Many people today, if you ask them, "What kind of business would you like to start?" They say. It's a very stupid thing. Why? The difference between GPT, Gemini, etc. Currently, Claude is winning. But I don't recommend anyone to buy an LLM. People are buying LLM aggregators. Pay ₹2,000 once and get access to all models. It's very similar to the dating scene actually. Like, you know us, committed to one model. Not for guys like us. Yeah. We need to talk about intellectual obesity. Pick up any random person. He is starting to think like an average AI. And this question people keep asking, "When will AI become as smart as humans?" If humans themselves become dumb, then their benchmark is falling. AI tools show a lot of foolishness. AI is not as smart as it is being hyped. It's literally like having a super-powered intern. An intern who's slightly drunk. No. The most stupid mistake is not building a team that will be able to exploit the tools of 2027. So, what can they expect in 2027? Like I am currently building my own second brain and I know that 5 years from today, I would be able to clone my brain easily because that allows you to become immortal. By the way, Meta has just bought a patent that they can run your social media even after you are dead. Wow. So they are all planning immortality. Do you think more entrepreneurs will emerge in the near future? 110% because there will be unemployment. Hiring has frozen, first of all. So what will these people do? They will be forced to start. They will be forced to start, and this is evolution. We are saying, "Stop solving this 'B' problem and step up."

Wow, we have moved so fast in this podcast. So much dense knowledge has been shared. It's such a practical podcast from your career perspective. Gone are the days when people thought that if you are great at prompt engineering, you will be great at AI. The world of AI is also advancing every month. Our country's top AI educator, my brother Ansh Mehra, who I believe is also a top thinker in our country. Today you will listen to him, learn from him, and I promise you, you will gain a lot of value from this podcast when you reach the end of it. Take notes. I hope that for many of you, this podcast will be a career turning point. This is the density of this podcast. This is the impact of this podcast. This is one of the top AI educators, Ansh Mehra, on Skill House, bringing you The Ranveer Show. I will try not to crack jokes today. But this is going to be that "whoosh" podcast where so much value will be packed and presented to the audience in a flash. There are some audience members who are beginners. I assume that everyone knows prompt engineering at this point. Everyone knows that their jobs will not be replaced. But I think this is going to be a hyper-practical, hyper-money-oriented, hyper-growth-oriented AI podcast in Hindi. Welcome, Ansh Mehra. I think you are literally the leading AI educator in India today. Hopefully, I look at you in that light, and I also know that as far as my understanding of uncles and aunties goes, they turn to you for the sake of AI knowledge. So you are at the forefront. I think the right person to explain these things to us. Thank you. Okay. Ansh bhai, where would you like to start? Because I actually want to keep college students in mind first, because many of my engineering college juniors are not getting jobs. And people my age, who are 32, are rapidly losing their jobs right now. So, keeping jobs in mind, what should be the current game plan for college students?

Bro, it's actually very surprising because the kind of questions I was practicing for my placements are completely outdated today. At this point, many people think that AI will cause a lot of unemployment. AI will take away many jobs. Whereas the sad truth is that AI is making you unemployable. More computer science engineers are becoming dumber and slower because what do they do? They get any problem and don't think, they tell AI to solve my list. By default, the result AI gives you is average. And they think, "I will be hired by Big Four." So when we were studying in college, coding was taught and programming was also taught. People think that AI will replace coding, and it will. But the second component, for which you actually used to get packages worth crores, was programming. Programming is your ability to take a complicated problem and break it down. Kids are not learning that. So if you want to get placed, I think you should double down on your programming skills and not just coding skills. Plus, here you are not just talking about software. You are using software as an analogy for every possible job. Correct. If you have a very big problem statement, breaking it down into small problems. Figuring out which problem AI can help me with and which problem only a human can solve. Having this intelligence is very important. Okay? At the beginning of this podcast, I would also like to say that as a founder of multiple companies, I want to talk to you because I am also facing a problem in terms of AI education for my own organization. Dhanish is also sitting with us. We work very closely together. And his sole mission in our office is to spread AI education among multiple companies, multiple departments, but practically speaking, it's not that easy. So, the next question is for founders, because I know that many founders are also watching this. What is the advice for founders?

Bro, in general, AI is giving significant gains in only three departments. The first is customer outreach or lead acquisition. Meaning, finding customers. The second is proposal design. You've got the customer, now what do you pitch? Can AI help me there? And then, after delivery, getting feedback for customer success, analyzing these things. In finance, now that Perplexity and Claude are coming, showing demos, it's very promising. It's still not perfect. And then, of course, there's basic marketing and content. Beyond these three spheres – lead acquisition, customer success, and marketing – AI is not as smart as it is being hyped. AI tools are very dumb. So when I say AI is smart, I am talking about agentic AI, where there are actually gains. AI tools show a lot of foolishness. Okay. You will have to explain the difference between tools and agents again, then we will move deeper into it. Correct. There are only three differences. Point number one: an AI tool can perform a task of a specific category well. ChatGPT is good at writing. Midjourney is good at images. Gamma is good at presentations. An AI agent, on the other hand, can do many things by using different tools. So my customer outreach agent will first go to ChatGPT, create a proposal, then go to Midjourney itself, create a cover image, then go to Gamma itself, create the proposal itself, and achieve these goals. The second difference: in an AI tool, you have to give a prompt. If you don't give a prompt, the AI tool will not take the next step. An AI agent asks you for a goal and asks you for tools, memory, and says, "Now leave me. For the next 10 days, I can do this work on autopilot." This is the second difference. The third difference: AI tools have already been made by AI companies. You cannot build a separate ChatGPT or Gemini for your company. But you will have to build an AI agent yourself because every company has its own nuances, its own SOPs. So every company will learn to build AI agents. They are learning, but that too is stupidity. Building an AI agent is not the right approach because very soon AI agents will be building AI agents. Okay? So, the AI you have seen so far was built by humans. The first version of ChatGPT, the first version of Gemini, was made by humans. But the latest tools coming out, like OpenAI launched Codex, they themselves officially said that GPT helped in the development of Codex. The new features of Claude, meaning Claude Co-worker, Anthropic itself says that AI and humans built it together. So AI agents will eventually build AI agents. You have to orchestrate, create SOPs, create workflows. Because in 3 years from now, everyone will have agentic tools. But not everyone will have those SOPs that can be used to onboard the agent. Okay. Basically, before moving on, I would like to say that there are many users who are absolute beginners in AI usage. Okay. We will address them. Before that, I want to ask one last founder's question. You said that it's not wise to tune your own AI agents because AI agents will be tuning further AI agents. Right? So, as a founder, which agents should I use or where should I buy them from? What should I be doing?

Yeah. So, bro, earlier when the concept of agents became popular, building agents was a very technical process. But now, with the new tools coming into the market, Perplexity, Computer, Claude, Co-worker, in all these agents, you just have to give written documentation saying, "Every morning when I wake up, I do this work for the next 6 hours. Now you learn this, and whatever I do in apps, I will give you the connection to those apps. I will tell you what I do every morning. Now you run this on autopilot." Wow! So life has become very easy. It's literally like having a super-powered intern. Not really, because where the stakes are very high, AI doesn't help. And I tell people that most people think that AI is a cheap replacement. But it's like an intern who's slightly drunk. No. So if you go with that assumption, you will use AI correctly. Meaning, today's agents in March 2026 are drunk interns. Or slightly drunk interns. But this drunkenness will reduce. Of course, it will. Hallucinations will also reduce. By when? In my opinion, by December 2026 is going to be insane. Meaning, there are still many hallucinations. But in the next 9-10 months. And that's why I am telling all founders that your biggest, your most stupid mistake is that you are training your teams for the current AI tools. You are not building a team which will be able to exploit the tools of 2027. So, what to do today? The very first step is to move all teams away from prompt engineering and start teaching them context engineering. I also made this mistake. I also used to tell everyone to learn prompting. But prompt engineering has failed now because if I go into Gemini deep research today and write a bad prompt, it itself improves my prompt. Context engineering has a simple framework, the ACE framework. A, C, and E, where A stands for describing the action, what you have to do today. C stands for context, why I am doing this, for whom I am doing it. Here you can also put constraints. And E is setting the expectation, that if you are making a logo, upload the logo you like. If you are writing a script, upload the scripts as a reference. Action, Context, and Set Expectation framework. Okay. Explain with two-three quick examples. If I have to make a logo for my real estate company. So what is the action? You need to make a logo for this specific name, say "Bear Biceps Real Estate." What are the constraints? Make a square logo, keep the design minimal, and use vector colors. Do not use gradients. Context is given. Constraints are given. Expectation: Here is the logo of Dumk, I like this kind of logo. Done. Okay. Let's take another example, scripting. I have to write an Instagram script. What is the action? This is my topic. Read the article and write an Instagram reel for me. What is the context? It should be 60 seconds long. My audience is a college student. Expectation: Include reels of five such creators on the same topic that went viral. AI will never disappoint you if you follow these three steps. Okay. And no difference between GPT, Gemini, etc. No, of course, they are all fighting with each other. Currently, Claude is winning, by the way. Okay? But I don't recommend anyone to buy an LLM. What do people do? They buy a ChatGPT. They buy a Gemini. But all the enterprises I am training are buying LLM aggregators. LLM aggregator means you pay ₹2,000 once and get access to all models. Because sometimes Gemini is advancing, sometimes GPT is advancing. So it makes no sense to be committed to one AI model. Okay. LLM aggregator. Yes. Can you name any? Copilot. Yes. Perplexity. These are the two top LLM aggregators for text. For images, you have Hicksfield and Freepik. Okay. They are the best LLM aggregators. Wow. It's very similar to the dating scene actually, like, you know us, committed to one model. Not for guys like us. Yeah, not for guys, not for us. Okay. So this will be the pricing model. And by the way, today people say that ChatGPT for ₹2,000. I can bet that next year this price will double. Otherwise, the economics alone will not work. AI tools are about to become more expensive than ever in the next two-three years. And at that time, because people will become dumber, we will have no option but to buy them. Wow. It's how economics works. Because the current economics is very expensive for them. Meaning, they will be at a loss. Yes, it's very difficult to make money right now. So for something to become profitable in the future, it has to make its product very expensive. No, I am very, very interested in what he will do now because OpenAI is shifting into hardware also. Which is a brilliant move, by the way. Robotics? No, like pendants, pens. Like he's paired up with Jony Ive. Right? Now, there's a very interesting analogy here. It is the Coca-Cola analogy. When refrigeration became very popular and refrigeration came into every home, the most money was not made by fridge sellers. Because a fridge is bought only once in 3 years. The most money was made by Coke. Because now Coke can be kept inside the fridge of every house, and Coca-Cola can be consumed daily at scale. So OpenAI was only providing intelligence. Now, if it enters hardware, it's a completely different ball game altogether. Like Godrej soft drink, or correct. Godrej versus Coca-Cola. Okay. So Godrej is making soft drinks now? Yeah. Okay. Moving on. Now, let's say there's a young professional, irrespective of the job profile, graphic designer, you know, wrestling coach, anything could be, a policeman. They just know that yes, GPT is a great thing for nonsense or for studying or for self-analysis. Their understanding is limited to this. Now, what's next?

Bro, I am going to give a very important recommendation. If anyone takes this seriously and applies this framework to their personality, they will be far ahead of 99% of people. And it's called the SHAPE framework. And I stand by these rules. The S stands for be SOP-driven. Whatever you are doing, write everything down because in the future, your writing skills are going to be way more valuable. Your vocabulary is going to be way more valuable than any technical skill. The H stands for Human in the Loop. Meaning, whenever AI gives you a result, you have to use your intelligence to build on top of it. So use AI results as a baseline, not a finish line. People who use it as a finish line are screwed. As in, use human intelligence to build on top of it. So S is done, H is done. Now comes A for Arbitrage. Whatever you are learning today will be outdated in 3 months. So set this expectation that you are in a limited window, an arbitrage, from where you can earn. Because in the old days, there was a mindset that if you got a degree, you would earn for 10 years. That's not going to happen anymore. So set the expectation of arbitrage. Learn how to learn difficult things. Because if you can learn something in 10 minutes, your competitors can also learn it in 10 minutes. So learn how to learn difficult things. P is Product is no longer a moat. Earlier, if you made an app or a very beautiful poster, that was your moat, that I can make this app. That is being eliminated because of AI. Because using AI, you can reverse-engineer anything. So the real moat is distribution and trust built on top of that distribution, which AI cannot do for you. So S, H, A, and P. The last is E, which says Stay Employable. And this is a very big point. Because you and I talk about this regularly, right? When fat and sugar became very accessible, there was a physical obesity crisis. And there was a guy on YouTube who said that if colleges and schools don't understand this concept properly, then we will have intellectual obesity in young children who will not know how to think clearly because nobody is teaching these people how to use AI. So if any person takes the SHAPE framework seriously, automatically by design, you will be far ahead. You are basically highlighting very basic level human skills only. Things that have been relevant for 100 years. They will stay relevant for the next 100 years. Yeah, probably. Then I would say read those old self-help books. Right? That's the first principle. Sun Tzu's Art of War. Agree. You know, the old, I don't know, Napoleon Hill kind of, Napoleon Hill. That kind of stuff, just understanding the thinking framework. Okay. So you said SHAPE framework, let's call it that. Correct. This is about intellectual obesity. What are people doing? Now open the media. Everyone is afraid that in 3 years from now, some random AI tool will start thinking like an average human. Everyone fears that. An even bigger fear is that in 3 years from now, pick up any random person, he is starting to think like an average AI. Ask him some questions, and he gives answers like AI. Because creativity is gone. Unique thoughts are gone. And this question people keep asking, "When will AGI come? When will AI become as smart as humans?" Of course, AGI will be achieved. If humans themselves become dumb, then their benchmark is falling. So my belief is that right now, these people who are summarizing, using Notebook LM, creating all mind maps. In 3 years from now, there will be a phase where the insights of 80% of people will be so shallow and so dumb that there is going

There is a huge demand for people who can consume long-form content. Who can read a lot. Who can come up with insights. Because if you want to be the human in the loop, you have to be worthy, right? Hmm. If you don't know more than AI, how will you build on top of AI? You indirectly answered the question of how to protect your own job. So you are saying that the average person will become so dumb that if you focus on smart things today, you will be able to protect yourself in the actual AI age that is coming. Correct. That's what the math is indicating. Correct. And you will be such a rare person. A person who can have creative thinking, who can write well, who can speak well. Hmm. Who can think well, write well, speak well. So in this too, we have a trinity-type framework: that the future belongs to those who have good health, great taste, strong judgment, high pattern-finding skills, and the vocabulary to describe those patterns. You need good health because if you have low energy and fuzzy thoughts, then nothing can help you. Right? You need good taste. Because taste cannot be taught. It can only be learned. Taste means that you are creating something, and your gut knows whether it will work in the market or not, and that is only known by standing on the ground. You cannot download that. You cannot download taste. In our media world, they say Shabana Azmi actually says, "Get exposed to different forms of art. It's just an easy way to build taste." Correct. Judgment comes one level above taste. Figuring out what we should do in the opposite position. Right? Judgment is also a very difficult skill that, by the way, only comes with experience. Then pattern-finding skills: if you are looking at four different case studies, you should be able to tell what pattern this is. This is not something new. This is just one of those. And at the very end, the most crucial thing is the vocabulary to describe that pattern. Because if you have something in your mind, and if you cannot write it down, then AI has no telepathy. Until unless that neural thing becomes very popular. This vocabulary to describe the pattern seems like a very easy skill to me. But that's my bias as a content creator. That's my actual core skill, writing. But you are saying that this skill is not very popular in the world. It's very poor. Very poor. And you are saying that this is the number one skill in the age of AI. Yes. I don't give it number one, but I give equal importance to all five: good health, taste, judgment, pattern finding, and vocabulary to describe those patterns. Okay. Because you have highlighted this particular skill. I actually want to go a little bit deeper into the skill because it also applies itself to prompt engineering, which was the 2025 narrative: work on prompt engineering, you will become better using AI. Then the narrative became: if you want to increase AI literacy, learn about tools. Today, you are focusing more on human skills. Bro, Jeff Bezos says, "Bet on things that do not change." We have to do that. The whole world is sitting here thinking, "Tell us today which LLM is coming." I am saying, "Bet on things that will never change." If you focus on this, then even after 10 years, you will be valuable. Bet on things that cannot change. This can also be understood as: if you want to start a business, focus on food, clothing, and shelter instead of tech. Correct. Food, grounded things. I would like to say one more thing to the audience: I think a story came out in the news that an AI agent made someone earn about $43,000 by trading in the stock market. Correct. And that was true. No, I don't think so. You think it was a lie? To promote some app? Yes. So now I will tell you, nuances come into play here. How did they earn? Was it day trading? Did they hold it for a week in equities and then exit? Did they bet by looking at charts or by going into books, meaning fundamental analysis? There was no nuance, so how can I believe it? For example, I will give you a practical use case. There was a guy in the US who reduced an almost 1 crore bill to 27 lakhs using Claude because when the hospital gave him the bill, Claude actually analyzed it and told him where the hospital was fooling him. AI can do all these things, by the way. Perplexity now has a built-in stock screener. That can do analysis and predictions for you. It is available to anybody right now. But to say that an agent is going and trading by itself, buying by itself, and exiting by itself. Right now, it's not practical. Maybe after a year. Maybe. And what else? I am pretty sure SEBI will block them too. Any kind of bot trading on any of these trading websites is going to get captured. Just like today when you go to a website, it asks, "Are you a robot?" So I am pretty sure that on the demat account website, "Are you a robot?" will automatically appear. We will stop it. The issue is that the model has to be so smart, its training must have so many decision-making frameworks that it can make the right bet. Okay. Right? If you expect a normal LLM to make crores of rupees for you, then it won't happen. But bet on things that can't change. Correct. So now let's go deep into the skill: vocabulary to describe patterns. How do you learn it? Yes, first of all, let's talk about pattern finding. It only happens when you read about it. True reading: you pick a subject and figure it out. Because if you don't have a word for it, you don't see it in the real world. There have been many such studies. There was some community on Amazon. They didn't have the vocabulary to define certain colors. So they couldn't spot the color at all. They saw green and blue in front of their eyes, but they couldn't see the blue color because they didn't have a word for it. So your vocabulary defines your perception. For example, if you are reading a lot about mental biases, cognitive flaws, then if someone is applying it, or if their mental bias is on, you will instantly see it because you have a word for it. If you do not have a word for it, your brain just rejects its existence. So, step number one is: read about it. And don't read about it through your general books; find classics. Books that have not become outdated for the last 50 years. Like Charles Darwin's book on evolution. First principles thinking: set your vocabulary. Practice pattern finding. It is just about writing and reflecting. You read about it, you wrote about it, you discussed it with people, and this is something that cannot be fast-tracked, like "tell me in one month to make me a great pattern finder." It's just not possible, man. I think based on the narratives of the last two years, I felt that our parents' generation would fall back a bit in the AI age. But based on what you are saying now, I feel they will be the forerunners. The irony is they have everything that it takes to use it to the next level, but there is so much fear, so much apprehension, that they are not using it. The parents' generation. Yes, those whose first principles are very strong. They are not using it. Wow! You are talking about corporate training. Yes. Wow. So in corporates, people are not adopting new work formats. Yes. And assuming, probably introducing them to tools, something. They were introduced to tools too. Firstly, there is a lot of fear, a lot of anxiety. They must also be afraid that what this thing is doing, I am also doing that. So there is also this fear: does that mean that now I have to step up? It's very difficult to digest that. Yes, Uncle ji. Right? Isn't it? Yes, Uncle ji? Of course, meaning, see, nature doesn't care about you. When any new technology comes in nature, there are three categories of people: opposers, builders, and users. And opposers are the first to be eliminated. So either be a builder or be a user. And I am saying, at least become a user. Okay. Now, on behalf of ambitious young people. Okay? Who want to grow at speed. I would like to ask you a question: If I want to grow my career at a very fast pace from April 2026 onwards, what should I do today? Based on the AI environment around me. So, bro, the answer becomes very finely tuned for every niche. So I will tell you the intersection. The advice that is common for everyone. You should keep the SHAPE framework that I told you firmly in your mind: "I must not forget this." Second thing, now you have two routes after leaving college. Either work for someone or start your own business. I don't recommend anyone to start a business right after college. It's better to work for someone. But pick a company that is already AI-native, whose founders are already adopting AI. If you go to a "lala" company or an outdated company right after college, then essentially you are moving in the reverse direction for 5 years. So, whichever company you are applying to, please make sure it's a small company, a VC-backed company, an AI-first company, so that you can be at the cutting edge of how the world is moving. If you want to start a business, if you are an agency owner, the simple funda is: to earn money in life, you earn money in life when you help other people earn money in life. If you can give something faster, better, or cheaper, these are your three options. Most businesses can only do two of these. You can never do all three. Now, you pick this. Do I want to be faster, do I want to be better, or do I want to be cheaper? AI can help you in all three. So figure it out. Whether you want to work for someone else, or for yourself. If you want to work for yourself, pick two of these three offerings. Search for a simple AI: which LLM is useful in my work? Create agents and reduce your service rate, and just be cheaper and better. It is the easiest way to win. Do you think more entrepreneurs will emerge in the near future, or fewer? 110%. Because there will be unemployment. Hiring has frozen. Firstly, people are not hiring that much. People are also firing. So what will these people do? People will be forced to start. They will be forced to start. And this is the evolution we are seeing: stop solving these foolish problems. One step up. If you are a freelancer, there are only three domains where you can sell AI services: sales, marketing, and product research and development. Not even product building, because many people think that I will just put an app idea into AI, and I will have a billion-dollar startup. The most stupid thing ever, because that product is no longer a moat. Even if you put an app in front of you through prompt engineering, if something goes wrong with it someday, you will have no clue because you have not built it. So, more than building, focus on how you can scale it, maintain it. I think in many people's minds, there is a cartoon of a button drawn by someone, where you just sit, press a button, and then the AI agent will do everything. Bro, most people think that AI is a vending machine, but it's a slot machine. You keep pulling in the casino, and only once in ten times do you get triple 7. People think it's a vending machine. Press the button, and it comes out. It's not like that. I told you, there is speed, there is density in this podcast. How are you liking it? Tell us in the comments. I really enjoy doing AI podcasts like this. I don't completely understand how deeply, at what frequency, you people want these AI podcasts. But this is the most important subject, man, in today's era. Very few colleges teach AI. I believe that if you want relevant education, you should listen to the talks of startups. Like Skill House, like The Cutting Edge School, which is Ansh's startup. Please educate yourself. Moving forward, like Fur. Such information, such knowledge will be shared with you. Please take notes and build your career further with this podcast episode. Skill House brings you TRS. Okay. Now, let's help the audience with some self-awareness. So how to rate one's AI literacy? In my opinion, the first question is: how do you write prompts? And I don't write prompts anymore. I speak prompts. So, first of all, I would recommend every person to start speaking their prompts. So that's the first level. The second level is: are you using connectors in AI tools, or are you stuck with just the normal, poor ChatGPT and Gemini? Meaning, booking.com has now come into ChatGPT. Photoshop has come. Many apps have come inside. So you can assume that many people have bought iPhones but haven't downloaded the App Store. So you can imagine where they are limited. They are limited to just phone calls and messaging. So that is level number two. Level number three is that start building basic agents either on Perplexity or on Claude. And I won't say building. Just start using them. Because I have analyzed your channel, my channel, all channels with Perplexity. It was insane. If I want to show you the report, you would say, "Bro, it would take me at least 5 days." If you are starting a business, you need to find sponsors, you need to find POCs, you need to find the email and phone number of the POC. Buy a subscription to Apollo in Claude. Add it as a connector in your Claude or Perplexity. This agent will understand your product, shortlist brands from the entire country, find the names of their POCs, find that POC's email ID and phone number, and based on what investments that person has made in the last 6 months, what kind of bets they are taking, it will create a pitch for you. In fact, you can also add objection handling: if this brand rejects my sponsorship, then tell me what I should do. So imagine how many people's work is becoming easier. Hmm. Okay. These are your inputs to increase literacy. So at least, for example, understand what is hidden in GPT. Correct. Can I ask you a question that goes something like: Explain the secrets of ChatGPT to me? First of all, the majority of people, I can bet on it. Okay? Tell me in the comments if this is true. Open your ChatGPT on your phone today. Click on your profile in the bottom left corner. Go to settings. Go to data controls. There is a setting there: "Improve the model for everyone." If that is switched on, which it is for most people, it means ChatGPT is already training on your conversations. OpenAI is one of those rare companies that are fully transparent: if you want your data to be private, turn this off. So, number one: most people don't even know that they can keep their data private. The same happens with Gemini. So whatever AI tools that you are using, please make sure to keep your data private. You can switch it off. That is one secret. Second secret is connectors. Many people are using ChatGPT and Gemini in their app. But if you click on "Gem" on the left side in Gemini today, you can make storybooks, you can make your own chatbots, you can make your own sales agents. There is so much that you can do with these GPT bots, with these Gems. What are Gems and connectors? Connectors are like your App Store for the iPhone. You are buying a phone. In the App Store, you download new apps. Those are your connectors. In GPT, when you create your own chatbot, it's called a GPT bot. In Gemini, it's called a Gem. It's not called a Gemini bot. But they are extremely useful because you can do repetitive tasks very easily. Like, for example, I am a very fussy eater. Right? And my mom gets tired of recommending me something to eat. So, I had created a bunch of GPT bots to recommend lunch items, dinner items. It's super easy to do that. You can upload your own preferences, and my mom can use that GPT bot. She has no clue about how it works. She just has to text that GPT bot, and the GPT bot will give her recipes. So you can do so much outsourcing. Hmm. Hmm. The next basic question is: if someone from our parents' generation is using GPT only for studies, is that also fine? 110%. Because you should learn AI to learn difficult things. For example, I am doing construction right now. I have no clue about construction. AI is helping me tremendously. We had to build our booth at the India AI Impact Summit. I knew nothing about booth creation. Gemini did a fantastic job at creating the entire booth design. How was the AI Impact Summit? Incredible. Yes. 100 out of 100. Like, I absolutely loved it. Why? We launched our sovereign models, by the way, which is a big day because sovereign models are very important for our country. Three models were a big hit: Sarvam, you have Bharat Jain, and you have Gyani. These are the names of the companies. Their models are something else. Which is a huge deal. And secondly, so much investment is going to be done in our country. Like, literally, 1 lakh crore, I think Ambani's are investing, to set up data centers, which is insane for our country. So it's like, as an Indian, this is the best time to be in the AI space. What should we build, bro? So it depends on your own capabilities. So Nvidia's founder, Jensen Huang, says that AI is a five-layer cake. The most foundational layer is energy, in which Ambani's are also investing: how do we generate so much energy, how do we create these data centers? So that's level one, which you and I cannot do. One level above that are your GPUs, which Nvidia is making. One level above that are chips, then infrastructure. This is very technical. But above that are models. Above models are applications. This upper two layers, models and applications, have a huge opportunity. Models, unfortunately, can only be made by technical people because it involves fine-tuning and training LLMs. But you and I can create basic applications that can help a business either save time or make money. God, basically, solve the problems of old-school businesses. Yes, but do it faster, cheaper, and better. Hmm. Okay, let's move on to the topic of AI tools. So what is happening now?

It's not that the fight between Anthropic and OpenAI is happening because there's currently a craze for Anthropic. There's a trend. Because Anthropic has launched incredible features back-to-back. They launched Co-work, launched Claude Code, launched Connectors. They are releasing them rapidly. And OpenAI is watching everyone very silently and patiently. But I know for a fact that they will 100% strike back. This is a recurring pattern for all these AI companies. Or, it's like basically your last piece of content defines your content creation career. It's the tech version of that. Yes. Yes. But bro, Sam Altman is a legend. Like, if you reverse engineer the entire story, he is the one who created this entire system. Like, he was the one who showed proof that this could even be possible. Wow! Okay. And you're saying that there are rumors that they are planning this whole hardware side of things. Yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes. With Jony Ive. Okay. They have made an official video also, by the way. Jony Ive, who used to, who designed the iPhone and everything. He and Sam, they officially announced that we are building something cool. Roughly, when will it be released? No clue. They haven't. What's your guess? I think it should be out within a year. Within a year. Yeah. So, in 2027, what can they expect? First of all, I assume those products will be expensive. A pendant, bro. I think it will be a pendant, 110%. Something that can listen to your conversations and help you create a second brain. Yeah, isn't that already? There are some hardware products. Minutes of meetings. Yes, but they will do it in a much better way because of this Jony Ive's credibility with your GPT models. But it will 100% be the second brain concept, that you can clone your brain. Because that allows you to become immortal. By the way, Meta has just bought a patent that they can run your social media even after you're dead. Yeah, wow. It's a proper reality now. We are all planning immortality. Oh, wow. That's the larger vision. But the definition of immortality is that I don't die my whole life. When I say immortality, it means that your consciousness remains on the internet forever. So, you as a human go away, but your thoughts and how you think in general, that stays on the internet. Like, I am already building my own second brain. I use Notion to build my second brain. I know that five years from now, I will be able to clone my brain easily. I think basically, if a lot of your content has already been uploaded on the internet by someone. If you have uploaded. A lot of your own content. You are pretty much a candidate to be immortal. 100%. If your transcripts, thought processes, and meetings are recorded. SOPs, this is what it's called. Whether it's video-based meeting recordings or something you've written yourself. SOPs is nothing but how you think about your work. All the podcasts, all the big actors, all the intellectuals who give a lot of interviews. But the irony here is that what these celebrities are saying in interviews, maybe that's not the real reflection of how they make work decisions. Okay. See, I want to make my clone so that the clone can take business decisions. I don't want a clone who will give interviews. What you're saying is that if we have to make an AI clone of Shah Rukh Khan, then his interview data is available, right? So, 100 years later, Shah Rukh Khan might come and give an interview as an AI clone. But if I want Shah Rukh Khan's brain to actually manage Red Chillies, even after he's gone, that is a separate kind of SOP training altogether. Which is happening actively with them. They can do it, and pendants can solve that problem, or you can do it manually on Notion also. Okay, okay. What about Anthropic and Claude right now? So, currently, you're saying that Anthropic and Claude are trending a lot because some of their products were very good. Which products were very good? Tell us that too. Majorly, Claude Co-work, because it could actually take control of your PC and open up your Gmail, create a proposal, send it to that person. But then what happened? People started getting scared that this is taking access to my computer. Then a new trend came where it became very viral that you deploy an open Claude, an agent that will go everywhere on the internet and work for you. Then there were two or three more gimmicks in between. I don't know if you have seen them. Molt Hub, you had like this website of videos that was serving to AI agents. They said that AI agents are creating their own religion, a cult. This was all a lie, by the way. I've looked into all of it. I've looked into Molt Hub. I've looked into Molt Book, which was supposedly the social media of AI agents. I saw the religion documents also. I have no clue who is doing this. It's all seeded. All these sensational AI news. A lot of this is live. Yes, because I have tested, I have opened them up. It's just gibberish. There are no AI agents talking to each other. Yeah, I think it's also one of those news cycle things where two topics gain a lot of views. One is the Ranveer thing. And the other is all these AI fake news. Yes. Right? These are. But keeping and the AI fake news stub. I think one of those hot headlines. It just catches eyeballs. AI does this now. AI does that now. Bro, fear sells. Okay. But these tools that Claude released. Yes, Claude Co-work. Claude Co-work. Yes. So, what does Claude Co-work do? It takes control of your PC and actually automates a lot of your work. But it was still limited in its scope. So, what happened? When Perplexity Computer came, it said, I can do everything that Claude Co-work can do, with one major difference. Perplexity Computer creates its own virtual machine, which is also called a sandbox machine. So, it never took access of your computer. It creates its own mini-computer. And it actually, you know, gets a lot of things done. And I want to show you the demo. You will be blown away. What are you showing? So, bro, I'm going to show you a very simple agentic tool now. There are two very big problems. One issue students face is, how do I prepare for the best of the best placements or packages? Or you can look at it for a business owner too: I'm starting a new business, tell me what are the best profits and guidance I should take? So, you tell me, can both be done? One can be done. Uh, no, both can be done. First, let's address the students. Okay. So, what is the problem statement? To get higher student pay packages. I need a higher pay package. Guide me. With 2026 context in mind. Okay. So, let's assume I'm doing B.Tech and I need a high package. I'm doing engineering in Delhi. Not even B.Tech, B.Com. Doing B.Com in Delhi. Okay? Now, the biggest mistake people make is they directly put prompts into AI. But first, I will ask the AI what I should ask. Okay? Because I do not know what I do not know. So, now what I'm going to do? Hi. I am a student doing B.Com in Delhi and I am about to graduate and I have no clue which job to apply. So, can you tell me in this month and even in the next six months, which are the best jobs? In fact, can you please tell me what I should ask in the first place so that I can actually crack these jobs? So, I've used a plugin now. It's called WhisperFlow. So, I just spoke it. Any reason you are using Perplexity? And because Perplexity is, first of all, an LLM aggregator, so I can choose any model here. I don't need to buy them individually. Okay. And any reason you chose Claude on it? Because Claude is outperforming in general. Everyone is saying Claude is giving very good results this month. Wow. So, this month, last month what happened? Gemini was very hot. But hotness, coolness depends on what? They have benchmarks, right? Benchmark results come out. Reports come out. So, in different evaluations, now this Claude is doing very well. Okay? So, now I submit this. So, now Perplexity has its own LLM, Sunar, but I'm using Claude. So, now you can see its thinking pattern. It's going to different websites. Indeed, LinkedIn, Instahall, it's surveying the entire thing. It has read 15 websites, and now it has the top jobs to target for March to September. Now, the most important thing here is that as a human and globally, I have to validate because this can be a drunk intern also. Hmm. So, I have to validate, of course. But now it's saying, what to study before interviews. Okay, let's go down. It's also giving a quick action plan. Now, I'll do one thing. Let's give it a very strong prompt. All right. So, I think I want to pick the jobs with maximum salaries. But I have zero skills. In fact, I have wasted my entire college life. And now I have like six months to crack this. Can you please guide me through this end-to-end process? How do you apply to these companies? What are the job requirements? In general, I am very good at talking and writing, and I have done a bit of, you know, co-curricular activities. But can you just find the top job descriptions? You know, just tell me what are they asking for? And also give me an action plan for the next 30 days so that I can learn these skills for free. So, only recommend me free stuff. And give me a prompt for this entire approach. Give me a prompt that I can copy-paste into my Perplexity Computer. Story over. So, now instead of me writing the prompt, I'm asking AI only to help me write a good prompt. Now, since this is a live demo, I'm not doing human intervention very well here. Otherwise, I would have read this prompt, applied my judgment, used my test, and improvised it further. So, you notice, it has again extracted all three bets: Business Development Executive, Financial Analyst, Investment Banking Analyst. It has said a strong fit, meaning, according to you, you should do this. Now, let's do one thing. It's writing this entire prompt. It has also given an action plan. But I think at the end, it will give the prompt that I'm putting into the computer. If you think this result is good, wait for the next five minutes, you will be blown away by what the computer will give me now. So, it has given me a complete week-by-week plan, and now an idea for a Perplexity prompt. In fact, I can put in so much specificity here, like what are my weaknesses, what are my strengths, this is my resume. Tell me, based on my current resume, what can I apply for? So, let's assume I wrote this prompt, and now we go to a new thread on the computer, and I've pasted this. Let's remove the quotes. As it's the same prompt. It's the same prompt. But if I want to do this personally, I would still like to read it and validate it. Now the game goes to the next level. Now you notice, I entered the prompt. First, what will it do? It will figure out how to approach this specific problem. But what is happening through the computer? So, now it is first deciding how many agents do I need to execute this task. So, it's thinking and analyzing. And the task is to create a roadmap for those three high-paying jobs and also create a roadmap where I can learn things for free. Okay, got it. I'm also thinking, in this entire process, what could go wrong for people watching this? And my answer to that same question is, I think people are just too lazy to give such elaborate prompts and to spend this person. First is affordability. This tool itself is so super expensive. How much is it? I'm on the max plan for ₹18,000 per month. 18? Yes, 18,000 per month. Hmm. But you can achieve similar results on the Pro plan also, which is cheaper. And the issue is that many students cannot afford this. This is the biggest issue. So, you can run the same prompts in Gemini also, and Gemini Deep Research also, because that is completely available for free. Of course, the results won't be as superior as these. Okay? Inside Gemini. So, for better analysis, let students use Deep Research. Yes. So, now you see what happened here. It said that to achieve this task, I will learn two skills. These skills are those SOPs. So, it has extracted one skill, Research Assistant, and its work is done with just one skill. Now, it has also said, I can see that we have discussed this before, Digvijay, because yesterday I was running for Digvijay. So, things, my name is Digvijay. So, now it has created a checklist. Slowly, it has run these threads in parallel. So, you notice these are different threads running in parallel. And if you notice it carefully, you will realize that even if I were to do this as a human, I wouldn't be able to think of all this. Because even I would not be able to figure this out. Then, at the end, you notice. It's reading a skill, how to make a PDF. Now, these skills, mostly, will be human work, writing these skills, because you can further improvise Perplexity's skills. While this is happening, I'll show you something else very interesting. This is Claude. So, Claude is also Perplexity's computer, and inside it, this Co-work has come. Now, the interesting thing about Co-work is that it can actually access your computer and actually get things done for you. So, it can achieve the same thing. If I go to the left side or to Connectors, then here are all the applications that a founder uses day in and day out. So, if you go to the category section and go to Data and Analytics, then you can directly link your YouTube analytics here. So that you can use Perplexity Computer for analysis. Similar for all the major applications that they have here. Any, any dangers? See, if you are using this for your business, then for every AI tool that you use, you have to check out the enterprise plans only, because with the enterprise plans, you get end-to-end ironclad security. Okay? So, whenever I give recommendations to any business or organization, because large organizations, above 1000 people, they have audits for this, that the applications you are using, do they have SOC 2 certification? Are they GDPR compliant? All these compliances are there that you have to fulfill as an enterprise. There, you have to only use the enterprise plan. But if you are a consumer, then you can use the consumer version and you can just switch off the training model. Okay? That's the safest bet. How much was the enterprise plan? It's usually 50% more expensive. Okay, got it. Not more than that. But it varies. And for every LLM, an enterprise. Yes, yes. Perplexity Computer also has an enterprise version. So, 18,000 is what you are using. So, basically 27,000. But I've never inquired about the pricing because they don't reveal the pricing upfront on the website. So, you have to, you have to, you have to, you have to talk to their sales team. It depends on the volume, so it's very subjective. Okay. But if you use the enterprise plan, it's all safe. Yes, yes. I mean, that's what they say. Okay. Flip side question. Suppose you are not a tech-oriented business, you are a real estate company, and you've heard kids saying, use an AI agent in your business. How do you begin? Again, there are only three areas where AI is helping me: marketing, customer outreach, reach. In general, product research and development. So, if I am a real estate developer and I need guidance on the map, I need to figure out what improvements we can make on this map. I am doing that already for my house, by the way. You can use AI as a second counsel. If you are trying to make a brochure and you want to make multiple brochures. So, I'll show you something very crazy. What we are seeing now, Perplexity Computer, is limited to text. Now, I'll show you an aggregator for images and videos. Okay? So, we use a product daily, like SuperYou, and let's assume I work in SuperYou, and I'm given the duty to create different social media carousels for different products. So, we have created this automation where if you put a normal picture of the product like this, then it will keep on making these multiple variations, and you don't have to individually do it one by one. Got it. Now, these are not production-ready. I'm not going to say that now you run ads on this. But imagine if a creator is starting from this, if this is someone's baseline, then compared to other competitors who are waiting for the crew, for a million things. I will show you something even better. So, this was for images. I put in three photos, I got different variations in operations. Would you say starting a business is easier now because marketing has become very easier? I think it's easier for a limited time right now. You have that arbitrage. If you exploit this information in the next six months, it's great. Because, you see, I put in these four products. I ran a prompt quickly. Now, see, it hallucinated that the wafer size is wrong. So, see, it's hallucinating. But then in batch processing, these kinds of visuals came out rapidly. Now, imagine how much time it used to take designers before to create this kind of result, even as inspiration. Now, if a creative team uses this as the baseline, imagine what a bunch of high-performing people can do. If they build on top of it, you know, today we had a session on our graphics team for not using AI enough, and they do. Yes, but to automate their work and make it quick. Yes. What is the future of graphic design? Yours is the past, right? Graphic design. Yes, mine is UX and graphic design, but in that, it will only be a game of my test and judgment because this execution is going to be human-led and agent-operated. Is it already human-led and agent-operated? Not yet, because the tools, you know, the issue now is. Everyone has a nano-maker. And think of it this way: if all the graphic designers in a team are only using nano-maker, then they have all hired the same employee. So, if they use that nano-maker image as the final finish line, six months later, all of their posters are going to look like each other. God. So.

They have to build a team that knows how to build on top of it. They are not doing this, poor things. God. Okay. So now it has also created a PDF. Now it has also put in an inspection for itself that I am going to inspect the PDF to make sure. Now, for example, if the rupee symbol got corrupted, it says that I have made the rupee symbol wrong. I am fixing that also. So it is auditing its own work. And this small part you are seeing, this clot code, it is deciding by itself which LLM to use. So if it thinks that Gemini is better for this task. GPT is better for this task. Then it is switching things by itself. But imagine if I had to do this myself. And by the way, we can run a parallel task also. So we can do the business owner one side by side. Do you have a business idea in your mind? Or can we go for a new IP idea? 100%, you tell me. So I want to plot the return of Fashion Bhaiya, but not for fashion and grooming videos. Fashion Bhaiya is just a name of the IP where I have my P2C treatment as an NPC to the camera. So I will just talk in front of the camera in a most serious tone and teach about different growth-oriented subjects. So this is the plan. Okay? Okay. So now, how to create this IP? From whom to take inspiration? What should be its structure? Ask AI once with references. Okay? All right. So assume that I am Ranveer Allah Badia, bear biceps, and I want to start a new IP altogether, inspired by Fashion Bhaiya. But I want to do it in a way that it goes completely viral. I want you to do some research on other creators, other successful IPs, find intersecting patterns, and tell me if I had to launch this across YouTube or Instagram or whatever you say? How should we do it? What kind of topics should we pick? What do you think is the sentiment overall in the social media community? Youth-driven. And it has to be youth-driven. We are targeting young people who are interested in, you know, self-grooming and self-improvement. So, let's do some proper research on this, so we can have a proper audit report. So I will change your name. Spelling is Ranveer Bear Biceps. Automatically it will detect. Everything else is fine. Let's click on submit. So the benefit of this now is that you can give both agents to work side by side. So notice, it has now opened the skill research assistant. So my own computer is not getting busy. I have multiple sub-agents working for me. We, we, what is actually working? It's the servers of publicity. Right? This is truly a very big crisis. Right? I really hope that they fix it on time because right now, a lot of energy is being consumed. See, as I mentioned, the five-layer stack, energy comes at the bottom. To keep these chips, these GPUs cool, you have to put in a lot of energy. So I think glaciers are melting because there, yes, because your, your heat is being generated so much. So much power is being consumed to run these models. But I am thinking that, you know, AI probably came precisely so that we could solve all these problems. By the way, there was a guy who had zero background in medicine. But he was able to figure out how to cure his dog from a terminal disease using these AI tools. Hmm. All right. So we have become from five LP and six months. Poor Digvijay has been caught. But let's download this, and you will be absolutely fascinated to see what kind of PDF this is. So let's open this up. Now, look at the formatting. Okay? First of all, it has created a role. Exact skills required. Top five companies. Now, for example, look here. I am not sure if it's Byju's or Unacademy. So here it has hallucinated. Most probably it is Unacademy. So here, look, there is an effort of hallucination, but it has provided the sources. So now you can easily validate. Then, free weekly study plan. These are my different plans. You went down. Top interview questions with the answers. So if you are applying for that thing, then it will prepare you for the interview also. Then it has done it for every role, whether it's financial analysis. It has given the plan for each. And let's go and see at the end once. For every single thing it has done. Then it has also given a 30-day quick start action plan on how you can do all these things. If you are not growing, if you are not getting employed, in the day and age of AI, it is your fault. Correct. No excuses left, right? Yeah. See, it helps you find signal from the noise. Yeah. You know, I hope that these skills reach the masses. But I don't think so, bro. I think the power law will increase even more. The gap between rich and poor will also widen. Because people are not learning. The biggest issue is, most of these AI tools expect you to prompt in English. So until the sovereign LLMs like Sarvam and others are natively integrated, or like this Whisper flow, until then, you cannot even prompt properly in Hindi. Okay. Okay. In the meantime, I would like to ask you, what do you think Mukesh Ambani sees Artificial Intelligence from? Bro, I am extremely impressed with him because the guy has decided to invest like ₹1 lakh crore setting up those data centers. That will be the foundation. I think the fact that they are going into that five-stack layer, that is the best move that they would ever have. And they are also opening up a bunch of, you know, free courses, education in general, because in India, there isn't much education happening right now. Practical education. Most of the courses that I see, even in colleges, they are like so outdated. It's pretty sad. I think he is older than 65 years of age. And I have heard him in multiple interviews that his passion is to educate India. So as people grow older, as they become senior citizens, it becomes much more about internal purpose. So AI-based education is one big domain. But why did you say that you are excited that 10 lakh crore is being invested? Because it is being invested in the right place. It is being invested in data centers, not in building apps. So that five-layer cake of ours, they are focusing more on the foundational layer. Like you are excited for India? Yeah, for sure. Because everything is happening in India. We have our sovereign LLMs all set up. Now we just need talented people to drive this ahead. Because if the infrastructure is there, chips are there, everything is there. But if people are not capable, then it is our fault. Hmm. Hmm. Hypothetically, if India did not have that much money and we could not set up such servers. Then we would have been far behind. In the world race. We would always be dependent on foreign LLMs. It is possible that, like tax is levied on goods and services, it is possible that after 10 years, people will also tax their LLMs. Okay. Okay. But if it is our country's LLM, then it is ours, right? It is trained on our context. It is trained on our language. So if it comes to servers, it is trained on our language. And the education system needs to understand this. Until now, what used to happen? In the education system, there were three components to teach any skill. One is 'What', one is 'How', and one is 'Why'. Most colleges only teach 'What' and 'How'. For example, if I go to graphic design college, they will tell me 'What' – they will tell me this is called a movie poster. 'How' – they will tell me to go to Photoshop and choose this font. But very few teachers teach 'Why' – why this font? Why this combination? So Gemini showed a demo of Project Astra. You must play it. If we can watch it right now also. It is a proper personalized tutor that can help you solve like JEE level questions. But it can do 'What' and 'How'. But the role of 'Why' will always remain with a human teacher. The problem is, the education system was built for scale. It was not built for personalized learning. It was optimized for profits and not for practical learning. So I really hope that they evolve. Otherwise, its value will keep decreasing with time. Hmm. And then there is another question, to continue this tangent of AI businesses in India. If you ask many people today, in which domain would you like to start a business? Even if they don't have an understanding of AI, they still say AI. That's a very stupid thing. Why, bro? Because first of all, half the people think that putting a chatbot or an LLM on their website is becoming an AI startup. Because it's a buzzword. But they are not realizing that this will become table stakes. If today I build a website and say that my website has images, that's a big deal. So 25 years ago, that was a big deal. So I would recommend any business owner to figure out what your niche is. What is your expertise? And just do it better, faster, and cheaper using AI tools. Correct. Hmm. Okay. Just be the absolute most affordable best option available. Hmm. Okay. So a very common business idea that is floating around these days in the world of agencies. Like all the video agencies in Bombay, Delhi, Borivali, they say that they will all be replaced in a year or two. No, bro, they will probably be replaced because two agencies will take all the customers instead of five agencies because they will be the first ones to adopt. And this trend keeps coming. Like there was a trend that AI influencers will come and like real influencers will be replaced. Which was like absolutely a fake lie. Right? So what do agencies need to do? They need to ask themselves, if I can provide an end-to-end package with agents. So, if I am a thumbnail agency, instead of just giving thumbnails, if I can promise five variations and also the analysis of the A/B testing and give them regular reports. Then that's a groundbreaking offering. So as an agency owner, you can offer a lot more at a much cheaper price. And then just eat the entire market. Hmm. Okay. How do we go about that? So give me an example of an agency industry. Any agency you want to do content, marketing, posters? I think most of the agencies now that at least I come across all do content. Yes, content. So, just say that you make short-form reels for any, your agency's job is to make short-form content. So the number one step would be to figure out that you pick a specific niche. Let's assume healthcare. And you specialize in scaling content for healthcare brands. Hmm. Step number one would be building a team who understand the first principles of how content works. Like many people think that I will go into AI and ask for a script, and that script will touch 1 million. What a stupid thing to say. Because a reel or a content becomes successful. On top of that comes the topic. If the topic is trending, then it's good. Then comes the writing. The core essence of what you are trying to say. Then comes sound. Then comes your camera lens, then comes the camera itself and the lighting. If a person has zero clue about the first principles of how healthcare content is working to your scrutiny. So step number one is build that team of aggressive pattern finders. Second is to figure out what are the deterministic problems in my entire workflow. Deterministic means that, brother, in the last 5 days, this thing is trending on this surgery topic of healthcare. Some celebrity had some open-heart surgery. So now everyone is interested in what happens in open-heart surgery. So now I do pattern finding. I put five Instagram scripts into perplexity. I ask it, tell me the recurring patterns. It will tell me, brother, these four-five patterns are coming in every reel. Now I will sit and write the script, building on top of that. Now it said that, brother, I don't have time to edit. So I need good B-roll. A B-roll that I can quickly put on top of the A-roll. Now, the aesthetic of that B-roll needs to be locked. So I went to NanoBanana. Said, tell me two-three aesthetics. Chose one aesthetic. Again, depending on my taste. I know that my audience will accept it. Then I went to HugsField or FreePik. Quickly made the B-roll. Put the B-roll on the A-roll. Reel finished. Now, if a person is doing this normally, they are sitting and reading scripts, sitting and making B-roll, going to Vato, downloading stock videos from all over the world. And it doesn't even look good because everyone's reels look like that. This is where you outshine and you deliver a one-hour deliverable in 15 minutes. Same as, you know, 10 years ago when we started Monkey, we used to solve only for speed. Yeah. Even today, in the age of AI, you are saying solve for speed using AI tools. So another enhancement is to solve for velocity because velocity is speed with direction. Right? So the speed has to be in a very targeted way, that we are running, but is this needed? Because this is going to be a very big problem. People will have an AI tool to literally solve any problem of their life. But they will end up solving the most stupid problems first. It's very similar to how most lottery winners go bankrupt. When you get so much power, you become unconscious. Hmm. Okay. Oh wow, a lot has come your way. So what's happening? Gen Z sentiment. It has almost made it. Latent is also coming. Of course. But it takes time, but the insights that come are worth it. I am sure. I am going to be downloading this file from you. I will send it to you. You know, I also feel that modern PR is going to become about what these LLMs say about you. Yeah, yeah. That's search engine optimization. Now it's evolving into generative engine optimization. Generative engine optimization. Yeah. That what do AI tools say about you? I am also optimizing for that. How do you do it? So everywhere, whatever your Instagram URLs are, or whatever is in your LinkedIn caption, make a description, make a template, that you have to write this at the end of the post because these are the cheapest ways to influence these AI tools. Hmm. So if you want to put a certain tagline next to your name, just make that copy-paste across every reel, every reel, and every YouTube piece of content, and LinkedIn, and wherever it is. Hmm. Okay. It is the best way. By the way, AI was also used in Dhurandar. Yes, and nobody even knew. What was this? See, when the movie Dhurandar was releasing, they released quite a few music videos. They have used HicksField templates like Khatakhat, and nobody was able to figure it out. You can tell, I can easily tell. I know the exact template also. That explosion template where the camera turns like this. But people don't care that much. If your final output is good, because they didn't just like literally put something in AI and just use it. So another movie came, I don't remember its name. They forgot to remove the Gemini icon from their trailer. Bottom right corner. That is bad because then you are using AI for the finish line. So you are saying that in Dhurandar, the AI use was done so well, it was integrated so well? Blends with the story. It speaks detailing by Aditya. But do you think movie making as a business is going to get fully overturned? No. I think people's taste and benchmark will go even higher. Because by the way, the creator of Tumbbad, he tweeted yesterday, he has launched a proper movie trailer. It is completely AI-generated. He made it on a budget of say ₹30,000, which is insane. Because Tumbbad took a lot of time simply because of pre-production and post-production. Yeah. You know, that's my conjecture on where content is also going. I hope to see a big web of fiction-based content, which is accelerated through AI. So are you watching those AI slop videos? Those that nowadays a Coca-Cola can or a vegetable says, my name is Coca-Cola, my name is this, I am getting some R-rated ones. No, I get them too. Let me tell you, I get them too. By the way, they can be made for free using Google Flow. If anybody is curious, like, what is this happening? They are made using Google Flow. Hmm. Hmm. The game is to build deeper, richer content. You know, fully built. Actually, I would recommend that you make more non-AI generated content. Because, see, there is a difference between attention versus respect in this. AI-generated content can get attention, but attention is very cheap. So it's not monetizable. To get respect, you have to do something which only you can offer. Hmm. And only respect gives you that distribution which is monetizable. Hmm. Now, what is the issue, bro? If AI slop is uncontrolled, and we start putting any nonsense on the internet, then in the next three years, the next generation that is reading will be reading AI's vomit. So by the way, there was this insane case where an AI did some research. It wrote a research paper, but the citations were also AI-generated. So that poor AI was creating a new research by reading some research by AI itself. Hmm. Hmm. Hmm. So we have to make sure that we are reading the right content as well. That we are reading the right people. I would not recommend reading from AI at all. Then your brain will definitely slow down. 100%, it will become brain dead. So what is the best way to learn quickly other than? You are reading something? Yes. You didn't understand a concept? You go to AI and say, explain this to me as if I am a five-year-old. By the way, both ChatGPT and Claude have an interactive section now, where if you are not understanding a physics or math concept, it can create proper simulations. Proper simulations that you can play with. No student ever had something like this. Where is all this going in 10 years? No true unpredictable can predict. Yeah, but the gap between the rich and poor is going to be massive. Because people will not learn. I used to think three-four years ago when ChatGPT came. I used to think, bro, this is insane. Like people will definitely learn. Nobody has learned anything. Nobody has learned anything. You know, I have also noticed this. I have been in very big rooms with big celebrities where I tried to tell them about prompt engineering a year or two ago. No one gave a shit. They are all celebrities in big rooms like, it just doesn't concern me. Yeah. And I was like, okay, maybe it's a celebrity problem. But I am sure they represent.

A large bunch of the masses. Yeah, of course, because people think it's not that good. Most people think it's not that good. And there is a three-level problem here, right? Most people are either using the wrong tools. If they have the right tool, they do not know how to prepare it. Or the third, which is the most difficult, they are solving the most stupid problems with AI. They are solving non-deterministic problems. They don't even know where to apply AI and where not to apply AI. Because the education system is not evolving. And it's not the education system's fault, you know. Because I can't blame a chancellor or a principal for not updating, because the organization is so big. How will you catch up, man? What do you think of all these? Like MBAs in AI and all this stuff, that AI is being taught in big institutes. No, I don't think they teach AI. I think they teach, like, if you said MBA with AI, first, I would have to see the curriculum to know what they are promising. First of all, it's, of course, it's a clickbait, right? Because they just want to show off. But the second thing is, I am so curious to see that are the teachers updated? Like, are they still stuck on Gemini deep research, or do these teachers even know that there is something like this in the market? I really think that for someone who is 50-60 years old, probably they are also turning to the Gen Z only. Yes, this time I have been very lucky that when I go to like a corporate setup, they are very happy that they are learning from a young person. So, earlier we used to struggle with this, we used to think that nobody is trusting us, let's probably get like an old IIT professor. Right? Maybe that will help us get credibility. That here the tables have reversed, or is the practical thing to do all about creating some sort of a text slash artificial intelligence department in your own organization. Correct. I, yes, I will tell you what the process is. This is like tried and tested, by the way. First, train the entire company on the basics. Give everyone tasks for a month and shortlist your AI champions. These are people who are inherently smart and curious about AI. You set up an AI lab within the organization. These 10 people, these AI champions, have only one task: to be regularly updated with AI, test use cases, whatever is working for them, drive the same adoption across the company. But what do many people say? They ask for shortcuts. Give us a Chief AI Officer. Give us a CTO. That's just not possible. Where will you find a Chief AI Officer with 10 years of experience when the subject itself is only 100 days old? By the way, in my classroom, I have had an 8-year-old kid and an 89-year-old guy also. So, when an 89-year-old guy is ready to learn, then age is not an excuse. It's a mindset thing. Wow, 89-year-old. Yes, I have taught him. Yes, I. Wow, wow. By the way, this is your entire audit. Executive summary, market opportunity, fashion deep dive, comparative landscape, 10 universal viral patterns, and a bunch of interesting stuff. So, this is about the industry, and I will definitely send this to you. By the way, he also analyzed on Reddit to see what can work on Reddit. Hindi and English are non-negotiable. So, your assumption is absolutely correct that you should do Hindi. Hindi or English first. Correct? But another quote is, you can't really trust because AI tells everyone, you are right, six white spaces, no one knows science, back men's skincare, men's fragrance, content, integrated gloer protocols, tier two, three budget, style in Hindi, fashion deep dive. What made fashion work? So, he is doing pattern finding across years. I think, yes. What went wrong overextension? I know better energy and content dilution. The lesson, the persona works. Budget first, India-specific advice works. What fails is ego, overextension, and the gap. Then comparative landscape. This is the Indian creator ecosystem. He has put all your competitors together. Then global viral IPs analyzed. He made a report of that. Let's go down 10 universal viral patterns. Meaning, here is where your game comes in, that you sync this pattern finding with your own pattern finding, that what he is saying and what I am thinking. And think about how much time it would take if you were to create this yourself. Hmm. Then platform strategy, long form, shorts, reels, and threads. Then he has also created a platform architecture. Audience sentiment, what kind of sentiment is going on across Reddit and all of these other apps. So, he has also told about Reddit threads. What Indian men want, and what they are tired of, what they actually want. Hard water hair. Such a strange topic. Because of hair loss, basically. In interior India, you mostly get hard water. Okay, got it. Yes. So, it's so like specific, right? Then you have your name direction options, glo up, sharp clean approach, content pillars, glo up. Then he has also given 30 high-virality video ideas. So, again, out of these 30, you can assume only five will be good. Now, which five those will be will be decided by your experience, your pattern finding. Okay. Basically, just places you that initial ball. You telling an intern, "Let's get into writing zone," and you going from zero to one. Correct. The baseline is raised. Makes your zero to one easy. Okay. Mehra bhai, what do they call you? I think Hans bhai, he is, perhaps. Okay, okay. I think we have almost reached the end of the show. A few final questions for you. You go to different offices. Is the most common problem that many people are rejecting AI? Is that the most common thing you see? Rejecting. Even the young are rejecting. They have a bias that how can it replace me? Then when I open my laptop and show a demo, they realize they are using the wrong tool. Hmm, hmm. So, their benchmark is flawed. Based on everything you learned about AI agents and where AI is going, do you think there will be downsizing in many companies? Now, I feel that people who are firing now are not doing it because of AI. They are firing to reduce their own operational costs, and they are taking the name of AI. Like, you know, everyone is trying to benefit from the situation. That's what's happening. They are blaming AI for it, but I don't think AI is smart enough that it can replace 10,000 people. So, I think they just had to fire people, and now they are putting it on AI. But I can see hiring freezing for sure. Hiring is going to freeze. 110% because they will realize that we already have enough people who can do basic-level work. So, either a boom is coming for startups. Many small startups will come. Which is again a great thing. We will just have more entrepreneurs in the country. Hmm, hmm. I don't know whether this is a bias on my part. But in my opinion, all the editors, all the graphic designers, they are in some kind of a step ahead of the rest. Because I feel that their job is always related to end outcomes. They have a vision in their head, they are working towards it. So, there is clarity in their process. That clarity can be translated into the domain of AI. And AI first gave incredible results in visuals. That's why a lot of editors and designers were super excited about it. Like, at that time, Midjourney was so popular. By the way, last year, Midjourney was everything. Now, nobody even talks about Midjourney. Hmm. Okay. Do you think editing is also going to become just writing? At some point. It's already happening, by the way. So, in Descript and Premiere Pro, there are plugins where editing is happening at the transcript level, not by watching the video. So, you get the transcript of a full one-hour podcast, and Claude tells you, "Take out these snippets." So, trimming is happening. But what people think is editing is trimming and cutting. But true editing is all about deciding what will stay and what will not stay. Hmm. Right? That is a difficult skill. Hmm. Editing is not like, "I have downloaded an After Effects template and, you know, added fractal noise, and made it look great." That is motion graphics. So, I think editors need to now understand retention, AVD, understanding metrics, having been more data-driven in their decision-making. If you are doing that, you will be so far ahead in the creative industry. So, actually, technically, every established content creator also has a serious edge in this new world. 110%. Pre-built distribution is a huge moat. AI can't replicate that for you. Hmm. Got it. Sure. Okay. I think this was the podcast for today. We covered a lot very quickly. Any signing-off notes? Because in my opinion, the right way to do this subject is to actually have an AI-themed podcast every two months because the subject is evolving so much. Am I right? Correct. What is your gauge on the next five years? I would like to tell all the audiences that if you write down the handwritten notes from this entire podcast, you will remember this for a very long time. So, please don't use your NotebookLM and that ask feature to get a summary out of this. You only have three things to do in your homework. Homework number one is watch the entire episode. Make your notes. In fact, put them on social media. I would also love to read what you are doing. Second is, whether you use Gemini deep research, which is completely available for free, or if you have a small budget, then definitely try any LLM aggregator. Perplexity, Hicksfield, or FreePick. Hicksfield and FreePick are for creative people. For normal business owners, Perplexity is more than enough. So, that is homework number two. Homework number three is that whatever task you are doing, whatever you are understanding, practice it. How can I build on top of it? If AI is giving you a career roadmap, specifically exercise yourself. How do I build on top of it? Otherwise, intellectual obesity will get you. This actually makes it the final question. Right. Intellectual obesity. You are basically saying, study, apply, use your own brain. Correct? That's never going to go. Because there is another angle, which is, what is the lazy way to become rich? Yes. And I think it's kind of a valid question in some ways. You know, you should focus on work smart over work hard. Yes. It's a good way of thinking, but it's also not fully practical. Correct. Correct. So, you are saying that because of AI, our brains will become a little lazy, become slow. This has been a recurring pattern in history. When the Industrial Revolution came, people started losing muscle because in the pre-Industrial Revolution lifestyle, there was manual physical labor. Now, what is happening is that if AI becomes very prevalent, then the brain will slow down. So, if you look at all our species, gradually all our body's physical assets are depreciating. Muscle loss was okay, but now intellectual loss has started. Now, to avoid that, the best thing that you can practice is to throw new problems at your brain. Every day, try to solve an interesting problem at your work. Read more, write more, because when you write more, you reflect more. And the habit of reflection is extremely valuable, by the way. I have been writing for the past five years, and that is the reason why my vocabulary is so sorted, because I have become so good at articulating my thoughts. That quality is not available in every senior management, or a VP or a director right now. Right? Because a lot of people in their current organizations have risen to the top because time has compounded, and not their skill set. They have spent 20 years doing something, but they have not spent 20 years doing different things. Right? So, don't be that guy. Hmm. Okay. Cool. Hansh Mehra, thank you. Anshu bhai, it was fun. So, very fun episode. I am just a little tense in a nice way for my own organization, for all the upgrades that are happening in the Gen Zs that I work with. Okay. And I want this to be a very sharp change. I think actually, we are already a very AI-friendly organization, but even then, I feel that we are not using it enough. So, if we are feeling this, I can't imagine what's out there. Imagine, bro, like the whole world doesn't know what to do. Can you imagine how big a crisis this is? People are taking loans to get degrees whose skills have no value. Yes. Flip side question. Okay? Because this has been on my life lessons. When I was in college, I took advice from 40-year-old, 50-year-old people, whereas I should have taken advice from 30-year-old people. I think you are roughly 27, right? What would you say to 12-year-old, 13-year-old kids? Especially their parents. Correct. So, I already teach AI to kids from 6th to 12th grade. Right? And I tell these kids that, first of all, avoid using AI until class 10th. Right? You can use it for games, but don't play too much with AI until 10th. So, for kids from 5th to 10th, I tell them, "Use AI for creative purposes." Like, if I am in 5th grade and I want to put myself in a comic book next to Captain America. What a beautiful instrument, right? What a beautiful use case. If I have a picture and I want to bring that picture to life, brilliant instrument. Right? You want to make an app, go make an app. But if you are using AI to solve your homework, if you are using AI to write your essays, it's such a bad thing. Right? So, I think parents need to instruct their children on where to use AI correctly. School principals and headmasters need to understand that you cannot deny AI. By the way, I am so proud to share this that IIIT Delhi, I think they were the first ones to do this. I think it has happened everywhere now. IIIT Delhi said that you can now use AI in your exams as long as you are submitting the prompt that you have used to get to this answer. What a brilliant move. Wow. This is how it needs to be, man. So, what do schools need to do? There are some schools that are rejecting AI. Huge mistake. That you cannot use AI. We will block AI. They are going in the wrong direction. Others are saying, "Open ChatGPT, and write prompts like this in ChatGPT, and you will get the answer." This is also stupidity because they are teaching it as the finish line, that accept whatever AI gives you as is, blindly. Both approaches are wrong. Imagine we have millions of students either ignoring AI or learning AI in the wrong way. Hmm, hmm. The right way is to also write the prompt that was used. Correct. Show that what you are thinking. Your creative thinking is reflected in this. Man, maybe, you know, exam questions will have to develop more, become more elaborate for you to come up with good problems. Gen Z has to update itself. At this point, all of our examinations are testing the "what" and the "how" behind things. They need to test the "why" behind things. The value of objectivity is gone. Every modern Indian exam, at least. Globally, it's a problem. Okay? This is not just an India issue. Globally, metrics because they are built for scale. Every old-world educational institute is built for the old world. Yes. Correct. Not an old world anymore. Right? Since the last two years, at least. Correct. Okay. Anxiety-inducing podcast, helpful podcast. Thank you, Anshu. It was fun. Hope to see you. We chill all the time. I will probably see you soon. See you. See you. So, basically, before every podcast, we do a human activity where there is an intention setting. Okay? Okay? My intention is, first and foremost, for my own team, that they get a lot of education. And I think that when you try to address a small cohort, you automatically address a larger cohort. So, personally, in my heart, my intention is this one is for my own team, but I know that the public will also be helped by this. Okay? So, do you have any intention through this entire piece? Bro, my intention is that many people are using AI at a very superficial level. So, I hope after this podcast, they have a clear call to action that if I want to actually change my life, how do I use AI? Okay. How to get started? Oh.

Get Our Mobile App

AI MASTERCLASS (HINDI): ChatGPT Vs Claude, Agents & More - Fast Practical Podcast | Ansh Mehra I TRS

Transcription