# 40 Jobs That Survive AI (and 40 That Don’t)

## Метаданные

- **Канал:** Varun Mayya
- **YouTube:** https://www.youtube.com/watch?v=JurOkE1p7Ug
- **Дата:** 05.09.2025
- **Длительность:** 36:47
- **Просмотры:** 355,107

## Описание

Get Warp Pro at just $1 by using the code “VARUN” at checkout at https://go.warp.dev/varun

Microsoft recently published a study on which jobs AI is most likely to replace, and which ones are safe. By analysing 200,000 Copilot conversations, they ranked roles on an “AI applicability score,” exposing which careers might be going away soon. And that’s what we’re going to discuss in this video. Because the results of this study were as interesting as they can get. Translators, sales reps, journalists, data scientists, and even web developers fall into the vulnerable category. Meanwhile, phlebotomists, nurses, construction workers, and other hands-on roles remain safe… for now.

This video breaks down why desk jobs are under more pressure than blue-collar work, why AI flattens the middle, and why being “average” in an “automatable” role is no longer enough.

Microsoft Research Paper: https://arxiv.org/pdf/2507.07935v2 

00:00 – Introduction
00:48 – 40 Jobs at Risk from AI
23:24 – 40 Jobs AI Can’t Replace
32:51 – Final Thoughts

## Содержание

### [0:00](https://www.youtube.com/watch?v=JurOkE1p7Ug) Introduction

A new study has come out by Microsoft which talks about the 40 jobs that are most likely to be taken by AI in the near future and also 40 jobs that are relatively safe. Now you might ask me why don't if a big company has put out a paper like this why would you listen to them? They have so much incentive to lie. But the truth is actually in this case Microsoft has all the incentives not to put out a paper like this simply because of the amount of fear and panic it'll create. And if you actually go through the list of these jobs, you'll see that it's very risky for Microsoft as a company to put this out because a lot of jobs that they say are going to be replaced are jobs that in a way their tool suite enables. So I wanted to actually go through the 40 jobs that are likely to be replaced and also the 40 jobs that are safe. Without much further ado, let's go to the 40 jobs that are first going to be replaced.

### [0:48](https://www.youtube.com/watch?v=JurOkE1p7Ug&t=48s) 40 Jobs at Risk from AI

Before we begin, I want to tell you the methodology used. They took about 200,000 anonymous conversations from copilot of people asking copilot to help them with their work. And then they looked at the work that copilot did and tried to figure out did it actually do the work end to end or not. Then depending on the user satisfaction of that work done by copilot they decided whether AI was applicable there or not. And at the end they created this AI applicability score for each of these jobs which is how likely is it that AI can do it end to end in the next few months or years. Anyway that was the methodology. Let's get straight to the 40 jobs that AI can replace. Number one is interpreters and translators. Now, this makes obvious sense, right? The transformer as a technology was initially created to do language translation. So, it's really good at language translation already. So, it can easily convert one language to another. And when I was at Google IO in San Francisco this year, I got to try out these glasses that they have that also do instant language translation. And it is pretty good. They even did a live demo of this. So yeah, it makes sense that if you're a translator or an interpreter, um, you know, AI could possibly do a lot of the job for you. The next job on the hit list is historians, which are jobs that focus around research, analytics, synthesis, reporting around historical events. Because all these AIs have been trained on pretty much all the history books out there, they can immediately recall as well as give analysis of these historical events. So you don't need to have a person around you who's sitting and giving you an analysis of history when the AI can do it for you. Now even if a certain book or a piece of data isn't in the AI corpus, you can very easily dump the book into the AI and say find me information from this book. Next is actually passenger attendance, which is roles built around helping travelers, answering their questions, providing directions. You know, if you go to an airport, sometimes there'll be a person sitting there who's helping passengers go in different directions or telling the person you should go here or this is the flight timings or whatever. It is very easy to feed those flight timings or even the train timings into AI and just have it answer these questions for you. Eventually, it's just going to be a kiosk that you talk to and it'll give you the answers. Or you know what, it might not even be a kiosk as long as you have the AI glasses. And even Mark Zuckerberg came out recently and said that they're really going to focus on the glasses, right? Even Google is like I told you. So it is going to be obvious that it's directly going to connect to the flight APIs and tell you where to go. So you probably don't even need to talk to an attendant there. Next is sales reps, but a very different type of sales rep. This is a sales rep whose daily work is writing proposals, explaining services, and persuading clients and responding to objections or queries. This is not the sales rep that will meet some a client in person and then convince them of something. This is a person who's doing some of the back office work for sales. So uh of course a lot of this process is writing proposals and emails. A lot of cold outbound emails. Uh it seems that AI is pretty much on path to replacing. Remember that a physical sales rep who goes shakes a client's hand and tries to convince them of something is still seems to be a safe job. Next is writers and authors. So whether you're writing books, articles, scripts, essays, uh it feels like AI will sort of take this over. And I think this is a rising tide phenomenon which is if you're not a very good writer like a midwriter then sure AI will take your job but writing is one of those jobs with a very high ceiling at the very top the people who are writing it's not really writing it's the way they're thinking and their life experiences. So someone who can explain something to you in a very crazy life experience that they've had and then connect it to you know whatever they want to teach you or communicate to you. I think AI can't do that because it's not had those life experiences. But if you're somebody who runs a news blog or a news platform and all you're doing is rephrasing news from another website, then yeah, this is not a very good job for you to be in. Next is customer service representatives, which is a role focused on answering customer queries. You know, actually in the early days, maybe five or 6 years ago, there were these bots that you could create for your websites that will answer queries for you. Now, it can just be an AI that retrieves data from whatever's on your website and then answers those queries as long as you've written enough FAQs. So this seems like it's very obvious that resolving complaints, drafting a response, providing information is something that AI can do. But remember there's an accountability part of this, right? Which is when do you give a customer a refund? Especially if it's a large ticket product, right? Like if it's like a you know like a Zomato or a Swiggy where it's like 100 rupees that they're losing and it's part of it's baked into their costs, then fine an AI can decide when to give a refund. But if it's a big ticket product, let's say you bought, you know, camera and now a customer service representative has to decide whether the customer who's asking for a refund is a true customer who is who has a genuine reason or a fake reason, that's something that still requires human accountability. But if it's not those tasks, if it's just drafting responses and then or giving small uh refunds back, then AI can probably do it. Next is actually CNC tool programmers which is jobs that involve writing precise code for machines right so this is highly structured text so a lot of these tools actually require a little bit of programming which AI can do pretty much off the bat now I want to take a second to talk about today's sponsor warp which is an AD an agentic development environment and it's already ranked number one on terminal bench here's how it works so imagine your terminal but designed entirely around AI agents it can understand natural language pair share with you on coding, create pull requests, and even spin up multiple agents for entire projects, all running in parallel. For example, say you deploy to Kubernetes and your app keeps crashing in a crash loop back off. Normally, in your terminal, you'd run cubectal logs, get blasted with hundreds of lines of text slop, and waste hours just to figure out what's wrong. But in WP, the entire log is organized into a clean, collapsible block. An agent then reads everything, your command, the output, the session history, and tells you instantly what's wrong. your app is missing DB host in its environment config. You can then go ahead and patch it yourself or let the agent fix it for you. Or say you're someone who manages multiple projects manually. In you can just assign each project its own agent all running in parallel and monitor them from a single panel. Plus, you can even one-click share everything since every session is saved as a workflow inside drive so your whole team can collaborate on projects. There's a lot to talk about but we think you'll get to know how powerful it is only once you use it yourself. Now, it's absolutely free to try out for a limited time, but my friends at have given me an exclusive deal. So, you can get Pro for just $1 by using the code Wun at checkout at go. wop. dev/warun. Now, let's get back to our video. Next is telephone operators. Actually, there's a role whose job is to connect calls. It's like a switchboard operator, right? Somebody calls and then they're rerouting you somewhere else, giving directory information, answering standard questions. It's like routine scripted communication tasks, right? like you call customer support of a company and maybe they have 10 different places they could send you sometimes when you call you know customer support of a phone company uh they could be like they'll ask you to dial a different number depending on what you want all that can be pretty much automated and there are still some companies that use humans for that task even though that's been already automated you already dialing a number and it's actually hard to reach a customer supports representative on call these days usually the you know it's just you press those numbers and then there's an automated voice on the other end now it gets even more automated now it's almost like you're into a real human being and the person reroutes you where you need to be. Next is ticket agents and travel clerks. There is an entire role of a ticket concurge whose job is to help you book travel, tell you where you should stay, plan your trip for you, provide information, handle all your common requests. Uh, and all of this requires structured language, right? Which is where do you want to go? Like what kind of do you want a summer vibe or winter vibe? You know, is are you traveling with family? Here's an itenary for you. As we all know, AI can do the itinary piece pretty well. Next is broadcast announcers and radio DJs. So, a lot of audio communications. If you look at our avatars on Instagram, right, the entire audio is synthetic. We don't need a human being to come and, you know, speak out loud like I'm doing in this video, right? We don't need somebody to do that, especially if it's just radio announcements. So, that's certainly going to be automated. It's going to be a clone of somebody else's voice that's playing on radio all the time. Uh they will tell you what track is up next. They will tell you, you know, a little bit of today's tidbits and news. It's very obvious that job can be potentially automated with AI. Next is actually brokerage clerks. You might not be familiar with a brokerage clerk because most of you use platforms like Zeroda or Grow or any of the others to actually transact stock yourself. But in the past and even some rich people today, sometimes they call a person from one of these brokerages and then they make the transaction on their behalf. Or sometimes it's not even making a transaction. Sometimes it's just information and it's pretty obvious that AI can if you tell AI what transaction to execute it'll actually execute it for you. It's not very far away that from this and plus getting information around specific companies what should I do? AI can give you that feedback and advice. Next is farm and home management educators. This is a very simple role which is in the US there is this role where if you're running a farm or a home you are managing it. You're one of the managers responsible for it. Somebody's educating you on what you should do and shouldn't do. It's pretty obvious that this information can come from AI. Next is telemarketers. All the spam calls that we get from all these different platforms which anyway none of you pick up at this point but yeah that can be totally automated with AI. Next is concieres. It means somebody you can reach out to who can get stuff done for you. Like I used to use this concier service where I could text them and say hey I want uh you know a VOUP from the US delivered to India. Can you handle it for me? And then they'll go talk to a bunch of people or ship it and pay customs and all on my behalf. And then I just have to make the final transaction. So there are services like this in India and now I think it's getting democratized. There are a bunch of startups as well. I invested in one recently. It's pretty obvious that if you are a human being doing all of this, it is at some point going to be replaced by AI because it's a bunch of calls you're making. websites you're going to and making clicks which as we've seen chat GPD agent can do pretty well. Next is news analysts, reporters and journalists. If you are somebody that is manually going out there picking out news, adding commentary on it, doing a deep dive, then you'll be fine. But if you're just paraphrasing news from one place to another, it's pretty obvious that AI can do this. In fact, if you look at my short form channel, and AI does most of the hunting for us, right? So, I don't need as many humans there. We have 1/5 the number of humans that any other newsroom would have. And it makes sense because AI can do a lot of the finding work for you. We just have to add our insight or our viewpoint on some existing piece of news, right? That's where our creative modification lies. But there are lots of other companies that don't actually do that who just take something that happened from Twitter and then just you post it and say here's what other people said about it. They don't give their own opinion because they're, you know, it is kind of risky for some of these news companies to give their own opinion on news because they're not individual creators. In fact, AI already does it. It doesn't mean that the news platform is going to die. Don't get this wrong. It means that the job of a person who's taking news from one place to another is gone. The news companies will continue to do it using a bot. And we've seen a lot of this with even with Indian news. You know there are lots of bots running right now that will take interesting information and then just you know write an article on it immediately. Uh there's no human involved. Next is mathematicians which is crazy right because if you look at three four 5 years ago we said hey I can't do math and this year both deep mind as well as openai have won the IMO which is the international math olympiad. So mathematicians actually do two things right which is at the base level they're probably teaching or you know reoing a proof that somebody else has done and at the highest levels like if you look at a terren tower right they are formulating new proofs they're coming up with new ways of looking at the world and I feel like the second part is obviously not going to be easily automated but as AI keeps getting better it's very likely that at some point it starts generating the proofs by itself so mathematician is not a very safe job according to this study at least although I believe there are two kinds of mathematicians and the second kind are we still need a lot of them. Next is technical writers which is you build something and you want documentation around it. Today if you write a codebase then you probably need docs around that codebase for people to interface with it. Somebody using your API, somebody using your SDK. But uh I think in the future technical writers who do this are not needed because the AI itself can go through your entire codebase and then give you docs. It will tell the end user how to use your API or your SDK. It will tell the user what your idiosyncrasies are. It'll allow the user who's using your API or SDK to actually chat with the codebase and figure out, you know, interesting ways to use whatever it is you're selling. So technical writers are not really necessary. Next is proof readers and copy marketers. Somebody who's going through the document to see if there are spelling mistakes or grammatical mistakes. AI can do a fantastic job of this. Next is hosts and hostesses. Now, it's not very clear here whether it's an air hostess or somebody hosting a party, but I feel in both cases, right? Even with an air hostess, I really don't think AI can do the job. If you're taking a flight in India, you know, you have no idea, right? Somebody might be pissing on the plane. Like we we've seen crazy stuff. So, it's a lot of having to deal with crazy people uh all the time and angry people and frustrated people and flight delays, which I do not think an AI would do. Like if there was a bot sitting and telling them that the customer would take the bot and throw it out of the airplane, right? Like uh you need a human being who's empathetic and who can say sorry like a million times. So, so I don't think that's going to go in India at least. Next is editors. And this isn't video editors. This is the type of editor who would take a piece of writing and then make edits on top of it. Right? So in a lot of newspapers, you have the editor whose job is to have this team of journalists under them and then the journalist will write something and then the editor modifies it to make sure that it fits with the newspaper's narrative. So they think that editors are already using AI which is improving the grammatical issues in a piece of writing, improving the narrative writing. That is something that is susceptible to AI. Next is business teachers because I feel like in general with teaching AI can help you with a lot of specific like let's say you have a very specific problem today teachers can't do one-on ones right so and we made a video on this right because teachers just don't have the time to do enough one-on- ones in a day they have to teach a class to 100 people but with AI you can ask really stupid questions and you can keep asking those questions till the AI helps you and tells you hey this is exactly what you should do for your specific situation or problem next is public relations specialists anyone who's doing PR now I feel like there's two parts of being a PR person right one writing out communication for you on your Twitter handle or on your Instagram or social media or whatever it is. But the second is also knowing journalists because sometimes when you want that narrative out the PR specialist has to go to a journalist and say, "Hey, here's what I'm thinking. Here's what I'm doing. " And I feel like that part AI can't do because it won't know enough journalists. You need to have built relationships with those journalists over time. And that's why even on the journalism front, right? If you're a journalist who's going and doing deep dives, doing investigative journalism, talking to people in the real world and getting some information, AI just can't do that. Next is demonstrators and product promoters, which is somebody who's doing a demo of a product. Now, this depends on whether it's physical or digital. If it's digital, makes sense that AI can do it. If it's physical, then you know you need a real person there, right? Next is advertising sales agents. And I'm going to go slightly faster here because some of these you already know. The next is new accounts clerk. So, if you start a new account, how do you help you with that new account? like where do you go, what should you do? Next is statistical assistants. Then is counter and rental clerks. So if you're looking to rent something, somebody who can give you some information. Next is data scientists and there are 192,000 jobs here in the US, right? So they do believe that data science is something that AI is already doing pretty well. Next is personal financial advisers. I remember talking to an influencer 2 years ago who said he's going to build a personal finance app. And I said, uh, what's it going to do? Said, well, it's going to give advice. And I said, "Dude, AI will do a better job because it'll know your context very well. It'll know how much you need for retirement. to keep aside. It'll know everything. It'll be able to query what the current inflation rate is. It'll give you like a perfect answer. " And 2 years ago, he didn't believe me. But today, AI can practically do this, right? You can talk about your retirement plans with AI. And it will tell you exactly how to invest, how to save, what are the different options, uh how much to put in this, that, and it'll have context. And the cool part is it will not just have the memory of how you invested. It'll also continue to give you advice as the market changes. Next is archivists whose job is to you know see something happening in the world and then you know put that data together. Next is economics teachers which is pretty self-explanatory. The next one is web developers. Yeah, I've made videos for 2 and a half years about this. Right. I got abused in the first year and now people are using Lovable and Lovable's gone from 0 to $100 million in ARR in a year. So you can imagine how much demand there is for something like that to be automated. It's going to happen. There's too much money going to it. Imagine being a web developer and sometimes I see this in my comments and I'm just like you maybe you don't understand the relative scale of things, right? But here's a company like OpenAI going out and making a $500 billion cluster to take on the problem of writing code because that's very critical, right? For you to achieve the singularity for AI to improve by itself, it needs to be able to write code. So they are going to solve it. And then there's Meta who's dropping like hundred million dollar a billion dollar salaries for researchers to come solve this. Like why do you feel you as a web developer are going to be able to compete in the next few years. A lot of life is knowing what battles you can't take. So a very important way to think about this is and there's a famous song somewhere you know I can't recollect the name but the idea goes something along the lines of Lord please give me the strength to know what I can and cannot battle and for me to avoid the battles that I should not be battling or that I have no chance of winning right it's about guiding me along the path and a lot of people and maybe it's ego right like at this point you know from the comments it just feels like ego right which is hey I've studied this for so many years and I'm good at this how dare AI be better no it's not better. Like whenever somebody ignores obvious evidence in front of their eyes, it actually comes from a place of ego and insecurity. But I'll tell you a truth about life as an entrepreneur versus having a job, right? As an entrepreneur, you could do your best work. You could be really good at what you want, but the market simply might not support what it is you're building or selling. Right? That's the truth. Like you could be very good. You could have a great product. You could have great marketing, but you know, the market just might not want it. or like if you built a social media platform today, it's very likely no matter how much money you raise, no matter how good your product is, you will lose cuz they're already winners and they already have that critical mass. But wisdom and something I learned in my 30s and didn't have in my 20s is knowing what battles not to take. On Twitter, there's a post from me in 2012 about teaching people how to use Heroku, Postgrace SQL, Ruby on Rails. I was doing web dev at that time, right? And I've literally written a tutorial about I had to evolve beyond that in that era. So, it's something that's very special to me and it's something that, you know, I've spent a lot of my years doing. But it's okay if the market has moved on and it's if something's being automated, it's fine. You can always move one level higher. You can still run an agency or a company and have the bots do the work for you. Kunal Sha had this tweet recently, right, which is AI might force everyone to be an entrepreneur if it does all the jobs. And you might say, well, if the jobs are gone, how can somebody be an entrepreneur? It's because an entrepreneur's job is to manage resources in some ways, right? and to know what problem statements and what markets to enter and what markets are too competitive or there's no money in it. Right? And I think there will still be webdev companies and webdev agencies and the lovables of the world. I don't know if there going to be as many web developers. So either you move one level up and say I will manage these agents who are web devs or I'll manage humans you move a tangential path and say this is not for me. I still want a job but this is probably not the path for me because it's too competitive. I can't compete with a meta throwing you know so much money at this or an openi building such huge clusters to solve this problem. It's not a good place to be. Next is management analysts which are people who give managers advice or companies advice on how they should run the company. Yeah, AI can pretty much do this, right? If you give enough data about the company, AI can tell you what you should or should not do. Next is geographers. Then it's models and this is like you know fashion models, right? Which is obvious. We've already seen a lot of newspaper ads where it's like good-looking model but you know it's not real. Next is market research analysts. Deep research can already do market research very well. And this is by the way this has 846,000 people in the US. Next is public safety telecommunicators. Next is switchboard operators. Then there is library science teachers. Again teachers. If you notice the thing that ties all of these roles together is they're all communication and language or writing oriented. Even being a web developer in a way is language. You're still telling the browser how a page should be rendered. So as long as you're doing communication, especially communication to other humans or to computers, those jobs are most in threat. But the nuance here is if you run a company, let's say you run a very big data science company or web development TCS, you will still be probably fine because all you're going to be doing is switching out the human from the job role with an agent. That's what you're going to be doing. And that's why a lot of really smart people across the internet now, a lot of founders I know are starting to talk about entrepreneurship as a necessity, which is a lot like the early days where either you were a blacksmith or a farmer or something, but you were an entrepreneur. You produced your own output. So whether you use a tool or whether you used your hands, it didn't matter if you were responsible for growing oranges. As long as you produce the oranges with a tool, with a hand, it doesn't matter, right? You were responsible for giving the market oranges and people would buy oranges from you. So the data science company that's offering data science as a service to their customers will continue to exist, right? But they will probably look more like a product company than a services company, right? If you look at AOS, it's kind of the same thing with content, right? AI automates a very big chunk of content creation. So we said we have to play one level above and be the content company where we'll use a mixture of human beings and agents to do what we're doing. And therefore, our margins are a lot higher than traditional agency margins. In fact, our company looks a lot like a product company at this point because it's mostly tools and agents going off doing their thing, coming back with human supervision. And across the world, there are a bunch of companies doing this in different spaces. So TCS can actually lay off 2% of their entire workforce, which is 12,000 developers without actually losing productivity. So it is important for a lot of people to understand this because if you were running a small web development agency and you were known well for it even if you had three clients or four clients and they had that trust with you, you can now switch over to agents and I think you'll be fine because in entrepreneurship you kind of control your destiny. But if you're an employee now is probably the time where you should start absorbing a little more risk. Like there's a famous quote which is that the riskiest thing you can do is not take any risk and be completely dependent on other people for your sustenance and income. Right? That's not a good place to be in. And yeah, I think that's the nuance to understand here which is if you can control your destiny in some way or build good trust and relationships with your clients or even your employer, you might be around a lot longer than if you think about all of these purely transactionally because the agents will do the transactional piece much better. Now let's come to the 40 jobs not being replaced.

### [23:24](https://www.youtube.com/watch?v=JurOkE1p7Ug&t=1404s) 40 Jobs AI Can’t Replace

The 40 jobs that AI has the lowest applicability score. The first one is a phabottomist. A flabottomist is somebody that comes to your house, puts a needle into your arm, and draws blood. No way AI is going to do that. And no way you're going to trust AI to do that. It's kind of like somebody asked me what job are you sure that AI won't automate in the next one or two years. And my answer was a person that gives me a haircut cuz I wouldn't trust a bot to give me a haircut, right? You're folding a scissor very close to me. uh instinctively I just don't trust it. Next is nursing assistant. Somebody who can you know you're sick or you fractured a bone and some a nurse comes to you and gives you the medication and sits next to your bedside. Yeah, no way AI is going to do that in the short term. Next is hazardous materials removal workers. I think it's such a beautiful world we live in. A lot of people don't like this because they've studied digital roles. But uh I think it's a beautiful world we live in that somebody who goes and collects waste or removes hazardous material that none of us want to do is the kind of jobs that will be protected from AI. All the cushion cushy jobs sitting in front of a computer like mine are going to be automated while all the jobs that make society run properly continue to exist. I remember saying this on a podcast. Tomorrow if I had to be a petrol pump assistant, somebody sitting in a petrol pump putting in petrol or a garbage collector and that was the only way to survive, I'd do it. I'm very adaptable like that. I have very little attachment to my current status. I will do whatever is necessary to survive. And when you've been an entrepreneur for 12 years of your life, you don't have a choice. It's almost like the market will make you do something to survive and you got to do it to survive. So I have less addiction to specific titles or job roles. I care more about uh how can I continue to exist and if I have to be a hazardous materials removal worker tomorrow so be it. Next is helpers which is painters, plasterers, you know people who come to your house and paint. Next is imbalmer right somebody when you die they embalm you in the US. Next is plant and systems operators. Somebody who's in a factory plant whose operating systems of course makes sense that AI can't replace their job in the short term. Next is oral and maxeloacial surgeons like dentists. uh very likely that AI can't take that job. I wouldn't let a bot next to my teeth right now. Uh next is automotive glass installers and repairers. Next is ship engineers. Then there's tire repairers and changers. Then there's prostodontists. Then there's helpers, which is production workers. Uh then there's highway maintenance workers. Somebody fixing a highway or a road. Uh the people we all complain about even though we don't know how hard their job is, they're going to be around. Uh next is medical equipment repairers. Uh then there's packaging and filling machine operations. Then there's machine feeders. Then there's dishwashers. The maid that comes to your house, washes dishes. It's going to be very ironic that they end up having a job and many of us might not. Uh next is cement masons and concrete finishers. Next is supervisor of firefighters. Next is industrial truck tractor operators. Then there's optalmic, which is eye medical technicians. Then there's massage therapists. Somebody who's giving you a massage. By the way, a very interesting stack. Most of the top rated massage people on an urban clap, you know, rarely I will get a massage and then I will ask the person how much money. It's kind of like I ask them how well they're doing, right? How much money they make. Like and I asked this to everybody. Like in New York, I asked a taxi driver this and he said he was making 1. 5 crores a year, which is which actually turned out to be true because I asked multiple taxi drivers in New York how much they were making. They all make way more than you know uh than you expect them to make. But the massage therapist and this is what surprised me in Bangalore is making roughly one and a half two lakhs a month which is not a bad salary but again it is an entrepreneurial job cuz there are some months where you make two lakhs and zero right and this is the thing right the more discovery you have the higher up you are on urban clap the more likely you get visibility orders so entrepreneurship has a very strong relationship with distribution the more eyeballs you have the more likely you will get more orders and more deals right in our business it's deals in the massage business. It's getting clients and I wouldn't be surprised if the top rated massage therapist makes a lot of money, right? The person who's most visible and probably is booked out all the time. So, yeah, massage therapists are going to be around. Next is surgical assistants. Somebody who's doing surgery, but you're standing next to the surgeon. Then there's tire builders. Then there's roofers. Then there's gas compressor, gas pumping station operator, which one day I might be. Then there is people in the oil and gas space. Maids and housekeeping cleaners. Makes sense. Like I'm really hoping they stick around while you know a lot of the uncles and aunties who purposely keep them below a certain you know range I've seen that a lot and that's very classist behavior right which is oh a maid shouldn't make this much money or this or that like you know what we are all subject to supply demand the reason software engineers made so much money and the reason you know content creators made some money for a while it's all supply demand right it's all the world needed more software engineers the world had space on YouTube for people to come out and do content fine but supply of good talent in many of spaces that have given people that you know that class that oh I am better than other people that goes away right and it goes back to pure economics right so if the demand for maids and housekeepers continue to go up while the demand for other roles go down especially digital roles go down I think it's a good thing and I feel it will normalize this dignity of labor that India has never had we need dignity of labor if you go to the US there's dignity of labor a petrol pump person is seen in the same light as a restaurant owner Right. As a software engineer, I think it's important for us to have that dignity of labor in India and it truly makes everyone respect each other a lot more, right? Instead of this weird new class system that we ended up creating. Then there is uh paving surfacing people. Then there's logging equipment. Then there's motorboat operators. Then there's floor sanders and finishers, people who do your floor. Then there's pile driver operators, rail track laying, foundry mold and core makers, water treatment plant and system operators, a bridge and lock tenders, and dredge operators. You know what this signals to me? This signals to me we are ending, we're entering the era of the end of knowledge work. Cuz all the jobs that look like they're going to be automated are people who have studied something, who have knowledge in their brains, and who are communicating that knowledge and connecting things on a computer. That era is probably coming to an end. And I said this on a podcast with Vant Trusty like 5 years ago right or four years ago where I said that it's easier for people to automate YouTube thumbnails and this was before a time you know we had all this AI technology out there right because I had played with Deli 2 at that time I had an early preview we have so much more data on thumbnails that if you want to create a model on creating thumbnails we can easily do it but the minute in the real world you want to have somebody who is holding a camera and walking around taking specific shots or you have somebody who's laying rail tracks we simply don't have enough data on that movement on how they lay it, how they're thinking about it. But for a long time, I know why digital jobs have looked cooler to everybody than physical jobs. Like I'm pretty sure of why this happens, okay? It's because the people with digital jobs use social media far more than the people with physical jobs. And if you're doing a job, you're going to make it sound cool, right? If you're, you know, somebody who has access to media, who can create their own media and communicate well, then you can tell stories better. And the world is driven entirely by stories. So the better you are a communicator, the better stories you can tell, the better you're going to make yourself and your job look right that is the entire thesis of this. And there are lots of people who are like hey on Twitter people are thinking like this or on Instagram Reddit people are thinking like this. I'm like it's never representative of the entire population cuz the population that's terminally online is a minority. You know why software developers have one of the best narratives and had for almost two decades is because when the internet first came about the first group of people who are on the internet everywhere from hacker news to even Reddit to old MSN to old AOL when I was a kid when I was 7 8 years old and I first got access to a computer it's mostly people in software or it's people who are very native to computers gamers. So it was that audience and of course they're going to say good things about their jobs. Of course they're going to go out and say our job is superior. What we're doing is superior. It's a status thing, right? And that happened for so long that we kind of forgot that all jobs are dictated by supply demand. Can you do something that is fundamentally rare, then you'll make money today? If you're a top AI researcher, you make $100 million a year because it's a rare skill. It doesn't have to do with the fact that you are an AI developer or you are in this field or you're doing this or you're doing that. It is how rare is the skill that you have and is it easily available in the market. That's it. We all are subject to the laws of economics and I have said this for three years people you know it's very hard sometimes to communicate with the audience right because for three years I've said we have an over supply of software engineers we're producing so much of them that we're breaking the supply demand curve and that's why TCS is laying off and so many software developers in India right now actually not just in India but all over the world are having a hard time finding jobs because we made it really cool and then everybody went out and did it and remember I studied computer science and engineering in college right so everyone went out and did it and then when you have a huge over supply you simply don't have enough roles for those people. Now if everyone hates the idea of software engineering and nobody does it in 5 years even with AI the role will be empty and then suddenly the demand will go up again. we are in this economic pendulum right like supply demand and I feel that we simply don't have at this even with the models right we simply don't have enough of a supply of data around how brick layers do their job or machine operators do their job and we don't have good models that can simulate what they do but the day robotics gets good and those models get good and they are able to compete with a human being then everything in this document changes and one last thing I

### [32:51](https://www.youtube.com/watch?v=JurOkE1p7Ug&t=1971s) Final Thoughts

want to tell you right even with this document if you are top 0. 1% % of your at your job or top 1% into your job. Most people in the top 1% of their job do synthesis. They don't just do one thing. Then if if you work in a news organization and you are top 1% of your job, you're not just writing news. You're communicating with stakeholders. You are talking to your investors, you are making a phone call to a person, a source that you know who and you're bringing that source, integrating that source in whatever you do. It's a synthesis job. You're doing multiple things. You're not just like a monkey sitting down and writing text. Same with software engineering, right? If you're a top 1% software engineer, you know, a lot of it is figuring out what needs to be built in the first place. What's the approach to build this? What are the constraints? How much time do we have? What kind of team can I put together? Can I do I know a team, right? Do I know how to use these bots? How much of it should be real human beings? What database do I use? All of these are synthesis roles. They are not just doing one thing. There's a reason that the top 1% of software engineers don't actually maintain their GitHubs because they have so much work on their plate that they would rather hire somebody to write code and they would sit on top of them and manage them. In fact, Carmarmac said this, right, which is he said that the highest leverage activity I can do is actually hire people and manage them, not write the code myself, right? So, this is something that if that you need to really take into heart if you want to be in a job, you need to be top 1%, top 5%. And it's not that hard We're entering an era where the top 1% doesn't mean you're the best at what you do. Your skills are the best at one specific narrow domain. It is actually that you can do all of these synthesis tasks. These five 6 10 things that is required of an executive. And in companies we have this role right CXO, CEO, COO, CPO, whatever it is. The minute you have that role or a VP of something, your job is synthesis. The job is to put all these things together and handle crisis and keep the business afloat no matter what. You are tied to the outcome of the business. whether the business does well or does badly. And people have forgotten this. We've had the people in HSR now who have spent 10 years in unprofitable startups. They've never worked in a startup that has made a profit. Yet every year their salary keeps going up. Nobody has stepped up and asked why. And AOS has been profitable for 3 years now. We've been profitable since almost the first week we started doing this, we've been profitable. And we're 400 employees now. It's a very different constraint. So my belief is the more you care about the outcome of the company. And Steve Jobs has a good quote on this which is he said between the janitor between the cleaner and the executive the person running the company excuses go away. A janitor can give an excuse. A cleaner can say bro I couldn't clean the room because the door was locked. But somebody who's a VP level of a company somebody who's responsible for the outcome of the company can't say that. That person has to pick up the phone find out who has the keys. If the keys are not there, ask call somebody from outside, get the door broken down and create new keys and then get the room cleaned. Whether they do it themselves or hire somebody else to do it or use a bot doesn't matter. But the job is to clean the room. And we are entering a world where outcomes matter a lot more than inputs. Everyone on YouTube especially people who are learning today are very input driven which is should I study this? Should I do DSA? Should I be good at these 10 questions? Should I study this? Should I be this? Should I do this? Should I learn the skill? That era is over. Today we're in the era of outcomes. What can you create? Like whenever somebody puts out a really good video, I'm like, "Oh, these this person had to do these five different things to put out a full video. " Whenever a person makes an app and they actually put out an output of their app, I'm like, "Okay, this person knows how to do this. " That's why on Twitter these days, when somebody goes out and talks about inputs, like, "Hey, I learned this cool technique or, you know, I'm studying this cool thing or, oh, this person doesn't know this technique. " Uh, everyone looks down on them today cuz everyone has realized the world is now about outcomes. And this entire paper is about doing specific jobs in terms of inputs because the AI will do the inputs. It'll keep skill maxing. And there's no nothing you can do to compete with the next GPD6. you know, everything that all the money that's being poured, you and I can't compete on skill, but we can compete on outcomes. If there's one thing you take away from this video, it is that even if you're in one of those high-risk jobs, the minute you're focused on outcomes, you'll do just fine. That's it for me. Hope you learned something. Bye.

---
*Источник: https://ekstraktznaniy.ru/video/11940*