Matt Shumer's viral AI warning draws 73 million views and fierce debate over job automation claims

The AI industry has a big Chicken Little problem

Please, sir, may I have some more AI? Credit: Getty Images / Bettmann collection Entrepreneur Matt Shumer's essay, "Something Big Is Happening," is going mega-viral on X, where it's been viewed 42 million times and counting. The piece warns that rapid advancements in the AI industry over the past few weeks threaten to change the world as we know it. Shumer specifically likens the present moment to the weeks and months preceding the COVID-19 pandemic, and says most people won't hear the warning "until it's too late." We've heard warnings like this before from AI doomers, but Shumer wants us to believe that this time the ground really is shifting beneath our feet. "But it's time now," he writes. "Not in an 'eventually we should talk about this' way. In a 'this is happening right now and I need you to understand it' way." This Tweet is currently unavailable. It might be loading or has been removed. Unfortunately for Shumer, we've heard warnings like this before. We've heard it over, and over, and over, and over, and over, and over, and over. In the long run, some of these predictions will surely come true -- a lot of people who are a lot smarter than me certainly believe they will -- but I'm not changing my weekend plans to build a bunker. The AI industry now has a massive Chicken Little problem, which is making it hard to take dire warnings like this too seriously. Because, as I've written before, when an AI entrepreneur tells you that AI is a world-changing technology on the order of COVID-19 or the agricultural revolution, you have to take this message for what it really is -- a sales pitch. Don't make me tap my sign. Shumer's essay claims that the latest generative AI models from OpenAI and Anthropic are already capable of doing much of his job. "Here's the thing nobody outside of tech quite understands yet: the reason so many people in the industry are sounding the alarm right now is because this already happened to us. We're not making predictions. We're telling you what already occurred in our own jobs, and warning you that you're next." The post clearly struck a nerve on X. Across the political spectrum, high-profile accounts with millions of followers are sharing the post as an urgent warning. This Tweet is currently unavailable. It might be loading or has been removed. This Tweet is currently unavailable. It might be loading or has been removed. This Tweet is currently unavailable. It might be loading or has been removed. This Tweet is currently unavailable. It might be loading or has been removed. This Tweet is currently unavailable. It might be loading or has been removed. To understand Shumer's post, you need to understand big concepts like AGI and the Singularity. AGI, or artificial general intelligence, is a hypothetical AI program that "possesses human-like intelligence and can perform any intellectual task that a human can." The Singularity refers to a threshold at which technology becomes self-improving, allowing it to progress exponentially. Shumer is correct that there are good reasons to think that progress has been made toward both AGI and the Singularity. OpenAI's latest coding model, GPT-5.3-Codex, helped create itself. Anthropic has made similar claims about recent product launches. And there's no denying that generative AI is now so good at writing code that it's decimated the job market for entry-level coders. It is absolutely true that generative AI is progressing rapidly and that it will surely have big impacts on everyday life, the labor market, and the future. Even so, it's hard to believe a weather report from Chicken Little. And it's harder still to believe everything a car salesman tells you about the amazing new convertible that just rolled onto the sales lot. Indeed, as Shumer's post went viral, AI skeptics joined the fray. This Tweet is currently unavailable. It might be loading or has been removed. This Tweet is currently unavailable. It might be loading or has been removed. This Tweet is currently unavailable. It might be loading or has been removed. There are a lot of reasons to be skeptical of Shumer's claims. In the essay, he provides two specific examples of generative AI's capabilities -- its ability to conduct legal reasoning on par with top lawyers, and its ability to create, test, and debug apps. Let's look at the app argument first: I'll tell the AI: "I want to build this app. Here's what it should do, here's roughly what it should look like. Figure out the user flow, the design, all of it." And it does. It writes tens of thousands of lines of code. Then, and this is the part that would have been unthinkable a year ago, it opens the app itself. It clicks through the buttons. It tests the features. It uses the app the way a person would. If it doesn't like how something looks or feels, it goes back and changes it, on its own. It iterates, like a developer would, fixing and refining until it's satisfied. Only once it has decided the app meets its own standards does it come back to me and say: "It's ready for you to test." And when I test it, it's usually perfect. I'm not exaggerating. That is what my Monday looked like this week. Is this impressive? Absolutely! At the same time, it's a running joke in the tech world that you can already find an app for everything. ("There's an app for that.") That means coding models can model their work off tens of thousands of existing applications. Is the world really going to be irrevocably changed because we now have the ability to create new apps more quickly? Let's look at the legal claim, where Shumer says that AI is "like having a team of [lawyers] available instantly." There's just one problem: Lawyers all over the country are getting censured for actually using AI. A lawyer tracking AI hallucinations in the legal profession found 912 documented cases so far. It's hard to swallow warnings about AGI when even the most advanced LLMs are still completely incapable of fact-checking. According to OpenAI's own documentation, its latest model, GPT-5.2, has a hallucination rate of 10.9 percent. Even when given access to the internet to check its work, it still hallucinates 5.8 percent of the time. Would you trust a person that only hallucinates six percent of the time? Yes, it's possible that a rapid leap forward is imminent. But it's also possible that the AI industry will rapidly reach a point of diminishing returns. And there are good reasons to believe the latter is likely. This week, OpenAI introduced ads into ChatGPT, a tactic it previously called a "last resort." OpenAI is also rolling out a new "ChatGPT adult" mode to let people engage in erotic roleplay with Chat. That's hardly the behavior of a company that's about to unleash AI super-intelligence onto an unsuspecting world. This Tweet is currently unavailable. It might be loading or has been removed. This article reflects the opinion of the author.

The flawed assumptions behind Matt Shumer's viral X post on AI's looming impact | Fortune

AI Influencer Matt Shumer penned a viral blog on X about AI's potential to disrupt, and ultimately automate, almost all knowledge work that has racked up more than 55 million views in the past 24 hours. Shumer's 5,000-word essay certainly hit a nerve. Written in a breathless tone, the blog is constructed as a warning to friends and family about how their jobs are about to be radically upended. (Fortune also ran an adapted version of Shumer's post as a commentary piece.) "On February 5th, two major AI labs released new models on the same day: GPT-5.3 Codex from OpenAI, and Opus 4.6 from Anthropic," he writes. "And something clicked. Not like a light switch...more like the moment you realize the water has been rising around you and is now at your chest." Shumer says coders are the canary in the coal mine for every other profession. "The experience that tech workers have had over the past year, of watching AI go from 'helpful tool' to 'does my job better than I do,' is the experience everyone else is about to have," he wries. "Law, finance, medicine, accounting, consulting, writing, design, analysis, customer service. Not in ten years. The people building these systems say one to five years. Some say less. And given what I've seen in just the last couple of months, I think 'less' is more likely." But despite its viral nature, Shumer's assertion that what's happened with coding is a prequel for what will happen in other fields -- and, critically, that this will happen within just a few years -- seems wrong to me. And I write this as someone who wrote a book (Mastering AI: A Survival Guide to Our Superpowered Future) that predicted AI would massively transform knowledge work by 2029, something which I still believe. I just don't think the full automation of processes that we are starting to see with coding is coming to other fields as quickly as Shumer contends. He may be directionally right but the dire tone of his missive strikes me as fear-mongering, and based largely on faulty assumptions. Shumer says that the reason code has been the area where autonomous agentic capabilities have had the biggest impact so far is the AI companies have devoted so much attention to it. They have done so, Shumer says, because these frontier model companies see autonomous software development as key to their own businesses, enabling AI models to help build the next generation of AI models. In this, the AI companies' bet seems to be paying off: the pace at which they are churning out better models has picked up markedly in the past year. And both OpenAI and Anthropic have said that the code behind their most recent AI models was largely written by AI itself. Shumer says that while coding is a leading indicator, the same performance gains seen in coding arrive in other domains, although sometimes about a year later than the uplift in coding. (Shumer does not offer a cogent explaination for why this lag might exist although he implies it is simply because the AI model companies optimize for coding first and then eventually get around to improving the models in other areas.) But what Shumer doesn't say is that another reason that progress in automating software development has been more rapid than in other areas: coding has some quantitative metrics of quality that simply don't exist in other domains. In programming, if the code is really bad it simply won't compile at all. Inadequate code may also fail various unit tests that the AI coding agent can perform. (Shumer doesn't mention that today's coding agents sometimes lie about conducting unit tests -- which is one of many reasons automated software development isn't foolproof.) Many developers say the code that AI writes is often decent enough to pass these basic tests but is still not very good: that it is inefficient, inelegant, and most important, insecure, opening an organization that uses it to cybersecurity risks. But in coding there are still some ways to build autonomous AI agents to address some of these issues. The model can spin up sub-agents that check the code it has written for cybersecurity vulenerabilites or critique the code on how efficient it is. Because software code can be tested in virtual environments, there are plenty of ways to automate the process of reinforcement learning-where an agent learns by experience to maximize some reward, such as points in a game-that AI companies use to shape the behavior of AI models after their initial training. That means the refinement of coding agents can be done in an automated way at scale. Assessing quality in many other domains of knowledge work is far more difficult. There are no compilers for law, no unit tests for a medical treatment plan, no definitive metric for how good a marketing campaign is before it is tested on consumers. It is much harder in other domains to gather sufficient amounts of data from professional experts about what "good" looks like. AI companies realize they have a problem gathering this kind of data. It is why they are now paying millions to companies like Mercor, which in turn are shelling out big bucks to recruit accountants, finance professionals, lawyers and doctors to help provide feedback on AI outputs so AI companies can train their models better. It is true that there are benchmarks that show the most recent AI models making rapid progress on professional tasks outside of coding. One of the best of these is OpenAI's GDPVal benchmark. It shows that frontier models can achieve parity with human experts across a range of professional tasks, from complex legal work to manufacturing to healthcare. So far, the results aren't in for the models OpenAI and Anthropic released last week. But for their predecessors, Claude Opus 4.5 and GPT-5.2, the models achieve parity with human experts across a diverse range of tasks, and beat human experts in many domains. So wouldn't this suggest that Shumer is correct? Well, not so fast. It turns out that in many professions what "good" looks like is highly subjective. Human experts only agreed with one another on their assessment of the AI outputs about 71% of the time. The automated grading system used by OpenAI for GDPVal has even more variance, agreeing on assessments only 66% of the time. So those headlines numbers about how good AI is at professional tasks could have a wide margin of error. This variance is one of the things that holds enterprises back from deploying fully automated workflows. It's not just the output of the AI model itself might be faulty. It's that, as the GDPVal benchmark suggests, the equivalent of an automated unit test in many professional contexts might produce an erroneous result a third of the time. Most companies cannot tolerate the possibility that poor quality work being shipped in a third of cases. The risks are simply too great. Sometimes, the risk might be merely reputational. In others, it could mean immediate lost revenue. But in many professional tasks, the consequences of a wrong decision can be even more severe: professional sanction, lawsuits, the loss of licenses, the loss of insurance cover, and, even, the risk of phyiscal harm and death -- sometimes to large numbers of people. What's more, trying to keep a human-in-the-loop to review automated outputs is problematic. Today's AI models are genuinely getting better. Hallucinations occur less frequently. But that only makes the problem worse. As AI-generated errors become less frequent, human reviewers become complacent. AI errors become harder to spot. AI is wonderful at being confidently wrong and at presenting results that are in impeccable in form but lack substance. That bypasses some of the proxy criteria humans use to calibrate their level of vigilance. AI models often fail in ways that are alien to the ways human fail at the same tasks, which makes guarding against AI-generated errors more of a challenge. For all these reasons, until the equivalent of software development's automated unit tests are developed for more professional fields, deploying automated AI workflows in many knowledge work contexts will be too risky for most enterprises. AI will remain an assistant or copilot to human knowledge workers in many cases, rather than fully automating their work. There are also other reasons that the kind of automation software developers have observed are unlikely for other categories of knowledge work. In many cases, enterprises cannot give AI agents access to the kinds of tools and data systems they need to perform automated workflows. It is notable that the most enthusiastic boosters of AI automation so far have been developers who work either by themselves or for AI-native startups. These software coders are often unencumbered by legacy systems and tech debt, and often don't have a lot of governance and compliance systems to navigate. Big organizations often currently lack ways to link data sources and software tools together. In other cases, concerns about security risks and governance mean large enterprises, especially in regulated sectors such as banking, finance, law, and healthcare, are unwilling to automate without ironclad guarantees that the outcomes will be reliable and that there is a process for monitoring, governing, and auditing the outcomes. The systems for doing this are currently primitive. Until they become much more mature and robust, don't expect enterprises to fully automate the production of business critical or regulated outputs. I'm not the only one who found Shumer's analysis faulty. Gary Marcus, the emeritus professor of cognitive science at New York University who has become one of the leading skeptics of today's large language models, told me Shumer's X post was "weaponized hype." And he pointed to problems with even Shumer's arguments about automated software development. "He gives no actual data to support this claim that the latest coding systems can write whole complex apps without making errors," Marcus said. He points out that Shumer mischaracterizes a well-known benchmark from the AI evaluation organization METR that tries to measure AI models' autonomous coding capabilities that suggests AI's abilities are doubling every seven months. Marcus notes that Shumer fails to mention that the benchmark has two thresholds for accuracy, 50% and 80%. But most businesses aren't interested in a system that fails half of the time, or even one that fails one out of every five attempts. "No AI system can reliably do every five-hour long task humans can do without error, or even close, but you wouldn't know that reading Shumer's blog, which largely ignores all the hallucination and boneheaded errors that are so common in every day experience," Marcus says. He also noted that Shumer didn't cite recent research from Caltech and Stanford that chronicled a wide range of reasoning errors in advanced AI models. And he pointed out that Shumer has been caught previously making exaggerated claims about the abilities of an AI model he trained. "He likes to sell big. That doesn't mean we should take him seriously," Marcus said. Other critics of Shumer's blog point out that his economic analysis is ahistorical. Every other technological revolution has, in the long-run, created more jobs than it eliminated. Connor Boyack, president of the Libertas Institute, a policy think tank in Utah, wrote an entire counter-blog post making this argument. So, yes, AI may be poised to transform work. But the kind of full task automation that some software developers have started to observe is possible for some tasks? For most knowledge workers, especially those embedded in large organizations, that is going to take much longer than Shumer implies.

Guy Who Wrote Viral AI Post Wasn't Trying to Scare You

You probably don't know Matt Shumer's name, but there's a pretty good chance you're familiar with his thoughts about AI. On Tuesday, Shumer published an essay to X, titled "Something Big Is Coming," which almost immediately caught fire online. (According to X's not-always-reliable metrics, it stands at 73 million views as of Thursday morning.) In it, Shumer, the founder of an AI company, warns that enormous advances in technology are poised to reshape society much more quickly than most people realize. He analogizes artificial intelligence's rapid improvement in recent months to the beginning of the COVID pandemic -- a looming, seismic societal change that only a small faction is really paying attention to. And he warns that the tech sector is the canary in the AI coal mine: "We're not making predictions. We're telling you what already occurred in our own jobs, and warning you that you're next." As Shumer's post ricocheted around the internet, it drew a predictably divided response. Some saw it as an incisive warning of things to come, while others dismissed it as another piece of disposable AI hype, or a naked money grab. I caught up with Shumer on Thursday to discuss the overwhelming response to his essay, why he used AI to help write it, and whether all of our jobs are actually in immediate danger. Your essay has been making the rounds in a way that few things do -- it broke social-media containment. What has the reaction been like? It's insane, because I didn't expect this. I didn't expect anything close to this. I originally wrote it for my mom and dad because -- I was home with them this weekend for the Super Bowl -- I'm 26 and I was trying to explain to them what was going on. I felt that there was an inflection point when GPT-5.3-Codex came out. I tried it and was like, oh my God, this is not just a step better, this is massively better, and it's the first sign of something a little scary. The way I view it, AI labs have focused on training models that are really good at writing code, and that's really important because what we see in the engineering space is like a year ahead of what everybody else has access to. So a model today is, let's say, at level 50 at writing code, but level 20 at law or something of the sort. The next model will be probably at level 100 on code and level 50 on law. It woke me up, and I felt that I had to share it. I was looking around and I was like, what can I give to my parents to help them understand this, so that they're not just thinking their idiot son -- that's probably a terrible way of putting it, but you know what I mean -- is saying "this is happening" and they have no way to know what or not to believe it? There are a lot of pieces of great writing in this space, but they're all extremely technical, and I think that's part of the reason people don't understand what's coming. They're written for nerds by nerds. They almost take pride in sounding as smart as possible. So I figured it would probably be important to write something that they could understand. And as I wrote it, I realized it could actually help other people. I decided to post it and it quickly broke containment. I have friends who are very much outside of the tech bubble and it's being passed around their offices, and they're texting me and it's a surreal experience. But I'm glad it's happening. I didn't expect my article to be the thing that did it, but it needed to happen. People need to understand what's coming. It may not affect them today, it may not affect them in a year, but at some point it will, and I'd rather people be aware and have the opportunity to prepare than just be blindsided by forces that they can't really control. Did you use AI to write any of it? I did. I actually posted a little bit about this because I think it's important for people to know. People are responding like, "look, this is AI-written. You should ignore it." And it's not entirely AI-written. I used it to help edit to sort of iterate my ideas and the ways of phrasing things, and it was incredibly helpful, but that's kind of the point. If this was helped by AI and got millions of views, it's clearly good enough. I didn't say "go write this article." What I did was feed it a bunch of articles that I have read over the years that I think articulate these points really well. I said, "Here's what I agree with, here's what I disagree with. Here's my unique spin and take on this." And then I said, "interview me -- ask me dozens of questions." And I spent over an hour answering those first wave of questions. Then we repeated it, and basically I ended up building this huge dossier of everything I believe, and everything I wanted to explain. And then for each thing, asked:how can I explain it in a way that's actually useful and understandable by the average person? I ended up writing a first draft based on that. Once that was done, I passed my first draft into the AI and I said, "Hey, from an editorial perspective, can you critique this? It gave me feedback and I adjusted it." So it was very much like having a co-writer, and it clearly worked pretty well. I wouldn't say it's the best-written thing ever. I didn't expect it to do this, and if I did expect this sort of virality, I would've put some more work into a lot of parts. One common critique of the article I've seen is that the AI revolution you describe in coding doesn't neatly apply to other fields, since coding is such a discrete task. Here's one post I saw: "Coders freaking out that it's replacing them and extrapolating from their extremely weird domain (as in unusual among knowledge work) to all of work is going to be a major theme of 2026 and kind of embarrassing by 2027." What's your response to that? I understand why people think that, and I think it's very easy to feel like, Oh, the AI can't do the thing I do because X, Y, and Z. There were arguments like that with coding for a very long time, but what we've seen time and again is if the AI labs are sufficiently funded and pick a goal to go after, whether it's to make something that generates videos or that writes code, given enough money and enough time and sufficient incentive to do it from a financial perspective -- which clearly there is -- they solve it. It's this very interesting technology where whatever data you have, if you train the model on it, it can learn it. My dad's a lawyer. Is AI going to stand in front of a courtroom? No, it's not, but I worry that associates are going to have a harder time getting hired. I had dinner with my lawyer a couple nights ago and he was saying that they're using one of these off-the-shelf programs and it's already at the level of about a second-year associate. Do I know exactly what that means personally? No, but I've talked with enough people who are pushing this stuff in their industries who aren't in the bubble, but are just like, "Hey, I want to see where this is going," and the rate that they're saying it's improving at is pretty clear. When you actually break a job into its steps, anything that could be done on a computer can theoretically be done by these models. But I don't think -- and I wanted to make this clear in the article, and if I knew how viral this is going to go, I probably would've spent more time trying to make this clearer -- that just because the AI can do something doesn't mean it's going to immediately proliferate across the economy. There are so many structural things, whether it's regulations, standards, or people's comfort with this sort of stuff, and that means that for certain industries it's going to take more time than others. Code is in this crazy place today where people are saying it's solved and you can build anything, and that is true, but we're still figuring out what it means for jobs. I don't know. No one knows. I think each industry is going to have its own separate reckoning, and it's going to look different for every industry 10 years from now. I think almost everything will be extremely different and almost unrecognizable, but in the interim, everybody has to figure out what this means for them and their industry. But assuming that the AI just can't do their thing and that their thing is special is not the right approach. Maybe that's true, but if there's even a 20 percent chance it's not, it's worth preparing. Another related version of that critique is that for a lot of jobs, dealing with other people is a huge part of it. I'm sure AI could beat a law associate at document review, if not now, then soon. But then you have to actually deal with the client. Most people's jobs have components like that. A lot of what we're seeing and what people know as AI today isn't actually the state of the art of what actually exists. I'm assuming most people that use it are using the free version. The paid version is dramatically better, but there's a whole level above that of more truly agentic systems, and that's the sort of scary stuff right now. When I go and I use AI to build an app, I am not saying, "Hey ChatGPT, build an app." I have a specialized program that has access to everything my computer has access to, and it can use tools like a person and go off and do things. So I say, "Hey, can you get this on the internet and then see if you can find some early users on Reddit and communicate to them that they should try it?" That is actually possible today. That is a little spooky. It's spooky as hell, and I've been one of these people that has been predicting this for years but predicting it and seeing it is a whole different story. Although many people were suspicious of the fact that at the end of your article, you advise that people pay for certain products and follow you on X to keep up with AI news. The following me on Twitter thing -- I agree. The AI products -- I have no stake in Anthropic, I have no stake in OpenAI. They don't pay me or anything. I can see why people might think that, but in fact, I have paid a lot of OpenAI bills over the years when I've tested this stuff. For example, for one of the startups I invested in, you basically worked with them to allow your AI to not just chat back and forth with you on ChatGPT, but to give them access to their own email inbox, where they can actually reach out and chat with other people and other AIs. I also oscillate back and forth between "this is interesting" and "this is terrifying," and I don't know which one is right. I think they're both right. It also strikes me that a hallmark of the AI industry from the beginning, as far as I can tell as a lay observer, is that people love to make sweeping predictions about what's going to happen in a year in two years. I'm pretty sure that in 2023 and 2024, I was hearing that by 2026, white-collar jobs would be totally endangered. Ten years ago, Geoffrey Hinton, an AI pioneer, famously predicted that radiologists would be obsolete by 2020. That did not happen, and it still hasn't come close to happening. Do you find it a little uncomfortable to be making these somewhat apocalyptic forecasts? Yes. The way that I think about it is, do I know this to be 100 percent true? Do I know this to be absolutely certainly going to happen? No, I don't. However, given what I've experienced and getting the preview that I have into the industry, I think there's a better-than-not chance it will. I find the analogy of the pandemic you use a little off because in January and February 20th, 2020, it's true that a very small number of people were actually paying attention to what was happening in China, and you had to be following the right people on Twitter. In this case -- to take an old school barometer of success, the Time Person of the Year in 2025 was "the architects of AI." These companies are widely used and in the news, and I've had a million conversations with people who are worried about the implications for their jobs. It's exactly under the radar. It's not. But people talk a lot about Terminator-style doom, and I don't see many people talking about the impact on jobs. In theory, if this could happen two years from now to your job, if your job happens to be one of the more exposed ones, maybe you should just focus a little more on saving today. That's the angle I wanted to take it from. Tell me more about this agentic stuff where, where AI interacts with the world by itself. Where do you think that could go next? I think it's just reliability. One of the key things that I've learned in AI over the years, because I've been doing this since 2019 -- I dropped out of college after realizing what this was going to do and I realized if I didn't put everything I had into this, I'd regret it for the rest of my life. Basically the best rule of thumb I can give to anybody, and this has been the one thing that's held true, is it's not about a specific prediction. It's not saying it's going to do X, Y, Z at any given point. It's just if a model can kind of sort of do something today, even if it's not good at it, even if it's unreliable in a year or two years, you can bet that it'll eventually be near perfect at that thing. I can't say the models today are reliable at using a computer, and I've actually been in this part of the space. My company actually built the first publicly available AI that could use a web browser and actually go in and order you a burrito. And was it useful at the time? No, because it got it wrong 50 percent of the time. But now we're at 80, 90 percent. It can almost use a computer today, which means in a year or two, you can expect that these things will be nearly perfect, probably better than people at using a computer. So if there's a task that can be done on a computer by a human and doesn't require going somewhere in person, it's very likely that AI will be able to do it reliably and well. Getting to 90 percent reliability for something like that or 95 or even 99 is great, but doesn't it have to be 100 percent? Because you don't want to entrust an AI to do something and it screws up one percent of the time, and it could be a very consequential screwup. I've thought about this a lot, and I go back and forth. You could take the argument that 99 percent reliability isn't enough, but then I've hired and worked with a lot of people over the last six years or so, and I would say that the rate of success is far lower than 99 percent for most things. So it is very much a perception thing. There are also tricks, and I think this is one of the things labs should be focusing on that they're not, to mitigate this. If you just tell the AI, "Hey, I want you to do this thing," there might be a one percent failure rate, But if you then actually have a system that has two ais, you say "AI One, do this thing," and then when it's done, you say, "AI Two, did they do it correctly or do they make a mistake?" To check the work. Exactly. You actually see that the failure rate goes way, way, down. This has been documented since 2020, when AI wasn't great at the time. If it was writing a paragraph and most of the paragraphs were awful, you could say, "Hey, can you generate 20 different versions?" And then you actually have a different AI critique them and pick the best one? The results are far better, and I think a lot of those sorts of things haven't been implemented yet in full. They're starting to be. You have been doing this a long time. There's been a widespread feeling among people who do this work -- at least among coders -- that they may be rendered obsolete or they already are being rendered obsolete, and it's bittersweet. The new tools are both incredibly helpful and sobering. They bring up all kinds of questions of human beings' value in the world. Are you feeling that in your own work right now? This is probably the trickiest question you've asked, because there's so many facets and I hope I hit on the ones that are important. What I've seen in my industry is a bifurcation, is the best way I can describe it. People who are really loving what they do, who are already working insanely hard and are adopting this, are pulling away in an extremely strong way. Somebody who was a top percentile engineer is now 10 or 20 times as effective as they were before, and they can do the work of many more people. Then you have the other side of things, which is that folks who really aren't determined and aren't top percentile already are not really getting the value out of it that others are, and it's not making that big of a difference for them. And I'm worried because we kind of have this social contract where you go to college, you get a job, and you'll be taken care of. But because AI is so skewed, at least in engineering, towards the hard workers already -- some people are just trying to get by, which is a totally fair thing to do. Not everybody wants to be exceptional. They're struggling. I'm a hard worker, but that doesn't mean everybody else should be screwed. And I don't know if it's going to be the same for every industry, but that's what I'm seeing here. That whole social contract -- going to college and getting taken care of -- has been getting harder in most industries anyway, and this could really magnify that. I think it's particularly an American thing to try to be an exceptional striver. It's easy to imagine other places, like Europe, resisting AI more than we do. Which might be a good thing. There are people in my life that I love who, even without AI, are really struggling to get work right now. Look I didn't put this out to scare people, although I understand that there are some elements of that. My goal is to help people see what they might be neglecting, what's not in their circles yet, because they should be able to know and make their own decisions for how to prepare or not prepare. It feels unfair that so many people think AI is a nothingburger when it's clearly not. Maybe it's not everything I say -- I think it will be -- but it's not a nothingburger, and no matter what, people should be thinking at least a little bit about it. And my hope is that this just gets people talking and thinking. So you're basically aiming at a place like Bluesky, where the idea that AI can do anything useful at all gets immediately swatted down. It's funny how people get into their factions about stuff like this. Everything takes on a political valence. It shouldn't. It should be what you need as a person. People get too tribal about things. It's different for everybody, and everybody should have a different response to this. For some people, it truly won't matter. Even if everything I say comes to pass, my nurse in a hospital isn't being replaced anytime soon. They shouldn't -- at least I don't think -- worry, but some people should, and I think it's important that they know. This interview has been edited for length and clarity.