0:00 On Twitter, CEOs are saying they are dead. Every day, you read news and you know that I'm fired tomorrow. 70% of workforce of US is white collar. White collar means software engineering, rest of it is blue collar. Carpenter, plumber and all that. White collar exposure is almost 80%. 80% of the jobs in today's world can be automated from AI. In blue collar, it's almost zero. If you look at data based on education, more degrees have, their jobs are higher on risk. How do you find a job in AI? Actually, let's do a very simple thing. Let's go to Claude. Claude has a connector which is basically called as Indeed. Now that I have Indeed, what I will do is, I drop the resume in Claude
0:45 and I will be like, okay, so you are an expert career counselor. What I want you to do is, and basically it will give me a proper list of jobs that I can apply for. So now I have found 5 companies to apply to which I have to apply. How do I apply? So here what I will do is I would basically go to Codex, the link of the roles. What happens in Codex? I can give it access to a folder from the resume. I'll just open that folder so it has access to that folder. And I'll be like, hey, I've given you access to my resume's cover letters. Now, I've told what it needs to be done. You have to go to plugins and there is something called as computer use. Now one crazy thing is that codex has access to
1:32 All your apps on your computer. And it's so beautiful when you see it working. It is just wild. Look at this. What Codex is doing is it's filling details on indeed.com for us. He's doing himself. Insane. And how is this information? From his resume. I haven't given him. I'm in Bangalore. This is my PIN code. I haven't given him. He's doing himself. Yes. Understand, I've landed two in five months. How do I do the interview preparation? Okay. So let's say, uh... किसी को भी अगर यह डूढ़ना है कि आपकी skills के according दुनिया में best jobs कौन सी है जिसमें सबसे जादा best पैसे आपको कैसे मिले उस पैसे को आप बढ़ा के सकते हो और रोज उन jobs के बारे में
2:21 आपको पता चलते जाए उसमें आप apply करते जाओ और उसके लिए prepare हो जाओ ताकि interview में आ� without any hassle with genuinely probably 5 minutes in a day, 5 minutes in a day, you don't have to do it yourself, you want to do AI, this is the episode where we will teach you how to build that automatically, you will apply AI jobs for a day and you will find which is the best job you are going to give. I have made an AI-focused community where AI is in-depth teaching. And if you have to learn to fully utilize AI to make
3:09 your work faster and better, you can join this community. Link is in the description below. It's completely free. I read a tweet in which you are talking about Mark Zuckerberg. You have written Mark Zuckerberg is reportedly building an AI to do his own job. So the CEO of the world's most powerful company is making AI himself so that he needs to do his job. What is that? So this is a thesis across. So this is Mark Zuckerberg, not all of the Sam Holt. They need a proof to tell the world that everything is replaceable. And the best proof for that is by replacing yourself. I think it's a statement
3:57 more than execution but what we are making basis of what we understand is a variation of the thought process, He has made an app, an assistant, which is his AI clone. Basically, what is he doing? We talked about second brain last time. He rent out to his employees and his concept is based on what we have read. If you come to me, before you come to me, go to my bot, he will refine you because
4:44 my brain in a way operates there. In the same way. After refine, you have to come. But that's what he's saying, his idea is how can every single employee of Meta access my brain? But will Zak do this? And the company will not be sold then. Why? Because only the unidirectional one brain will be used. No, this is access. No, you're right. If everybody is like, I'll ask Zak's bot first, I'll ask what I'll do, then probably that will be done. But this is more of access, like I said, this is more of a statement. than anything else that this is possible if my job is automatable how much is your job worth it's a statement it's like replaceable theory I am replaceable who are you and Sam Altman has
5:32 also said one thing no I think AI will be able to do my job before most of the jobs because CEOs decision making thought process use they are all usually backed off data and what is inference that is based of data points, CEOs and that is why it is a very replaceable thing most of the layer is replaceable thing but the part is not replaceable, it is your CEO, it is your own, it is mine and it is the taste everybody has their own taste you know there are people who do their jobs who do their jobs like tools, i mean i have to do this step 1 step 2 step 3 step 4 step 5 and i will finish it those are your average people in your company who
6:21 don't think those jobs those are the jobs that are in risk but when you have taste when you have your own thought process coming in when you are an original thinker when you are trying new things out when you are not doing a work like a tool you are going to be fine in fact you are going to thrive at this point of time because the thinker has the ability of doing something new and the taste has What are the person doing? They don't enjoy the job themselves. They delegate and they are doing A.I.K. through. So they are able to do a lot more. True. Interesting. Then you had a tweet about replacing jobs. It was about task and purpose. Yeah, this is exactly what task and
7:11 purpose is. Exactly what I was trying to say. Task is sending an email. purpose is why the hell are you sending that email so you think why the hell are you sending that email, this will not be replaced because that's the purpose that is what you do the job for but because this is a part that is confined to you, this part that is what you do yourself it means that when you look at the world most of the people are doing tasks in the name of a job that is where the fundamental problem is I don't understand If you go look into your office, look at people who are doing tasks like step one, step two, step three, step four, those are the people who
7:57 are in risk, who can replace because today I rely on the taste that this is a problem, how to solve it and it will solve some new ways and take it, those are the people who won't be replaced and how do you develop taste? It's a great question actually, how do you develop taste? I think there could be a lot of ways for you to develop taste. One way I believe is by getting yourself exposed to more things. If you know all these possibilities automatically you build your own taste, you find your calling and its simplest framework according to me, which I have done for now is experimentation. There is no yes or no. Either it works or it
8:43 doesn't. There is no good idea, bad idea. Either the idea works or it doesn't. So I know you also operate like that. Right? We do a lot of experiments. As you do experiments, you build that muscle of taste. But I think the first step for this or experiment is constraint. If there is no constraints, it can't develop taste. there has to be some constraint, constraint either time or money or where will we pay rent tomorrow, you know there has to be something, job is going on, something like this is constrained, where it has to be, that is not coming until people don't put their brains in experimenting and doing something faster or better and because they don't do that, they're not able to
9:32 build yes and no, white and black, good or bad and in that case, taste doesn't develop, like in those They don't have it in them to experiment a lot. No, I think it also has to do a lot with the purpose and hunger. That comes out of you don't like your life right now. I mean, it's become philosophical. The situation is become philosophical. But this is true. The current version of your life, you, okay, like is, let's say maybe a strong word. You think there's a better life out there. which is way better than the current version of your life that's ambition right that's ambition so are you saying that people I mean obviously if you are someone who doesn't have hunger who doesn't have ambition who just tick boxes I'm
10:23 saying that they are tasking not that they don't have to pay bills they also have to pay bills but they just don't enjoy doing what they are doing maybe they should find their purpose maybe they should find what they enjoy doing but that takes us to a very different tangent of tangent altogether so you are saying that the monotonous regular tasks without a purpose will go or will go yeah you absolutely believe that everyone will go like twitter on all CEOs or news in all tech companies are saying that they are gone Every day you read news, you know that I'm fired tomorrow, I'm going to go, I'm going to go, forget employees, I don't work at all, I feel like I'm
11:12 going to go. Bro, I don't think we should put this on a position of fear at all, but I want to talk about this on a position of realization. I want everybody to realize this, that companies don't work for your benefit companies work for the beneficiary of their shareholders they have to give return and value back to their shareholders and they will do whatever it takes to make that happen right now let's talk about what is happening okay two examples will pick up first you know who Jack Dorsey is right ex twitter I mean founder, I call him the
12:01 founder, right? Last founder. Yeah. Now they're only owners. So Jack Dorsey, after he moved on from Twitter, he built something called as Block, which is Square, which is very popular fintech you might have heard of in the US. This tweet shook up the internet and everyone else 64 million views on this. What was this? His team had 10,000 people. His business was growing faster than ever. And the reasoning is this, we are making this tough decision because not because we are in trouble, our business is strong, gross profit continues to grow, we continue to
12:49 serve more and more customers profitably. But something has changed. We are seeing that the intelligence tools we are creating and using paired with smaller and flatter teams are enabling a new way of working which fundamentally changes what it means to build and run a company. Just, I got chills when I read this because, what is it saying? It's saying that it's not your fault first of all. Business is good, AI is not bad for you, it's growing up, growth, profit is growing. But we have seen that those people who are using AI tools are able to considerably build and do so much better that we don't need so much.
13:41 But there is a caveat here. What is that? The caveat is paired with smaller and flatter teams and we are already seeing that when people are working, when smaller teams are working with these tools, so what does it mean about the other 4000 people? That they were not working with these tools. Yes! They were not expanding themselves, they were tasking. Now I tell you what is event raised here? What? Meta laid off 8000 or laying off 8000 people, Oracle laid off 30,000 people, India is also very effective. 30? 30, 30, 20% of their workforce. globally including a ton of people in India okay, the story is the same, business is
14:29 doing great, best quarters ever, highest profitability ever but laying off people, why is this happening? this is what the scary part is, large companies, there is one side of the story where they are like efficiency has increased, we have no need so much, they have given people, who are doing good, who are able to multiply, they are continuing to have jobs the other thing that is happening with oracle, meta and all of that is they are laying off people at meta is also almost 10% 15% something oracle is 20% taking that money if you lay off 10,000 people that billions of dollars of money saved in salaries they take that money in and you know where they are investing in AI infrastructure they are buying
15:14 more GPUs and AI infrastructure so what is it? What will happen from other computers? It's a bloody vicious cycle. It's a cycle over cycle over cycle. This is one side of the story. Now, the thing that worries me a lot is every time a company doesn't do the right thing, the market doesn't reward them. Here, the market is rewarding. Boss, it's a leak, the stock is 10% increasing. And 10% is not increasing in 100 million companies. You get my point. Trillions of dollars. I mean billions of dollars is being added.
16:05 So, it's on us how we want to conclude of this. But things have started to happen. This has started to happen. But here's a question. Like 10,000 people, 4,000 people came. They said that we will retain 6,000 people. We fired 4,000 people. They fired 4,000 people who were able to make a smaller team with AI. They could not be able to make AI productive. Here's a question. What are the children who are still alive? What are the others doing? What are the others doing? You wrote something about observed exposure. I didn't write it but I'll tell you what it is. Is it related to
16:51 it? That the job of observed exposure can be done? No, I'll tell you what observed exposure is. In fact, today we do the tool that Jack Dorsey said, we have built internal tools with which people are able to do more work. That internal tool is open source and I've been using it for two years. Nice. I'll show you that. Okay. First, observed exposure, then what is it? Yes. Observed exposure, is basically a coin that is turned by anthropic. They had a launch of income index, which observed word coin. They went through hundreds of jobs, looked at what people do in those skills, and looked at what percentage of jobs can be automated using AI. Let's
17:38 say software engineering. Software engineering using AI can be automated. Actually, how many AI are using inside of software engineering is observed exposure. Okay. What is the potential of AI versus how much are people using AI for? Okay. So, in their math, on an average, what is happening? This is based on US report, by the way. This is fascinating, if you go deep into it. Okay. 70% of workforce of US is white collar. White collar means software engineering, We work in computers. Simple. Rest of it is blue collar. Carpenter, plumber, etc. White collar exposure is
18:24 almost 80%. Potential exposure is 80%. 80% of the jobs in today's world can be end to end automated. It has been done, it has not been done. Observed exposure has been reduced, but actual potential exposure is much more. In blue collar it's almost zero. Yes, I mean how plumber will do that? Correct. Now if you go deeper into this, if you look at data based on education, the more skilled people are, the more degrees they have, their jobs are higher on risk. So it's not the value. Now people will say, you're getting a degree, you're getting a degree, I am going to go to I am low. Right? We are not. The reality is that job skilling is happening now. Right? Now
19:10 this is the high level, this is what observed exposure is, potential of AI is so much and actually you have not seen it, it will be seen as a graph. Nice little web graph. Very very cool, it was completely viral. Are you seeing it? People think that it doesn't come to the spelling. which is true, this one, this became very popular, this is the observed exposure, many people
19:57 have read it No, the blue that is a theoretical AI coverage, that is what AI can automate. So, I mean, my job can go. All of them can go. But the red, that is observed AI coverage, I mean, how much push has been? AI coverage. Right? Now, what happens when the exposure or the coverage becomes maximum? Okay. A ton of jobs will go, because AI can automate all of them. This is US, we will sit in US and do what we will do in India. Now we flip in India. This is not made of study. So I was surprised
20:45 that I would have to apply India. So I told myself that I have taken my tax data. How much tax in India comes in? Around 45 lakh crores is the total tax. I have done this digging. Total government revenue of 44 lakh crores. Out of that, 30 lakh crores. Direct and indirect tax. In that income tax alone is 16 to 17 lakh crores. These 16 to 17 lakh crores are coming in me. How many people come in you? 2 crore people. 2 crore people pay 16 to 17 lakh crores. What do these 2 crore people do? White collar jobs. Wide collar jobs and we
21:30 are seeing exposure, if it is software engineering outside, it is also software engineering, let's say 85-90% of the jobs actually vanish. How will the government fix the deficit of 15 lakh crore? US will not be worried about it. I'll tell you why. Why? Because I told you about the conversation with Alex from Meta. Tell me. Where we spoke about, I asked him this question. Alex who is head of AI for Meta. I asked him what happened in the world. He said one simple thing is, look, when we start a job, universal basic income will come. What is it? How will everyone get the pocket money? How will the government get the money? So these AI companies, they will tax the
22:18 government 65, 70, 80% on that money. but it will give people as universal basic income now this works for US great open AI, Anthropic, Google, all companies there China also understood this but Chinese companies started to make Chinese companies now let's talk about India where do we stand jobs will go but are there companies which are AI first which are killing these jobs in India to tax If there is a universal basic income in India, then how will it happen? Who will give the companies? And how will the companies earn money? Because who will buy the companies? If there
23:04 is no jobs, then there is no consumer. That is a new thing. What do you know about that concept? There is another theory for that which is universal high income. Where this is said that we are watching it in a silo. We are just watching the internet. the biggest bottleneck is in energy, that is what is costing money, we are all paying money for everything because of one thing, because of energy everything works, if energy costs are negligible, then everything will be negligible, because there are robots, which can do most of the work in 5, 10, 15 years from now, there is AI which is very very smart, so the factory is running, the robot will run, the brain is AI, but
23:53 It still costs one thing, energy, electricity, where will it come from? That's why if you see majority of the investments, Sam Altman has not, he is not part of, and he doesn't have a stake in the open aim, but people have invested so many investments, in which one? Nuclear fission and fusion, because that is the next leg of energy. Energy transition. On top of that, if you can see, Sunar Pichai, Elon Musk everybody is talking about Dyson swarms. Dyson swarms is now solar. If we want the energy that we want from solar, the amount of land coverage that we need is insane. And there's a lot of
24:39 loss of energy. So Dyson swarms was a theoretical concept. Imagine sun is like, sun is saying, you are almost covering the sun with Solar, in a way, not solar but in a way, that is Dyson Swarm, okay. It's called Dyson Spear by the way. This is what Dyson Spear is. That's the sun and you put a thing around it, which is capturing the energy of sun to the closest. And data centers, all these GPUs that need data centers, will also be in space. Okay. So if you look, Jeff Bezos has invested big money to set up data centers in space. Google is talking about data
25:26 centers in space. Elon Musk said, I have SpaceX made this. He has forgotten about people moving to Mars. He's talking about data centers in space because they've all realized the money, the fuel is energy. Now this is theoretical by the way. This is pure. Because the whole cover is big. So swarm It's a small brother, so think about 3 or 4, it's not so big, small part, small part, small part, small part, small part, now they are thinking about do we have to go to sun or we can go somewhere else, how close do we go to sun, it's for material science investments, because things will go so close, how do we transmit inference, now in cloud, AWS or Google cloud, it's hosted,
26:13 it's time, it's latency, But when it's in the sun, like when it's in space, how do you transmit energy? That's why tech is also being made. They're talking, a lot of investments are going in. On the earth, there's nuclear fission and fusion. And then there is this moonshot. Crazy. That is Dyson. So if this happens, if energy becomes free, the whole world will go into devaluation. Inflation will not be deflation because everything will fundamentally become free. Cheaper. when fundamentally everything becomes free then you're watching Netflix Better Life you're living a life of fun where everything is free right but this is the good side of the story
26:58 but people will buy what? what will they buy? what will they buy? what will they do? exactly that is the big question right now the questions that people are asking on the good side of the story is humans always look for purpose in life Work gave them purpose. Exactly. Solving complex problems actually kept you sane. People say, you know what, we will go back to the medieval times. When your work, when your life was family, was chilling, was hobbies, was painting, was art, was writing. You find your calling. They say. But in a distant future, I think there is a possibility. I mean,
27:43 in theory, it's good in the air. It's like a concept, it's like it's possible. But does it make a good side of the story? There has to be an ugly side of the story. The ugly side of the story is of the hungry people. They're like, no, this is mine. I've made it. Right? There could be that side. And that is what the world fears. If China has... So, you know the concept of AI, AGI, ASI, right? Artificial intelligence is what we are using now. When AI becomes smarter than theoretically, I mean AGI definition is very loose. Artificial general intelligence is an AI that is smarter than every single person in the world together and can literally do anything and everything much better than every single person in the world. That is AGI which people were chasing towards.
28:31 The reason companies chase towards AGI is because when a company reaches AGI They unlock very smart thing. They unlock something called a singularity. What a singularity. When AI can develop itself without AI researchers. In the world, the whole AI wave is moving purely because of maybe 1000 to 1500 people in the world. Probably not even that number. All these are AI engineers. That's why Mark Zuckerberg gave 1 billion dollars for one person. These are AI engineers. They are the highest value people in the world today. Which is more AI engineers than smarter brains, they move faster. Now, when the AI engineer
29:17 itself is an AI, then we are living in singularity because AI is self evolving and making itself better. When AI can make itself better, you basically have unlimited AI engineers working 24-7, leading to something thousand times more powerful which is super intelligent and that is ASI so this is a wave curve okay, many people defend it too some people say Sam Altman can't be said that there will be a point of time where there will be rapid acceleration now there will be rapid acceleration so I can't imagine what it actually is but Daryo Amode and all these people say no there will be rapid acceleration the moment AGI happens Because self learning has
30:06 been ended end to end. Then you grow so fast that no other catch up will not be able to do. That is why every single lab is behind one thing that we have to do AGI. So that the brothers will reach AGI, they will be the first set of people to hit ASI as well. And you will have such an unfair advantage over everybody else. That you are driving in Ferrari and everyone has a cycle. So what race are you racing? So that is the game that's happening as well. Okay, let's come down to today. Sorry. We were talking about observed exposure to India. We were talking about that. Can you open that again? Yeah. So we were at
30:56 observed exposure. So how many jobs are going in India? I mean white collar jobs. Whatever is there. Now, There is one more data to this. So India, you said that there is no company that will pay for India. Correct. So what happens I don't know. But one smart thing India has done. I used to complain. But I think now it has done a smart thing. India has always pushed every country. If you need tax, if you want to pay for tax, open your company here. We have something called OIDAR. Where we put GST, even on a transaction, from an international company coming to India so they would get extra 18% that was like a tax that was being applied, because
31:43 it is in a way a tariff for digital products so because companies have started to build their basis here if they build that basis here, value can be utilized here for example open AI, usage of data centers usage here I feel that's a smart play India playing that data center card I could not understand initially but when you look at the big picture that makes sense right now because where data centers are data centers are not for India, for region so all the dollar value will flow through this of course tax holiday for next 10-15 years, after that tax will be done so there is a good side of the story which India has played there as well though we were not able to build our indigenous AI which I am still hoping that
32:30 we will make something because a lot of work happening but we have in a way figured out a way to be part of that value chain by doing data centers and why is that a story? Google is investing close to 15 billion dollars meta is investing they're building after us largest data centers in the world are being built in India in Vishakapatnam Why does one mean that during covid they built out that you know internet infrastructure under water wiring and all there is sport and India gave those subsidies where tax already it's a herd mentality dude like one google came and it's finished. Who is it? Meta is coming and all these guys are building there
33:20 are conversations that are happening. Everything is happening. is called as the AI capital of the country right now, it will in a way become the AI capital, second AI capital of the world. Just like Hyderabad had become the IT capital of the world. Then there's something about this chief minister. Yeah, man. I don't want to sound political. But yeah, no, but it makes sense. Chandrababu Naidoo, he, because the IT boom in Hyderabad was here and now, if you are talking about Vishak, if you are talking about the second revolution, if he has played such a substantial role, it means there is something he understands. Yeah, something he understands is more stoned. He plays the tech game like a boss. Because
34:06 Hyderabad, out of nowhere tech city, nothing was there, it was nothing. And today it's It's giving competition to all top cities in the country. Probably is better than most of the top cities in the country. Hyderabad is the second highest AI talent density pool in the world. The highest amount of AI hiring that is happening is happening actually in Hyderabad. In the world right now. And the foundation was led by this one. You know how much of work of AI work happens in Google of India, it's crazy. Many good teams are sitting out of India. Microsoft has a huge pool of talent sitting
34:52 in India. I think strategically, Anthropik, OpenAI, all these companies are putting their shops in India also because of talent pool density because initial talent is made in Silicon Valley. After that, that is only comes to India. Last wave was also done. So they are seeing that way possibly to happen right now as well So hopefully So there's scope for jobs as well Yeah scope for original thinkers Scope for original thinkers that is exactly what we're talking about Now because we're talking about that there's one other very interesting thing okay You know Karpathy? No I've heard about it about Karpathy's job Andres Karpathy was the co-founder of OpenAI Okay Was the head of AI of Tesla If a brother tweets, he
35:41 talks about it, he is that credible. As simple as that. He built something called as a US job market visualizer. What is in this? It's very simple. He mapped all the jobs of US. Okay. And basis of the jobs, he said, what is the exposure to AI? Total wages exports in the US is $3.7 trillion. 143 million jobs are exposed to digital AI. Now the thing that is exposed, look at this. Customer service representatives, 9 out of 10 exposure. That is the thing. Green is a good job. For example, construction laborer, no problem. Continue hustling. Janitor, continue your hustle.
36:26 General maintenance, all good. Carpenters, all good. Now look at the red ones. Information clerks, financial clerks, receptionists, bookkeeping. Material recording, customer service, secretaries and administration. These are one side. Accountants, 8 out of 10. Management analysts, all MBA people get into this, 7 out of 10. Project management, 7 out of 10. Financial analysts, 9 out of 10 exposure. Market research, 9 out of 10. You see the darker the red is, the higher is the quality of talent that we were always told it is they are the highest paid people, highest paid people, software developers 9 out of 10 again guys this is coming from Andresh Karpathy not from me I'm reading the report right
37:16 computer support specialist blah blah blah you can see this but overall on a map if you see where green is cooks 3 out of 10 I would like to ask Sanjot to intern and now this is a very interesting thing which is education so you can see this which education is affecting the basis of that also is there no degree doctorate which channels are they falling in so that is a breakdown here as well and the same way monthly median pay what is their impact again you'll see there's an indirect correlation higher paid people have a highest chance of impact now when I saw this report I was like why did they make it for India So we have researched some
38:02 of the things we have done and we have made India job market visualizer right what is India job market visualizer well it's the same thing it's the same thing okay same thesis but for India by the way I want to make this clear this is what I mean by building right I built this like three hours before we started How did I build it? I went to perplexity computer I just said this is Andresh Karpathy's report Okay Look at this This is the world that we are living in The report is this This is Karpathy's job thing You can go check it out What I want you to do is I want you to do extensive research of data of India
38:50 That is available And build a very similar exposure dashboard with data of India Because I don't have to do all of this because the hard work is already run by Karpathy He has spoken about thesis, how did he research it, what would I do now? Task, what was the task in this? Figure out all the things that Karpatri did, what are the similar report, crunch the data, visualize the data, but what was the purpose here? What does this data mean? This is the outcome. As long as I know what the outcome is, I can build it and that's what I said in a single prompt, right? It was told actually one hour before and we have the report ready. Okay. which basically talks about 127 occupation groups in India and the same thing where our exposure actually is 616
39:38 million workers but in high exposure it's 42.6 million which is 6.9% why? Blue collar white collar split India's 2 crore people are saying 4 crore people are saying the tax paying now if you look at two graphs next to each other right The US was red, almost red. But India, this is what was... Okay, 143 million, 616 million more people who do work on an everyday level. But exposure, which is in India, is 42.6 million people. There is 90, wait, 34.1 million, negative outlook. But it's okay, sorry. Here is not mentioned
40:26 in the count, here is added in the dollar value but okay. What I want to look at it from here is India looks all green. Because majority of the egg bag look at this. Crop growers which is agriculture. Animal producers. Shop sales people all okay. Animal producers all okay. All small low cost jobs are staying in India majorly. Blue collar is 80%. Now These are all our white collar. Financial professionals, software developers. What is the most? Keyboard operators, clerical support like banking, secretaries. So if we look back one step, it looks
41:12 like India is all green, all good, but the parts that are red. The 2 crore people are spending 17 crore tax, they are still there. Yes. Yes, that is the problem. But I think our 100 crore problem is not our 2 crore which is progressing in the country, which is increasing for freebies, which is increasing for tax collection, which is going to be going to be the people. That's what we are seeing. Now this is cool. What we'll do is we'll link this up for people to play with it. If someone wants to see that in India which jobs can go through AI and will have more impact on this chart we will put links to all of this,
41:59 then people can just click it and download it done done so yeah that is where we are with job exposure I don't think anybody has done it outside interesting okay here's the thing so last time you said that jobs who are doing jobs, will make our own episode. So I went out and asked everybody. So employees have all set of problems and questions. Okay. What are their problems? Are there in their life, in their professional life? Are there problems with me? Or are there any particular boss or manager? I don't know. This is a general.
42:46 What are the things? So I'll ask one-one. Because maybe these are all problems that most of the employees will be in their career, in their jobs. Okay. And then one by one we'll try to solve it. And then we'll make it generalize later. Okay. Do you want to do generalize first and then specific or do you want to do specific first and generalize? Go with the flow, I don't know. I don't even know what these are. I'm a little concerned now. Okay. Okay. So here's the thing. First is, if your boss is micromanaging you, Okay. How can you use AI to show progress and reduce constant check-ins? Wow! Who is this boss? I don't know. Well, I don't micromanage. Sure.
43:36 Actually, I don't. I'm sure this is not fair. For me it will be that my boss doesn't care, so how do I care? That's it. My problem is other problem. Okay. Okay. If this problem happens, if If there is a very shrewd manager who is trying to micromanage everything The problem with micromanagement is that we have to give updates I will basically wipe code a dashboard I will basically wipe code a dashboard of progress Now I have made a perplexly computer One task, one task, one thing that I am working daily That end of the day report manager will be available so that they didn't ask me what happened today, how did I
44:21 work, all that stuff. Because here's a problem which I think is a good problem. Many people think that if you don't sit in the office for 10-12 hours, you don't work. Which is absolutely wrong. And unfortunately, many people are in front of the bosses. So, I'll tell you what I'll do. If they're doing something, they should give daily progress report. First of all, I think if your boss expects that, You should get the hell out of the job. Okay. World is too short. If you are affecting it, means you are putting effort, then it's happening. So you will find a better place. But what you said, I will give a tech solution. Okay. And to make a birthday report. What should be a birthday report? Someone has to
45:10 keep seeing what you're working on. Right? So, I would go to something like Omi desktop. Okay, what OMI desktop is, it's an app. Okay, it's an app which you can download on your computer, it's in macOS, in windows, it's open source. Okay, what does this do? I will not download it, but what does this do? It looks at, look at this, look at the screen. It basically scans everything that you're doing on the whole day level. It is made so that you were doing something and you forgot what you were doing. Or you saw a website and you are losing context of that. You want to remind yourself that three hours ago I was watching a podcast in it, I
45:55 was talking about something, what I was talking about, I remember it. So, OMI desktop is created to capture everything that you are doing on your computer 24 hours and makes that into your memory. So that you can ask a question, what I was doing. I sent a message to someone by making this report. I don't remember who sent it. Omi please remind me. So it will remind you. I will do Omi's desktop. And Omi will tell me, how did the whole work have done today? I made a document. Nice. It was not made for your boss, it was made for you. But you are using it for your boss. Nice. Pretty interesting. Nice.
46:43 And you do this? What? For yourself? No. No. I'll tell you, no. I'll tell you, no. I have a lot of time to say. But I have optimized it. The reason is, inside of codex, there's a similar feature. Where I spend a lot of time to build products, chat, GPT, everything. There's a feature in codex. which keeps watching my screen to understand context so that's not a problem I use comet browser in comet browser if I can just say hey could you open that website which had data for jobs
47:29 and impact of all of it I was checking that tab sometime back I can't remember what that is Instead of Comet I could just drop it something like this. Comet is my job, my browser's history. So it will understand what I'm talking about and it will open that tab back for me. Nice. All automatically. So I do this in micro instances. Instead of 24 hours. Instead of 24 hours because I know how to get out the information. So for me I don't need an extra app of Omi desktop per se. But someone who's using multiple apps, my work is in Chrome, so Comet is doing that work for me already. Comet is by the way really
48:16 good. I've started using it from our day one. I can't go back to any other normal. ChatGPT Atlas is also very good. I tried Atlas, I think it's good now, it's good now, the entropic is bigger than chat g. Yeah, they're saying 60 billion dollars in revenue right now, the growing like wildfire but when OpenAI dropped GPT 5.5 I was a huge clod buff, huge entropic buff because I was very disappointed in OpenAI, what are you doing bro? Why are you not shipping? But Codex GPT 5.5 dropped, I've stopped using cloud completely. I've moved completely to codex. In fact, if I
49:02 want to codex, I'll take three hours just codex. I've stopped using chat GPT, I've stopped using cloud. I just use codex. What codex? Codex is an app which is designed for developers to build products. But what they did, because it is built for developers, they have transitioned that into everybody's app. What do you mean by that? So everybody can use it right now. Only developers, what was happening in Codex? So you can do anything, it looks very developer centric. All the code is seen. Now they have MCPs also added which is connectors. They have added their own version of skills. They have added their own version of agents. I don't know if you saw a tweet or not. I got Codex a
49:52 job. of building content OS, which I showed you, that glimpse, I'm making it right? You know how I work with it? At night, before sleeping, I come and I say, okay, what did we accomplish today? What are the things that are needed to be done, which will take a lot of time? A lot of tech debt, data-based cleanups, migrations, technical debt, I'm learning all that now. I'm learning to be a software engineer without writing a single line of code. Not that I wrote code any time before but I thought this is very important. So there will be a lot of tech debt. I will say, I have to execute this, this, this, this, this, this. Don't stop till everything is done. You have access to everything. If you are getting stuck somewhere, you have
50:39 access to my browser. If you are still stuck somewhere where you really need my help, you have access to Wapi. So you can make a phone call to me. and I'll wake up if you're not able to move but don't stop and run autonomously. Every night it runs for 8 to 9 hours without stopping. I had a tweet that screenshot and Greg Brokman has liked it and bunch of open AI people retweeted it and whatnot. 9 hours an AI, GPT 5.5 in fast mode. Fast mode means it works 2.5 times faster than regular AI. GPT 5.5 high Basically it's like the best AI coding model ever in the world is running 8 hours all by itself. That's equivalent to I
51:26 believe 10-15 software engineers work for 15-20 days. It's going overnight. Bro, by next podcast maybe content or else will be ready. And I will show that. It's a personal app that I built for myself. I am so bloody surprised. I'm someone who cannot write code. I'm so bloody surprised, that I can make this product by myself. It is bloody complex, I know, that is a proper SaaS app. I mean, a team will take a 15-20 people for the product to make this product. Easy. That week-end, 2-3 hours prompting you to go and multiple agents are parallel working. It's insane where we are
52:13 heading towards. It's wild. I am doing something else that can be interesting for people. You know every chat GPT and all of these have something called as instructions. My codex instruction. This is what is custom instruction. This is okay. This is chat GPT instruction, some old one. Leave it to me. There is also a codex instruction. What did I give instruction on my computer? I have said I don't know how to code. I'm not a software engineer. I am building because I like building. But I'm also building because I want to learn and I want to know how to think like a software engineer. So when you're doing something because Codex keeps telling you I'm doing
53:04 this next, I'm doing this next but it talks in a technical language. So don't talk technical language. Talk in a language where I can understand what it is but also tell me the technical word for that so when I wipe code, when I solve my problem, it is also showing me the technical language and a simpler way of explaining me like I'm a 10 year old So I'm not only building by getting AI to do it for me but I'm also learning in that process Why are you saying that you should first design the database before designing the front end We have a product, we have UI, AI said that 80% work is in the back end, I'm like what do you mean by the back end? Explain that. Why
53:51 do you have to do the database? Why can't we do the UI? I think UI first, I said because I'm a visualizer. And that is what differentiates you from a great software engineer. You have to think architecture first. He is studying, I am teaching as I'm building a product. I'm getting better. 1% better every single day. So all these little things has nuance. And when you talk, when you explore things like this, in fact, why do you have to do this for only the questions that you don't understand? I fundamentally go and do this for everything. If there is any idea, I talk about it. I explore. Tell me what more can I do? Tell me what are the things that I'm not thinking through? Tell me why this
54:37 will fail? interesting people should just keep talking yes like a lot yes tell me basic problem because you're talking about basic problems another big basic problem of people is that there is a meeting, you have to come to the meeting for 20-30 years, you don't know what to do, you might have to present something, prepare something, at least okay present after just like I want to be prepared for the meeting I have asked for the last 15-20 minutes, how do I prepare the meeting AI? Let's say meeting is about last month's
55:26 growth growth. Let's say brainstorming on this and present your data and you have to do everything. I mean, what are you doing in the job? I'm just making it very very basic job person like. I mean if I don't even have data I don't know what to prepare. That's a fundamental problem. I want to use it by removing all of it. So, you say I need data first of all. Data is for anyone to use that. That's a social media case. You are a social media agency. who does media for creators and you have a bunch of creators now I don't know if you know this if you go to business.facebook.com there is something
56:14 called creator marketplace I know this creator marketplace basically gives you data of all the creators that you want you didn't know this but the world didn't know this So I don't know if you have... I'll tell you another one. Okay fine, tell someone else, tell your name. I'll tell you anything. Okay bro. I'll search possibly not. Now what is crazy about this is that you can literally see the data of a creator, which was only creator before. Right? You can see the followers, you can see the reach. accounts engaged, this is reached,
57:00 accounts reached not impressions. Right? And not just that, these are insights that was only available to you as a creator. Demographics, countries, cities, everything. Everything. Right? This is actual data. I mean, you'll be like, for example, here is Varun, right? It's a friend, so we can open, it's fine. Right, so I know I'm not Varun. Right. But look at this, I can see all this data. 4.6 million people reached. Right. The age, the demographic, third biggest country for Varun is Pakistan. Yeah. For you also? Yeah, yeah.
57:48 For me also? I don't know. But you get the point, right? This is some insane data of a creator that is available to you. of any creator for that matter. So the reason I'm showing you this is okay, now that we have this data, what I can do is, let's say, so I will just take a, I don't know, full page screenshot. It's an extension that I have, which is called GoFullPage. You can scroll the whole page and you can take it like screenshot. Nice. Okay. Whatever data you have, I can download it as a PDF. It's called GoFullPage for people who are I will go to a cloud or a chat GPT or whatever. Okay. Dump it right
58:35 inside. Hey, this is Instagram data of a client of mine who I'm helping with social media. And I have a call with him to make, to have a conversation with him to talk about his progress in the last 30 days. I'm not prepared at all. But I've shared all the data with you about the creator. You have his name and also you can happy to research about him as well. Do all the research and build me like a, you know, super extensive document of about full performance of the last 30 days with soft and hard data that you can find on the internet and from the screenshot and everywhere else around it so that I can have a solid conversation with the client for 30 minutes, which will be insightful for the client. as well. And towards the end, give me like 10
59:21 talking points that I should stress on and spend most of the time on, which could help my client grow faster and also position me like a super smart guy. I mean, pretty straightforward and simple stuff, right? No rocket science. But this is what orchestrating is. Being able to understand how to get data, what do I do, how to put data, what do I ask, That is basically what we need to do. And then you just read this just 15 minutes before the meeting. You probably read it during the meeting. Until you say hi, hello, when someone is talking, you do it. I have 10 talking points. I mean, there is no such trouble that you have never seen. You just need 10 talking points. If you get
1:00:10 high level points, you will talk. Fair. Okay. Another problem is If you want to automate repetitive tasks in your job, where should you start? Repetitive tasks? That if you have to start that task every time, then how would you start? Okay, if your tasks are on a computer and uses internet, then I feel the simplest platform for you to build, you have to build an AI agent for this. The simplest platform for you to build an AI agent is called a Zapier. You have heard about it. Many people will listen to Zapier but a new product is Zapier Agents. This is Zapier Agents. This is text to agent. I will
1:01:01 just say that I have to automate this and it works in Symfony. For example, I'll give you two solutions. Tell me what you're trying to automate. Daily trending news to build content pillars on. It's a very repetitive thing that somebody has to do in my team. Daily trending news related to our domain and then giving that as a piece of let's say usable dialogue or usable note for me to build content on. Great. I like this idea. Okay. This is a very simple agent actually and we should build it. Now, where does the daily trending news come from? Where does the step
1:01:46 number one come from? Google News, correct. So I need a way to pull the news from Google News and there will be some topics, this topic, that topic, let's say you create content on politics, I don't know, just making it up, politics. So you want to be sure that Indian politics, whatever content is on internet, you are able to, your AI is able to read through it and is able to do this. So you need a platform for that. What is that platform? So for this, a good platform is Feedly.com. Feedly is a platform which aggregates all the news. It's a news reader. I know. It aggregates all the news. Now, Feedly also has an AI extension. Okay. So what I
1:02:35 can do is, and I know it's on Zapier. So usually when you have to automate something, then we see that it's not in the tools. Feedly is a platform. So my first step what I'll do is, I'll say Feedly Zapier, I mean, there's a integration in Zapier, okay, great. Right. So I have to make an agent and this report should be used in my email everyday. So what I can say is I can come here, which is we are on Zapier agents first bot and I'll be like, what I want you to do is I want a news aggregation on Indian politics every single day. I'm a creator who creates content on politics every day morning at 9am on everything that
1:03:22 happened over the course of last 24 hours. I want you to pull all the critical news that has happened in the last 24 hours on Indian politics from Feedly by using the right keywords. And as you get that information, I want an AI to be able to go through all that information, cut out noise, put things that matters. and make a list of 5 news articles or 5 summarized points of each one of them in form of a paragraph and send all of that information to me on my email exactly at 8.30 am. So over the course of last 24 hours you aggregate the information find the top 5 pieces put them all together send me an email and that is the agent that I want you to build.
1:04:08 That's it. I'll explain what I want. I'll click on create agent. You know what Zapier agent will do? It will automatically look for the connector. Ideally it should be able to find Feedly, it should be able to build that layer of AI in the middle, it should be able to build the layer of Gmail also. That's it. It's a simple look at this, it has started to ask me connect your Gmail account because I have to go here, connect a new account. But you get the idea. So that is what it does. And then you send an email. So once the whole thing is done, what you will see is, this is the flow. You have built the flow right now. Right now it's asking,
1:04:56 set up this tool. In fact, what is it done? It has a web search tool. So it picked that up directly. And the knowledge source is that it can be mentioned here that these are the kind of content that I create. These are 10 topics that have gone viral. These are 10 topics that have not gone viral. You can put that information to make it better. And that's it. Your agent will be built. You just have to publish it and your agent will be built. It's the simplest way to get started to build very, very simple agents. And of course, right? Everything that we mentioned, right? It converted all of that into a prompt. That's all it did. And it connected tools and it gave it tool access. That's it. Can I do something like this for the newsletters that
1:05:43 I really like? There are like bunch of newsletters, they are 15 days, 10 days, on my email, some random person. I want to make some content that day. I have subscribed to Webhub, I send me once a week. I really like that newsletter. I want that there is a daily newsletter. Absolutely. I can just tell him, give him knowledge center, example. I'll tell you how I would go about it. I'll tell you the problem state. I'll tell you how I'll solve this problem. What is the problem number one?
1:06:28 That newsletter looks good. So first thing that I'll do is find more newsletters. What is that? That's the same one newsletter. Then it's a different, even easier problem to solve. So my random name is like Chamath, one week. No, like staying ahead, which is my newsletter. I'm going to spend weekly three times. I need you to do it daily. So what I will do? I will go to that newsletter. I will download 10-15 samples of those newsletters. I will dump it. inside of a cloud or wherever and I will be like understand the DNA of this newsletter what they create I really like this what is the structure of this newsletter give me a DNA once it gives me the
1:07:15 DNA I will go to an agent like this and I'll say this is the kind of content that I want to get exposed to every single day so build a newsletter like this for me by doing all this web search and do all of this put together build me this newsletter and send me an email every day at 9am The base work here is not getting the newsletter. The base work here is how do I want that newsletter to be? Do you understand? Yeah, yeah, yeah. And then you'll be able to do it because I have seen that it's repetitive. One time, but it's not fresh ideas. As long as it has new source of information. New source of information. New source of information can't scrape yourself. That is exactly what you're saying. Every day, read up the internet. find what is
1:08:04 working basis of that tell me information don't make up your own stuff in fact there is a much simpler thing that you can do for this as well right like for any of these things you can go to co-work of Claude okay it's even simpler actually okay and what I can say is hey I'm really interested in learning about AI and there's a lot of AI news that is coming every single day. I follow up newsletters like Staying Ahead which keeps me updated with all of AI but Staying Ahead sends a newsletter once in every three days. What I want is I want the AI newsletter to land on me, land on my email every single day based on whatever has happened in the last
1:08:49 24 hours. What I want to do is do a quick research of the internet to find everything interesting that's happening in the world of AI and bake that in into a newsletter. Even better what I can do is, even better I told you right, I have to give some samples. So I'll open some samples of my own newsletter. Here are some samples of what I like.
1:09:35 And of course I can save it as PDF also. I get it, it's faster. Yeah. So what this will do is co-work. The reason why I'm using co-work is it'll create a newsletter for me in the first iteration. After that, because it's co-work, I can set up a scheduler. So this task that I have to do every day at 9pm. So once I streamline this process and get to a format where I like the output, that needs to be done for one time. After
1:10:21 that, I can set it up as a scheduled task so every day at 9 am it will do the task and it will send me the information so it's almost like Zapier agent only of sorts what's better Zapier agent or this? Zapier agent has a lot more connectors for this simple use case I have Kovac but if my use case is more complex let me tell you what I need data from websites from websites I need data from tweets I need data from Reddit threads. I need data from product hunt. I have five different sources. What will that happen? Then you go to a Sapir agent. Because it will have connections for everything in some form or fashion. It's a much more complex thing. When do you want your daily newsletter to land? 7am. What
1:11:11 format do you prefer? Staying ahead. What should be the newsletter be delivered? Okay. directly to me what topics matter you the most I will say all and it has started to build it up for us and then it will be made and send it to me okay coming back to this coming back to this yes so the repetitive task will be made through agents and ZPI and you can use look at this it's all done it's all done yeah it's all created it's saying basically connect and then I just have to click on schedule and every day 7 am it will drop in my inbox as simple as that another thing which is actually forget this basic for people,
1:12:01 simple things that are the biggest problem for people, like in jobs which is how to find jobs in AI and which job to find, which job will be good for me all that, how do I do that? Very interesting. High paying jobs. Okay. So, let's make this hypothesis, that you've become very good with AI. Okay. And we know that we're going to be talking about a lot of jobs. So I want to get into a job which is resilient to AI. Forget about what I do today. Because there is no difference. If you do something, I need
1:12:46 to go to this place or this job that will affect the AI. Actually, we do a very simple thing. What anyone can do. Let's go to Claude. Claude is a connector which is basically called as Indeed. You know what Indeed is, right? It's a job platform. Right? We just want to connect it. I have to keep it connected. Now that I have Indeed, what I will do is I drop in the resume I drop in the resume and I will be like Okay so you are an expert career counselor What I want you to do
1:13:35 is I want you to go through my resume, you can see my resume and I am learning a lot of AI today I have become a lot of proficient in figuring out AI. I have seen a lot of podcasts. I have seen a lot of information and I have tried a lot. And now I have to switch to AI job. Which is resilient with AI changes and I don't want to go. So it doesn't matter what I have done. But my business exposure is now. The job exposure is now in the resume. And I want a job which is of AI first. Also another problem is I don't need coding. But I have AI agent, wipe coding, everything I have to do. I am an AI generalist. So you will find such a job for me. And also I can speak in
1:14:23 English. So that's not a problem. So find me a job which can pay me at least 100,000 dollars per year. 1 crore per year. So I will find a remote job. I can work out of my hometown. I have good internet. and I have a good computer. For one crore, I will go to another city. Yes, okay. I will go to another city but I prefer remote. I won't get a visa, Trump won't get a visa. So, because Trump won't get a visa, we need a remote job. Which can pay up to 100k. A minimum of 100k. So, you research. Look at which job is in which AI tools, AI workflows, AI first, who is resident in the chain. and set my profile to
1:15:10 which you don't need to be able to use software engineering. You can look up on Indeed based on my resume and give me 10 jobs that I can apply for basis of this on the internet. I have written full khichdi prompt. There you go full khichdi prompt came. Hindi, everything else. Sometimes it happens, you said so much and it doesn't catch it. Whisper? No, it happens, it saves the backup in the app. Transcripts are saved, I just have to click on regenerate. Very unlikely. But anyways. It has failed. It doesn't happen. There you go.
1:15:58 Now I said so much that he is answering in Hindi. I understand your resume and requirements. You are a senior generalist in AI deep hands on, he is not talking about you, same exact as you are talking about. But now this is what is crazy according to me, it used indeed in the back end and he is searching for different keywords, not one keyword, multiple keywords parallel search, which we will not be able to do at one point of time. It's doing all these searches in the background parallel based on what things can do well and look at this. Let me do more sorts specifically for AI generalist. Since that's the role Webber mentioned.
1:16:49 It's going through my resume and pulling information. So it will do all of this. And basically it will give me a proper list of jobs that I can apply for. Actually, we are doing something else. I've never tried this so I don't know if it will work or not work. Now let's do the same damn thing while it's happening on Comet. Because Comet's advantage is that I have my browser access to it so I can find it on LinkedIn, not in Indeed. So I gave the same resume, I'm not sure if it will work or not, we'll see. Okay. So same problem.
1:17:35 In a way. But indeed it is not. Hey, could you go through my resume? I'm looking for AI first roles. It could be remote, hoping for at least 100K in salary. I want you to take over my LinkedIn, look up for jobs which are remote, which fits my job description. So it fits my resume. So go
…
