, because we know the numbers, it is going to happen soon.

The second part which is harder is making a world. Till now, we've been talking about how to fit into another person's world and make a modification.

But making a world is really hard. I

fundamentally don't think it's possible to make a world from scratch. You have

to invent a new language and all of that. The closest I've seen is to take a

that. The closest I've seen is to take a little bit of somebody else's world.

Like you cannot start a world completely from scratch, but you have to take a little bit of somebody else's world and then spin on top of it. I'll give you a best example of this. Have you watched Dune? I love Dune. I've read the books.

Dune? I love Dune. I've read the books.

I've watched the movie. I've played Dune Awakening the game. I think it's one of the best examples of world building ever. But let's talk a little more about

ever. But let's talk a little more about Dune now. But before that, I want to

Dune now. But before that, I want to take a second to talk about today's sponsor, Warp. You know how engineers

sponsor, Warp. You know how engineers today are running multiple AI agents locally to fix bugs or to implement features? Well, while the agents work

features? Well, while the agents work fine, there's a big problem with it.

Context switching. Either you lose track of agents or your machine just gives up since all these agents are running locally. And to solve exactly that, WP

locally. And to solve exactly that, WP has launched Oz that lets you run practically infinite number of agents in the cloud. The setup takes just 10

the cloud. The setup takes just 10 minutes and you can trigger these agents from almost anywhere like Slack messages, GitHub issues, linear tickets, or even the W terminal itself. Plus, you

get a single dashboard where you can manage all your agents and even schedule them to trigger on a recurring basis. On

top of it, if an agent is going in the wrong direction, you can jump in live and steer it in the right direction.

Having said that, you should check out Wu today and get an extra 1,000 AI credits towards Oz. Link in description.

At the most immediate level, the language of Fman and Dune, the native inhabitants of this desert called Arachus, is full of Arabic and Islamic

terms. The messiah figure Paul Atrades is given the title madib which in Arabic translates to teacher or educator. His

feminin name usul derives from the Arabic word meaning principles of fundamentals. Other examples around such

fundamentals. Other examples around such as lisan algib the tongue of the unseen in Arabic. The religious and little bit

in Arabic. The religious and little bit of like you know the messiah vibe of dune are also heavily influenced by Islam. The concept of the Mahadi, a

Islam. The concept of the Mahadi, a prophesized redeemer in Islam who will appear before the day of judgment is a clear parallel to whatever's happening with the Fman and Lisan Algib. So

there's a little bit of that context, a little bit of world building taken from somewhere else that people can understand. Now here's the spin.

understand. Now here's the spin.

The author put in sand worms. massive sand worms and so much lore around it.

Why they exist, what their properties are. Then to root the world, lots of

are. Then to root the world, lots of consistent examples. In the real world,

consistent examples. In the real world, this could be water or later oil. The

most valuable substance in his universe, the spice milange is not a mineral to be mined, but a biological byproduct of the sandworm's life cycle. It's their

excrement, their waste. They excrete it into the sand and then it's exposed to the elements to become the spice. Now,

here's the rooting. This immediately

creates the central conflict of the entire saga. The great political houses,

entire saga. The great political houses, they don't just want to control a desert. They want to control the

desert. They want to control the treasure in the desert, the spice. The

entire galactic economy depends on this worm's byproduct for travel and skills.

Spice isn't just collected. There's the

spice blow that happens. There's a

massive explosion of sand and raw spice, which is a direct consequence of a pre-spice mass growing deep underground and reacting with water. This makes

this, you know, anytime they harvest spice, it's a dramatic, dangerous, and very time-sensitive operation. The Fman,

the people who live in harmony with the worms, they secretly control huge stockpiles of the spice, and they're the wealthiest people in the universe.

Technically, the Fman have learned how to master them. They learn to ride these big sandworms, and that's their primary means of long-distance transport. And

it's a physical skill. It requires like a hook used to pry open a segment. You

open a little bit of the worms plating.

It exposes the sensitive flesh and allows them to steer it by causing irritation. A fman who can summon and

irritation. A fman who can summon and ride a sandworm. A sand rider, it's called a sand rider. It's like a title of immense respect, like a job title. It

becomes a right of passage proving their courage and showing you how the fmen are deep inside the aracus ecosystem. And I

think if you've read any religious textbooks, many religions have sacred rituals involving psychoactive substances, drugs where you transform and you connect with the divine. So in

his books, taking the spice makes you high. Also in the game, it gives you

high. Also in the game, it gives you extra skills. All consistent

extra skills. All consistent storytelling. And in order to be

storytelling. And in order to be consistent and give examples, he goes super micro like the hooks on the sandworm. Any other storyteller wouldn't

sandworm. Any other storyteller wouldn't have gone that deep. So he's taken an existing world and added so much depth to it. And the more depth that's added,

to it. And the more depth that's added, the more it's consistent, the more it's rooted. Any other storyteller wouldn't

rooted. Any other storyteller wouldn't have gone that deep. By all these small micro interactions, the world has depth.

It is rooted. Anyway, let's go to the next segment.

People really like Dune. You can get totally immersed in the world, the relationships between the characters, and whenever you get transport into Arachus in the movies, it's just so

good. And movies, you can even get

good. And movies, you can even get deeper into the storytelling by adding in music and visuals. Again, with its influences and then subtle twists to root it. Every time you see a sky shot,

root it. Every time you see a sky shot, a certain type of music plays. Every

time it goes to the bad faction, some different music plays. But it's

consistent. If it shows the sky shot, that old music plays again. This

relationships between music, characters, locations, going micro, it's very important.

Like I've always had a problem with Superman. When he saves people who are

Superman. When he saves people who are falling, he's going really fast. In a

city like this, that should kill the person he's catching. It's like being hit by a 6' long bullet. It's

inconsistent with physics. when he wears glasses, why don't people recognize him?

Both of these in the comics are explained in a very dumb way. In one of the Superman comics, they said that Superman doesn't hurt people when he flies fast because he has a protective barrier that emits from him. With the

glasses, they wrote in one of the comics that it's a hypnosis glasses. Like,

these are not good, welloughtout examples. It's not like Dune. And I

examples. It's not like Dune. And I

think Superman did well because it was written at a time where there's not so much media. So, it's a little bit harder

much media. So, it's a little bit harder to be caught in the world of DC comics on the big screen, and that's what's happened to them. Whereas on the big

screen, Marvel does it better. Watch.

>> Wait, who are you?

>> You know, Marvel managed to erase some of the stupidity in the comics. I like

the micro interactions between Doctor Strange and Spider-Man. The reason that Thanos belts the Hulk the first time they meet is to set up Thanos as this badass. They showed Thanos in many

badass. They showed Thanos in many movies as this big bad. Again and again, you'd see Thanos in the ending of those movies. They referenced him many times.

movies. They referenced him many times.

And his first showing, you can see he's the real threat because he beats up the Hulk, which is at that point one of the strongest characters they had built up.

That's called foreshadowing. All of the small sequences of Thanos before the final movies is all setting up a small part of the universe ready for the final reveal later. That's how you build the

reveal later. That's how you build the universe. And once you've designed your

universe. And once you've designed your world, you need to make it easy for people to enter and explore. This is

where the craft of writing comes in. The

principles of good writing are all about cognitive hospitality. You're making a

cognitive hospitality. You're making a world that's friendly for people to enter. Anything that the user strongly

enter. Anything that the user strongly disagrees with, break immersion. Masters

of writing like Paul Graham and William Zissner all preach the gospel of be simple. Use ordinary words, write clean

simple. Use ordinary words, write clean sentences, omit needless words. It's all

about reducing the load on your brain.

The brain has an immune system that rejects new ideas and new words. The

less energy your reader has to spend decoding your writing, the more the brain takes in the idea. That's why GPD writing is so bad right now. It's all

very, very wordy. This brain and immune system idea is very important.

Do you know why the most likely people to start believing in random stuff like chakras and motivational gurus and some obscure stuff when they're broken in in life? Because the brain becomes

life? Because the brain becomes suggestive. The immune system of that

suggestive. The immune system of that brain drops. Whereas it's very hard to

brain drops. Whereas it's very hard to convince someone with a high ego to change. People go to temples and

change. People go to temples and churches when they are at their lowest.

Psychedelic drugs, for example, also make you very suggestive. You're more

likely to believe things on them.

Do you know why AOS has done so well in the last 2 years? I keep thinking about it because we accidentally found something. We found that very good video

something. We found that very good video editing drops the immune system of the brain. It was a totally accidental

brain. It was a totally accidental finding and because we have a we have we were anyway producing video editors. We

are producing the spice. You just don't see it yet. And are you seeing what I'm doing right now? I'm building a universe. I'm rooting my story in

universe. I'm rooting my story in examples that I previously gave you within your world. You can use these principles to guide your reader. The

first piece of information you provide heavily influences what comes next. Your

title and first sentence are powerful anchors. An app that does rag with LLMs,

anchors. An app that does rag with LLMs, it's a boring conversation. But saying

Rahul is using this app to prepare for his exam in 5 minutes immediately shows a compelling human benefit, something that people can understand. And humans

are fundamentally wired for stories, not for numbers. A concrete story is very

for numbers. A concrete story is very memorable, very persuasive. Instead of

saying our solution improves productivity, tell a story. Himish, one

of our top reps used to spend all his Friday afternoon updating spreadsheets.

Now that process is automated and last Friday person closed two extra deals.

That's five lakhs extra they closed.

This makes the benefit tangible. Like I

said, you need to keep it simple and not stray too far from what the reader believes. It's not possible to change

believes. It's not possible to change the world that quickly. You need to change it in steps. people thinking I'll come and change the world entirely change all of this rapidly changes the world will not believe you the more unbelievable the story the harder it is

to buy it I know a guy who makes a 100 crores a year liquid when he goes on a podcast and gives examples of people's annual earnings he always says just five lakhs just so people relate because he knows and even I know that if he goes

out there and talks about well making 100 crores nobody would relate it's not rooted in any world right so people be like this is this is all in the air and I think the greatest enemy of clear writing is what the cognitive science

Scientist Steven Pinker calls the curse of knowledge. This is the inability to

of knowledge. This is the inability to imagine what it is like for someone else. This guy understand because he

else. This guy understand because he knows people won't understand the 100 crore example. I'll give you another

crore example. I'll give you another example. Right? Zerva founders have been

example. Right? Zerva founders have been making thousands of crores per year, no one cares. No one bats an eye. The

one cares. No one bats an eye. The

minute news came out that they're taking 100 cr salary, everyone lost their This is because people relate to the word salary more than dividend. Breaking

this curse requires a conscious act of empathy. You must constantly ask, "What

empathy. You must constantly ask, "What does my reader know? What is the simplest possible way to say this? What

context am I assuming? And this leads to a very obvious conclusion about simplicity. Writing simply is not

simplicity. Writing simply is not dumbing it down. That's not what writing simply is. It's a sign of respect. When

simply is. It's a sign of respect. When

we make our reels, when I do my short form reels and make it simple, it's not because I'm not technical or I don't understand it. It's because we respect

understand it. It's because we respect our wider audience and fit into their worldview what they can understand.

Because it's simpler, it gets a lot more scale because more people can understand it. Sometimes I get attacked right for

it. Sometimes I get attacked right for for making these simple res. The people

attacking are mostly young. So I don't blame them. The easier you make, the

blame them. The easier you make, the easier you scale because you fit into everyone's worldview. It's why Tan

everyone's worldview. It's why Tan tweets about cricket and Rani talks about ghosts. Everyone understands to

about ghosts. Everyone understands to write simply to use ordinary words, clean sentences is all important, but you need to make sure it's rooted in a world that other people understand. You

are giving them a gift, a clean, efficient program that runs smoothly in their operating system instead of you asking them to download your operating system. The best storytellers have

system. The best storytellers have something called reach from ground truth. I I invented this this concept,

truth. I I invented this this concept, right? Without telling a lie. How far

right? Without telling a lie. How far

can you stretch a person's brain from what they believe? How far off their own axis can you get them to go without having to tell lies? That is the true test of a storyteller. Can I convince somebody who hates this country that's a

good country? Can I convince a person

good country? Can I convince a person who doesn't want to have kids to have kids? Can I convince a person who is an

kids? Can I convince a person who is an atheist to believe in religion? All

without telling a lie. And I think the amateur writer like they try to brute force this. They'll bombard the reader

force this. They'll bombard the reader with facts that contradict their world.

They be like, "No, no, you're wrong.

This is how it is." That never works.

The skilled persuader does the opposite.

They'll begin by validating a part of the reader's belief system. You are

right to be skeptical of new marketing fads. Most of them are waste of time and

fads. Most of them are waste of time and money. That's why it's important to look

money. That's why it's important to look at the few that are grounded in fundamental principles. Once you build

fundamental principles. Once you build that rapo, it lowers defense before you introduce a new and different idea. And

I'm telling you, the world hates new ideas. You might think the world likes

ideas. You might think the world likes new ideas. It doesn't. And the only way

new ideas. It doesn't. And the only way to deliver a new idea is to first start with an old idea and step by step.

The second thing is we judge the likelihood of an event by how easily an example comes to mind. This is also the engine behind powerful storytelling. A

very concrete story is way more persuasive than any numbers or statistics. 92% of customers are

statistics. 92% of customers are satisfied. No one cares. that story the

satisfied. No one cares. that story the first one of some Raj Raju uh you know suddenly seeing using your product and doing well and making more money becomes more available to the mind. I'll give

you an example in countless adventure movies quicksand is a mortal threat. You

know quicksand you walk into quicksand you'll sink into it. It's a terrifying way to die. The reality is that while quicksand is a real thing it's a real phenomenon. It's almost impossible to

phenomenon. It's almost impossible to die in that way. Humans are less dense than quicksand, meaning we are naturally buoyant in it. You would sink, but only up till the waste. The real danger is

not being sucked under, but rather getting stuck and dying from dehydration, exposure or starvation.

Annual deaths worldwide from quicksand are negligible. Maybe single digits out

are negligible. Maybe single digits out of, you know, 8 billion, 9 billion people in the world. Many people discuss sharks, terrorists, plane crashes as threats. Let me just tell you, more

threats. Let me just tell you, more people in the world die from diarrhea than from sharks, terrorists, and plane crashes. combined. But you can't use

crashes. combined. But you can't use diarrhea in your story as a threat. You

would have more success with terrorists or plane crashes because they're more recent and they come to mind easily. So

let's go back to telling a person a story that changes their mind radically.

The easiest way is to induce something called dissonance. Present the reader

called dissonance. Present the reader with two ideas which they believe are true but which are in conflict. I'll

give you an example. Okay, so you're saying code is more valuable than content. Correct. Many people agree,

content. Correct. Many people agree, especially engineers. But how many apps

especially engineers. But how many apps have you downloaded last year versus how many new creators or channels or Twitter accounts have you followed? What is the number? If you argue that funding,

number? If you argue that funding, salaries, etc. come because there is usage of something. Therefore, there

must be producers and producers need talent and capital, then you're saying that content is going to explode as an industry because you're watching so much more of it right now than downloading new apps. You're watching it right now.

new apps. You're watching it right now.

The app you're watching it on, YouTube, is a very old app and doesn't require too many engineers to run anymore. But

that one app supports hundreds of thousands of creators who all need to hire. You see where I'm getting at? This

hire. You see where I'm getting at? This

is the way, right? First you establish something that you think you know that people believe in and then you say but I'm deviating a little bit. Then do you agree with this follow on? That's the

thing. And the last thing as a storyteller is you need to provide resolution. Your idea is not just

resolution. Your idea is not just another proposal, right? Like you want to release them from this problem of having two conflicting thoughts. Now you

put the reader in a problem that they need to solve. That is why almost all the stories these days I tell are true.

Everything I have told you in this video is 100% true. My job is not to come up with stories. It's to actually find

with stories. It's to actually find hidden but true stories or stories that the world doesn't believe in but are true. Like the devs, you know, going out

true. Like the devs, you know, going out of business over a few years. One more

thing is that most writers make the mistake of treating all their points as equal. They'll present one laundry list

equal. They'll present one laundry list of reasons and hoping that you know if I give you 10 points as to why this will work, it'll be persuasive. This is the engineering answer syndrome. You, you

know, in an engineering answer in college, we just put many points and hope that you'll get more marks because one of the points will be right. In real

life, there are no part markings. In an

argument, the highest leverage point is the single core assumption upon which the entire opposing view rests. For

example, for a long time, people said that making AAA game has a $100 million budget. You break that down, it's about

budget. You break that down, it's about $25 million of dev budget and $75 million of marketing budget. Let's break

that dev budget down. $25 million of dev budget the starting salary of an engineer in the US is $200,000 for a decent C++ engineer in India for a starting salary isn't it onetenth of that cost so technically you're saying

the entire dev budget in India for a game of this quality could be onetenth of what it is in the US and then you say well there are new tools like photoggrammetry and this and that so it should be even lesser than you know

onetenth of the cost that is how you flow in instead of fighting 10 different surface level symptoms uh and get into a bad argument the master writer and even the master businessman identifies the

one flawed premise at its root. Why

should it cost $100 million to make a game in India? Should cost much lesser than that. And on the marketing side, if

than that. And on the marketing side, if you're into content, you'll figure out marketing for free if you know how to make memes. I'll give you another

make memes. I'll give you another example, right? I was hiring a CFO and

example, right? I was hiring a CFO and let me tell you, a CFO is expensive. We

have 500 employees now and we absolutely need a CFO. And I called the guy whose company I was hiring from. He's actually

shutting it down. So, he's letting the CFO go. And I said, "Hey, give me five

CFO go. And I said, "Hey, give me five reasons I should hire him at this price." It's a huge price. He said,

price." It's a huge price. He said,

"Bro, this is a gut decision." And

instead of falling short of his expectation and giving him whatever, two, three lakhs less, whatever amount it is, just close one more deal. Having

a CFO means you can close your eyes and go faster. No, he'll help you close 10

go faster. No, he'll help you close 10 times more deals. And I trust him on that. And that's it. That line convinced

that. And that's it. That line convinced me. Not one five arguments as to in his

me. Not one five arguments as to in his CV what he's done in the past, this that just okay, if I hire this person and I pay him a little more, I can go much faster. Like I said, he could have given

faster. Like I said, he could have given me five reasons, but ultimately he had to just fit my worldview of okay, company's growing. He already knew that.

company's growing. He already knew that.

I just needed someone to handle the back office and the amount difference doesn't matter if you're growing that fast.

Better to have the best person for this than haggling for a small amount here and there. And the minute he said, "I

and there. And the minute he said, "I trust this person," that became the most important reason because I trust the person I was speaking to. So instead of glazing five points, one was enough for

me and that one converted me. And now I just want you to know I tricked you.

This video is not about writing. It's

not about persuasion. This video all it while was about prompting. The best way to think about good prompting is that it is worldbuilding for the AI. Let's get

very technical. My claim that effective prompting is akin to worldbuilding for AI is not just an analogy. It's an exact technically accurate description of how

we guide large language models. If you

can take all the principles I just taught you and move it into prompting, you are essentially constructing a temporary reality for the model to inhabit, you're creating a universe for the model and you're influencing its

behavior at the fundamental levels of machine learning, especially when it comes to how token embeddings work and attention mechanisms work. I'll now get a little bit technical and then tell you why everything you learned in this video

is also going to help you no matter what job you're doing.

An LLM in its raw state, you can conceptualize it as a vast high-dimensional probability space represented by its token embeddings. Let

me make that simple. Every word or token has a numerical vector representation.

And the relationship between these vectors defines the model's understanding of language and concepts.

Without a specific prompt, the model exists in a state of just potential capable of generating nearly anything from its training data. A well-crafted

prompt acts as a set of physical laws for a small universe. So every time you prompt something, you're constraining the universe. Here's the technical

the universe. Here's the technical breakdown. First, there's something

breakdown. First, there's something called an initial state vector. When you

provide a prompt, the text is tokenized and then converted into a sequence of embedding vectors. This initial sequence

embedding vectors. This initial sequence of vectors defines the starting point and a trajectory within the model's enormous embedding space. So vectors for similar concepts will be very

mathematically close. the vector for

mathematically close. the vector for gang signs would be closer to the words gully and hood than it is to five-star hotel. A generic prompt like tell me a

hotel. A generic prompt like tell me a story places the model in a very broad undefined region of the space. On the

opposite side, a very detailed prompt like write a detective story set in 2049 cyberpunk India where the protagonist is a detective who only communicates in gang science immediately teleports the

model to a highly specific constrained space in its embedding space. You're

creating a universe for the model. Every

time you prompt, you're creating a universe. So smaller the prompt, the

universe. So smaller the prompt, the more it's putting in context for the universe from the generic whatever statistically likely. But the minute you

statistically likely. But the minute you are very specific about what you want, you are creating the world. And just

like in Dune, you can get as specific as the hooks inside the sandworms and then your world is even more richer. Next is

the contextual modulation of embeddings.

Don't get confused by this technicality.

Token embeddings are not static. There's

something called contextual modulation.

Meaning that the representation of a token is influenced by the tokens around it. A very rich prompt filled with

it. A very rich prompt filled with specific entities, relationships, any sort of cues. It creates this contextual field we spoke about. This field you

know it physically warps the local space making everything more defined. It's

trying to make a world where your concepts are now more probable and less the models. Think of your prompt taking

the models. Think of your prompt taking the model from basically everything to a very specific universe using your words.

You're basically going from a zoomed out world where everything is generic because that's what AI surface is to a very specific world. You are shap you're like a it's like it's better seen as like a you know clay pot. You're not

putting the pieces of pot together.

You're just shaping it with your hands.

Right? Similarly with your words you're shaping the world into what you want.

But if you only you know shape a little bit only one part of the world will be what you want. The rest will be generic pot. You are actually spawning universes

pot. You are actually spawning universes every time you prompt. And some of you are making very shitty universes which are very generic. It's like making a clay pot but you know pressing only on

one side. The rest of the side is just

one side. The rest of the side is just going to be generic whatever the pot was. Anyway, the transformer

was. Anyway, the transformer architecture which is the foundation behind every modern LLM. It has

something called attention. It has a mechanism called attention. Attention

allows the model to weigh the importance of different tokens in the input when generating a new token. So, a detailed prompt is essentially telling the model what you should pay attention to, what

you should focus on. Think of the attention mechanism as a spotlight.

Okay? The model shines this spotlight on the context. A very vague prompt is

the context. A very vague prompt is giving the model very little direction on where to shine this light. Tell me a story. It makes the model aim, you know,

story. It makes the model aim, you know, wander aimlessly. It's going to, you

wander aimlessly. It's going to, you know, pick up the most generic story it can think of. A detailed worldbuing prompt provides a very clear script. For

example, in our detective story prompt, the attention mechanism will naturally assign higher weights to tokens related to detective gang science, cyberpunk India and then generate subsequent text.

And I think all advanced prompting techniques can be seen more directly as a form of world building. That's what it is. Advanced prompting is world

is. Advanced prompting is world building. For example, AI always

building. For example, AI always generates sentences with negations, right? It's not just X, it's Y. You've

right? It's not just X, it's Y. You've

seen AI generate, AI writing sound like this. We're not just writing a script,

this. We're not just writing a script, we're baking cakes, stuff like that. But

you can steer your world not to do that.

Give it that specific example and say, "Don't do this." Setting the physical rules for your world and you can nudge its outputs towards any desired style, personality or knowledge base. This is

the technical equivalent of setting laws. Just like we have laws in our, you

laws. Just like we have laws in our, you know, you know, cars can't go specific.

I mean India breaks all laws but you know in most laws the entire reason for laws is you can tell what the world will be could be you know anything can happen but you know these sort of things will very rarely happen because laws exist

whereas in the world with physics physics is our law set for the entire universe right you know nothing can violate the rules of physics so you can set those rules you are the rule maker you are the law law maker you are the

god that's spawning universes and I think it's quite lame to be a god and say you know to to just have a threeline prompt like that's that's not what you should be doing. The ability of LLMs

also to do in context learning is important. If you provide examples and

important. If you provide examples and very explicit instructions, you are essentially giving the model a very condensed instruction manual for the world it's about to simulate. Like I

said, you're writing laws. I told you in normal writing examples ground the story. We've spoken about this on the

story. We've spoken about this on the storytelling part. When you provide a

storytelling part. When you provide a few examples, we call it few short prompting of the desired output or style. You are giving the AI enough of

style. You are giving the AI enough of that world building and laws and lore for your world. The model then uses these examples to infer the underlying patterns and rules of your world. To

give you an example in Dune, if you don't mention Dune to AI and you just tell it, okay, there are sand worms, there's this sand, there's these multiple waring nations, etc., etc. You know, if somebody wanted to ride the worm, how would they do it? What device

would they use to do it? And believe it or not, AI will actually give you the idea of the hooks because Dune is there in its training data. But that's the thing, you don't need to specify everything. You just need to specify

everything. You just need to specify enough of the world. For example, if you ask what are the top two successful Indian businesses, it'll give you some examples. But if you ask it for more

examples. But if you ask it for more specifics, you give it like very specific examples, say I want businesses like this. Then it'll give you a very

like this. Then it'll give you a very different type of business. For example,

if you ask for what are the top Indian businesses, it'll give you, you know, Amanis and Adanis and whatnot. But you

can get very specific. Do you want bua type businesses, food businesses, startups? It really depends on your

startups? It really depends on your examples because the AI is thinking in a very different way, is spawning a very different world, but you can shape how this world goes. Anyway, the entire point is that world building when it

comes to writing for a reader or prompting are exactly the same thing.

They're two parts of the same coin and if you get good at one, you get good at the other. Till now, most people with

the other. Till now, most people with hard skills in India didn't really need to get good at world building or writing. But after AI, I think it's a

writing. But after AI, I think it's a critical skill that everyone needs to know. And I hope this video has gotten

know. And I hope this video has gotten you started thinking about writing and thinking about thinking because writing is a form of thinking. So definitely go read some more, practice as much as you

can and do not lazy prompt either AI or any of your readers. That's it for me.

Bye.