Limitless Podcast - Will AI Take Your Job? Andrej Karpathy's Report Shows Who Will Survive (and who won't)

Episode Date: March 16, 2026

Drawing from Andrej Karpathy's report, we discuss the effects of artificial intelligence on the job market. We uncover which jobs are most at risk of automation, such as financial analysts, while hands-on roles remain safe.With 150 million jobs potentially impacted, we highlight the evolving nature of work and emerging opportunities.------🌌 LIMITLESS HQ ⬇️NEWSLETTER:    https://limitlessft.substack.com/FOLLOW ON X:   https://x.com/LimitlessFTSPOTIFY:             https://open.spotify.com/show/5oV29YUL8AzzwXkxEXlRMQAPPLE:                 https://podcasts.apple.com/us/podcast/limitless-podcast/id1813210890RSS FEED:           https://limitlessft.substack.com/------POLYMARKET | #1 PREDICTION MARKET 🔮https://bankless.cc/polymarket-podcast------TIMESTAMPS0:00 AI and Job Replacement1:08 Andrej's Report6:31 Blue-Collar Jobs vs. AI8:32 The Future of Work11:23 Economic Indicators and AI14:49 The Role of AI in the Future16:37 Jobs and Human Demand20:19 Using AI21:33 Closing Thoughts------RESOURCESUS Job Market Visualizer:https://karpathy.ai/jobs/Josh: https://x.com/JoshKaleEjaaz: https://x.com/cryptopunk7213------Not financial or tax advice. See our investment disclosures here:https://www.bankless.com/disclosures⁠

The number one question we keep getting asked is, will AI replace my job? Well, now we have an answer. The godfather of AI, Andre Carpathie, released a report which tells us which jobs are at most risk of being replaced by AI. And it's not quite what you might think. Jobs like financial analysts, computer programming, and transcriptioning are at most risk.

But what shocks me the most are the ones that are the safest. Plumbing, bartending, being a mechanic, turns out AI won't replace manual labor because it just doesn't have any hands. It still requires human bodies. But on the other hand, if your job involves a computer, you might be kind of cooked at this point. But it's not all doom and gloom. The report measures AI exposure, which doesn't necessarily imply replacement, but rather the evolution of job roles. What your job looks like today is very much going to look very different

a few years from now. Yeah, so a brief timeline on how all this stuff unfolded. On Saturday morning, Andre Carpathy, this is the guy who coined vibe coding, he co-founded OpenAI, he led AI at Tesla. I mean, this, this is the guy. He spent a few hours vibe coding this project that scores every job in America on a zero to one scale based on, like you said, EJAS, how exposed it is to AI. Now, this is not a direct correlation to how likely it is to be replaced, although there is some indication of that. Basically, he publishes this project and it goes absolutely nuclear. Elon's replying, Fortune picks it up, it's in all of the major publications, all within hours, and then he deleted the repo. It was gone off of the web, completely and entirely. Thankfully, I was playing around

with it that night. I forked the repo and I republished it on GitHub, and that's what we're looking at here. And this is the project. This is what shows all these jobs that are going to be replaced. If you see your job here, that's okay. We're going to walk through a bunch of categories about what is most affected, what is least affected. And I guess kind of the impact that AI is going to have. So, Ejaz, there's some high numbers here on this left tab here that maybe we can start by walking through. Yeah. So what are you looking at on the left is the total number of jobs that are affected. And it's quite a big number. It's almost 150 million. Now, if you scroll down, you'll see the number of or the amount of wages that are exposed to potentially being automated by AI,

$3.7 trillion. That is like one of the larger numbers that we've seen when it comes to displacement. But if I shift your focus to the right here on the main graphic, you notice that there's a lot of red and a lot of green. It's pretty self-explanatory. If you see your job and it's red, it's at most risk. It's at the highest exposure. If it's green, it's probably the most safest. So let's start with the worst medical transcriptionists. AI exposure 10 out of 10. 10 out of 10. Josh, we're looking at a median pay of almost $40,000 and there are 44,000 jobs that do this in the

U.S. How much is that worth? That is about $1.65 billion of value that has 10 out of 10 exposure. Now, there aren't many 10 out of tens. In fact, I think this might be the only one that is truly 10 out of 10. I'm not going to lie, this one might be cooked. And the way this is categorized is kind of in these like higher level brackets. So there's the 10 out of 10, which is the medical transcriptionist.

If you are a transcriptionist, I would work very hard to try to use AI and maintain your position as best as possible. But then beneath that, I think one of the categories that surprised a lot of people is a very popular subcategory of jobs, which is like the software developers, the designers, the people who are working on computers every single day. A lot of these jobs are at an eight or a nine out of 10. They're very high relative to a lot of others. And there's a lot of them. If you look at the software engineering section, we've got an average median pay of 100. $130,000 a year, but they're almost two million of these jobs. And this is taken from like, I think, 2024 data, so it's probably a lot more today. But it's not just technical roles. It's also

graphic designers, data scientists. They're all exposed. Wait, did I see anything about podcasting in there? Oh, my God. I saw video production. Oh, that seems scary. No, you know what, Josh, I actually haven't seen this at all. So I think we're actually safe. And I'm going to choose to ignore it and not look for it. But market research analysts are exposed. Customer service Reps, 2.8 million jobs, nine out of 10, general office clerks. So what I'm seeing here, I'm noticing a trend in the jobs that are affected. It largely involves computer-based tasks, but ones that are very repetitious and most likely to be automated, kind of like low-hanging fruit, like back office work, like moving files around basic analyst stuff. The other trend I'm noticing is these sound a lot like

entry-level jobs, Josh. Like when I'm reading the descriptions of software engineers here, it's mainly talking about like tasks that junior level engineers might do or junior level graphic designers might do, which kind of mirrors the trends that we're seeing on the job market right now where entry level jobs in the US at least are completely evaporating. And it's also important to note that the AI exposure does not directly correlate to the job loss. It's just the involvement of AI in the actual job. So we're seeing at companies like Anthropic, Open AI, the companies that are building this software, they're hiring faster than than ever. And they're limited purely by the constraint of quality people to hire versus the actual

need and the want to hire. There are some places in which this could harm people. So like if you're a software developer right now, if you're a designer right now, AI is such a great tool for leverage. You can do so much more. You can get so much more output if you use these tools for leverage. In the case that you don't, or in the case that you can't, in the case of the medical transcriptionist in which it literally is just like dictating notes into a page, it's going to be tough. So this isn't direct. a job replacement board, but it just shows how much an impact AI will actually have on your job. Now, there's a lot on this board that is green. And that means that not only is AI not coming to replace the job, but there's not much

AI involvement at all in the job. And I think this is a bigger conversation around these lagging indicators in the physical world where AI is wreaking havoc on software. But when it comes to just your dude down the street who's coming to fix your sink or your toilet or paint the walls, like, there is no competition for that at all, right? Yeah, it's completely inverted where typically technology advances will kind of kill blue-collar jobs or completely evolve that landscape. We're seeing the opposite here. So if you're a janitor or a building cleaner, one out of ten, you're not exposed, you're safe.

If you're a grounds maintenance worker, again, safe, if you are working in any kind of labor or construction, home health perspectives, things that require your hands or a human body or human intuition in that sense, you're pretty much safe for now. And the main reason for this is probably obvious. AI or AI models can only really talk and do computer-based tasks. They can spit out characters and words and numbers and analysis, but they can't action that in the physical world yet. Robotics is still comparatively very young at this point, a young industry, and it's probably going to advance, but it's going to take a little longer than AI LLM's will right now. And it's a really stark contrast. Like when I think about technology replacing people's jobs, I always think

that it might start from the bottom up, but I think this is a clear example of it happening from the top down. Yeah, and it's funny because when you look at the people most exposed, they're generally the highest earners. The people least exposed are the lowest earners. When you look at the average, the median pay for all these jobs, it's normally the ones with the highest median pay that are most susceptible. And I think it's because, I mean, generally speaking, software has been such a lever in terms of how much productive output you can generate relative to the physical world in which you are one person. And that lever has created a lot of value. But now that value is able to start to become either subsidized or replaced by these AI tools so long as you're not

using them for the leverage yourself. So it's this interesting chart. I've really enjoyed looking through it. Like if you are, what do we have, child care worker safe, general maintenance are safe, software developers not, cashiers, maybe not. There's this middle section that I find interesting where It's like the heavy tractor drivers, kind of towards the top. And there's the retail sales workers and the cashiers. So if you're in retail, you're probably okay for now. Is there anyone who's truly safe EJAS who has zero exposure to AI? I can't find it right now because it's probably minuscule.

But if you are a truck driver, you are at a zero out of 10 AI exposure. So you're the safest job in America and in the world right now. Congrats. And I hope you're listening to Limousis on your truck ride this morning. Yeah, exactly. But anyway, to kind of like tie a loop around this, The directional trend is pretty clear. AI is going to automate a large amount of jobs.

It's not just going to be in one industry or one vertical. It's going to be across the entire board. And actually, Elon had a very poignant tweet that he put out here, which he said, all jobs will be optional. There will be universal high income. And this was in response to Carpathy releasing this report. And the point that he's making here is no job is safe. But equally, the optimistic take is in a world where no job is safe and humans don't really have a

purpose when it comes to whether it's physical or software-based labor, they will still get paid, not just a basic income, but a very high amount of money, which I'm still trying to wrap my head around and understand, but it gives us a more optimistic end to this story. Yeah, and some caveats too. This doesn't cover the entire U.S. job economy. I mean, the way it works is the data source he scraped 342 occupations from the Bureau of Labor Statistics. Yep. And this covers about 143 million total U.S. jobs. And then he fed that into chat GPT, scored each one. So this is a very generalized approach that isn't quite accurate, but it's enough to make you think about it, right? It's enough to kind of give you a generalized sense of where things are going, who's most affected. Overall, the weighted

average of the AI impact was 4.9 out of 10. So almost everyone is feeling the effect of this, to some extent. Jobs paying over $100,000 a year, like we mentioned, 6.7, which is much higher than jobs paying under $35,000 a year, which is $3.4. and then 42% of all jobs scored seven or higher, representing 60 million workers and $3.7 trillion in wages. So the impact is real. I think when you talk to people who are employing AI, they're not quite seeing the productive output just yet

where it's replacing jobs. But the impact of this was clearly noticed because surely after publishing this, Andre deleted it. He took down the entire thing. He deleted it because of this exact response that it conjured up. everyone started freaking out and thinking, oh my God, my job has a 50% risk of getting replaced by AI. I don't know how much time I have left. I should start panicking.

And the reason why I deleted it was he said, listen, this was a two-hour vibe-coded project that I did over the weekend. I didn't even announce it online. You guys just found it and you've massively extrapolated it into your own story and caused fear. And his point is, this is based off of raw data. But again, we don't quite know how any of this is going to pan out. and I'm likely going to be very wrong about it. The point he made was, although these jobs are exposed to AI, it's not necessarily going to take away people's jobs.

It's going to evolve those roles into different kinds of jobs. Like, for example, if we take software engineering, everyone's freaking out about AI replacing coders. The truth is, the coders will just become managers of AI agents that do the code for them. Have we ever had a job like that before? No. So it's the same as like the Industrial Revolution, where we have this like weird evolutionary period,

we don't quite know what's happening, but we know that hopefully we're going to be okay by the end of it. Yeah, and he followed up that original post with the exposure was scored by an LLM based on how digital the job is. This has no bearing on what actually happens to these occupations, which has to do with demand elasticity and a lot more. This is noteworthy. And I think, again, the backlash comes from, there are strong headwinds that point the opposite direction, right? Like, just recently we had a jobs report that came out in February where the U.S. economy lost 92,000 jobs in February. On unemployment rate, it ticked back up to 4.4%. Healthcare, which was an industry that was really holding the jobs economy together, it shed 28,000 jobs in February, and December was then revised

down to lose 17,000. So unemployment rate is much higher. I mean, when you look at this chart, if you scroll down just a little bit more, it shows there's some serious losses happening. It's been happening in a couple years. Is this a lagging indicator of the overhiring that happened in COVID? is this an effect of AI? It's probably a mixture of sorts. We can go to our friends over at Polymarket to kind of seek truth and understand this a little bit better. And we'll take a look at this Polymarket that shows how many jobs will be added in March. Now, you'll notice the number was much higher recently before crashing down of 50 to 100K. We were at what? It was 44% and now it's dropped down to significantly less under 30%. So it looks like the Polymarket is kind of

guiding towards a worse jobs report than we think, how high will U.S. employment go in 26? There is a 58% chance now that it will be up to over 5%, which is noteworthy because we're currently at 4.4. So this is a trend that Polymercic sees happening. And then I guess we have to ask the question, will there be a U.S. recession clause by this by the end of 2026? Thankfully, the number is staying strong at 31%. And trending downwards, it looks like, over the course of the time that this market's been in place. So it seems like maybe we're losing some jobs, but it's not going to cause a recession. And it's just, it's a good time to learn how to use these AI tools, leverage them for your own well-being, for your own productive gain. And thank you

very much for Plymarker for sponsoring this part of the episode. I think it's important to mention that Carpity's report doesn't live in a silo. There are a bunch of other reports and studies which actually confirm his data. Anthropic themselves released a study a few weeks ago, which looked at the labor market impacts of AI. And they had this pretty crazy chart on the left here, which basically shows the ability for AI to automate particular jobs or skills. So like business and finance, computer and math, you'll see the blue section shows that AI can basically do a lot of that job right now. And then the red basically shows how much AI is actually penetrated that particular job or market in the real world. So the point being made here is AI today can actually

do a lot of the work that a lot of human employees and workers do, it just hasn't diffused yet, which is a lot of scarier of a proposition than I initially thought. I thought, okay, well, AI isn't just smart enough yet, and so it's going to take some time. But the truth is, it's already here. We just haven't dispersed it into everyone's hands just yet. And it's more of an adoption issue going forwards. And this is the take. I think a lot of people might not realize when they want to ask AI to slow down. They want to decelerate. And the reality is that these models are so powerful now and they're so capable that it's no longer matter of increased intelligence. It's more a matter of diffusion. I find that most of the impact

that AI will have on the world around us, we have the models that are sufficient in order to enact that change. It's like, what do we need more intelligent models for? Solving new physics, solving new math, solving new science for solving novel breakthroughs. But the things that we do every day, it's really a dispersion problem. It's a matter of getting the AI into these systems and automating them because it exists today, it is here. And I think one of the, I mean, if you don't believe us, we can defer to Elon's post, which I think was one of the more noteworthy comments on this whole Andre's story because he replied to it saying, all jobs will be optional. There will be universal high income. And I think he really believes this. I've listened to Elon in a series of conversations

in which he's kind of laid out the path for the future and what that looks like. And it seems like it's going to be a very difficult train to stop, particularly once we get humanoid robots. So one of the problems with this publication that a lot of people mentioned on X is that it doesn't factor into account what the world looks like. Once a year breaks out of its box, once there are physical robots kind of moving around, once there are robots in factories, robots going down the street. Recently, last weekend, Travis Kalanick released his new company. The founder of Uber, he created the follow up called Adams. And Adams is entirely based around moving these physical atoms and automating the movement of these atoms. So I think as we start to move forward

into the Elon-based world, into the Travis-based world, where the AI breaks out of the box, oh yeah, here we have it here. This is the vision from Travis Kalanick's new company named Adams. And the idea is that AI is breaking out of a box. It is becoming physically manifested through these robots. These robots will become so efficient that the cost of doing a productive service or task will basically collapse to the net goods, the cost of the net goods, and the cost of the energy expenditure and that labor cost will be removed from the equation. And a lot of these people are betting their entire companies on this equation being true. The craziest part just going to mention this Anthropic report, Josh, is this came from a frontier AI lab. So they're building the tool

that's going to replace or automate a bunch of this stuff. So it's a stark display of honesty and transparency, which I'm quite impressed by. But the other point I want to make, and I want to end on a positive note is I think a lot of this AI stuff is going to result in more demand for human jobs, which might be a very hot take at this point. But I think what's the law, Josh, where if we have more of the thing, it actually leads to a higher demand? That is Jevons Paradox, my friend. Thank you, Jevons Paradox. So Jevons Paradox basically states that if you have more supply of something, let's say more compute out there, you might argue that, okay, well, it's going to cheapen the use of AI. The result is often the opposite, which is more people can do things with this cheap

AI, so they start creating more jobs and more use cases and consumption and production absolutely go exponential. And that's what I think is going to happen in the job market. I think initially we're going to come across a bit of turmoil because people don't really know what that next level of their job looks like. But once things settle, once the dust settles and once people realize, oh, we're living in this new world where AI agents can do things for me where my personal agent can handle all my payments and shopping and all that kind of stuff for me, it frees you up to do much larger, bigger things. Everyone can vibe code their own app or product or launch their own company and run it completely by themselves. And that's a new world order, which I don't think

people have wrapped their heads around. I certainly haven't. I don't really know what that looks like. But I do think it's going to result in more jobs and maybe even higher pay per person. I think people will become smarter as a result of AI and become more productive. And it's, it's okay that we don't know what it looks like. And I wouldn't expect anyone to know what it looks like. We're going from a very certain thing to a very uncertain thing. We know what the jobs market looks like. We had a very clear projection of what was going to happen. And now that has changed. Does it mean that it will result in less jobs, perhaps in the intermediary, in certain subcategories? But net net, at the end of this, I mean, we've just, we've continued to

evolve. We've gone through the agricultural revolution, the industrial revolution, the industrial evolution, each one of those times, jobs disappeared and then reappeared in so much more abundance. And every time there is more opportunities, they're more fun opportunities. I mean, we solve a series of pretty mediocre problems. How many people really grow up wanting to be a medical transcriber? Is that your dream? Perhaps not. Perhaps it just enables you to do something that is more valuable, more exciting, more interesting. And I think we're going to continue to see that across the stack. So it's scary because it's uncertain. But I think the one certain thing is that we have continued to evolve to create more ways to spend our time that yield in a better result of living, a better

quality of life for everyone. And I guess that's the optimistic hope we'll take at the end is things will be uncertain. Things will be weird. It's moving very quickly. It's changing quickly, but hopefully everyone wins. And at the worst case scenario, Elon's giving universal high income for everyone, right? So yeah, I know. I'll take care of it. Yeah. So to summarize, if you are a college guard right now and you're freaking out about what job you're going to have, the most highest thing you can do, even if you are currently employed, is to start using these AI tools. Josh and I try and use it day to day in everything that we do, whether it's research for the podcast, or whether it's trying to help teach ourselves something that we don't understand. We use it for everything.

And the point is, the earlier you can start using these tools, the more acquainted you're going to become with what this new world order is going to look like. And the better proposed you are to evolving into whatever that job eventually becomes, the jobs that you're probably going to a take in the future probably haven't even existed yet or haven't even been created yet. So there is a cautiously optimistic approach to all of this. And I truly think that if you use the tools, you will better set yourself up for success. Very well said. And if you're interested to see just how affected your industry is, we'll include a link to this project in the description. So you can go and check it out for yourself. I would love to know what profession you are in.

How cooked are you? How affected are you? If a one of these? Yeah. Are you in the box? Are you one of these people? Let us know. I'd love to know what everyone's up to. What is their exposure? If they're worried or not, I think there is a lot of worry. But some people also are kind of pragmatic about it.

They're like, hey, I'm actually using AI and it's making my job better. It's making me better at my job. So actually, that's a good point. Like, are you using it in your job right now? If you see your job here and it's red, like, are you, do you see the opposite? Like, are you getting more productive? I'm curious. I would love to know.

But that wraps up our Andre Carpathy Jobs Report episode. Thank you, as always for watching. If you enjoyed this episode, share it with a friend to either scare them or enlighten them or excite them. Whatever emotion you wish to elicit, I hope it finds its way into their inbox. If you enjoyed, please don't forget to rating it on your favorite podcast player, commenting, subscribing on YouTube, all of the good things. It's really been helpful. We've been doing great. The numbers are up into the right.

We had our best week ever last week following a best week ever the week before. And it's all thanks to you and all of the support. So thank you so much, as always, for watching. Just one last thing, there's 3,500 newbies here, right? Is that right? Did I get that figure right? We're getting a lot of new subscribers. Thousands.

Thousands. Well, welcome. And if you are new to limitless, please subscribe. We post four times a week covering the latest and greatest things in AI and Frontier Tech. And published great newsletter twice a week, which is also linked in the description. You can find links to everything and anything down below. Thank you so much for watching as always. And we will see you guys in the next one.