Transcription
AI is collapsing futures, and most of us are missing what that really means. We think collapsing as in destroying. That's not what I mean here. Collapsing as in compressing is what people are missing because AI is collapsing multiple different dimensions of our work lives into a single thread pointing to the future. And we're, we're missing the deeper implications of that.
The first collapse is horizontal. Engineer, product manager, marketer, analyst, designer, opsley. These used to be distinct career paths with very distinct skill sets. They're all converging now very quickly into a single meta competency: orchestrating AI agents to get work done. If you cannot do that, none of the rest of the domain knowledge is going to matter in late 2026. And yes, I don't want to lose the fact that we still have folks who have 10 years, 15 years experience in these individual domains in front-end design, in being an operational lead, in doing deep back-end engineering. But you don't have value there unless you can do the orchestrating AI agents piece, certainly by late 2026, early 2027. That is how fast this space is developing, and I don't think most of us are ready for it.
The second collapse is temporal. The leverage you thought you could build over the next five years. The way we've been trained to think about career ladders as these steady steps: wait two, three years, next promotion. That timeline is compressing into months. The rate of AI capability improvement nearly doubled in the last year, and it's just going to keep going faster.
Both collapses point to one conclusion: Now is what matters. Not your 5-year plan, not your eventual intention to get up to speed on AI, because the future keeps arriving faster. Preparation means engagement. And I'll add one more piece here that I think is absolutely true across everyone who engages with AI productively. This is an art you learn by doing. You do not get to learn to ride a horse by reading a book, as a friend of mine called out. You do not get to learn to swim by sitting in a deck chair and watching the ocean. You just got to get in. And that's very true of AI because it's an experiential technology.
Let me go a bit deeper onto the differentiation between knowledge work roles because I think a lot of times when you see, "Oh, the knowledge work roles are collapsing. They're all going to be the same." It feels like a big claim. It feels like it's overhyped. Gartner's predicting that close to half of enterprise applications will integrate task-specific AI agents by the end of 2026. That's up from less than 5% in 2025. It's absolutely exploding. It's an eight-fold increase in just over a year. 57% of companies, as of 2025, claim to have AI agents in production. Now, those can have varying degrees of competency. They, the direction is nothing but it's exploding.
So, what this means is that specific domain AI expertise is going to be mediated through these universal skills. The differentiation is going to be whether you can apply your marketing skills, your engineering skills, your finance skills, whatever it is, in an AI agent-shaped way. Think about what a product manager does today versus two years ago. The job used to require synthesizing customer feedback, writing specs, coordinating with engineering, managing stakeholders. And now, increasingly, the job involves just prompting models to draft specs and using AI to analyze customer data. And you're often now using agents to update tickets. You're using agents to directly build in production. Your entire job is radically different. And that pattern repeats across every function. Legal teams using AI to review contracts are compressing jobs that took weeks into hours. Finance teams can now use clawed-in Excel to build projections that used to take days. Customer success teams can run AI agents that handle 80% of initial inquiries, or 90, or 95. There is going to be a fundamental turnover of skills across every one of these job families. What used to be 50 different specializations is going to converge into variations on a single theme: humans directing AI with good knowledge and good software-shaped intent toward an outcome.
I've talked about software-shaped intent before. I think it's one of the biggest skills we're missing when we direct agents. We need to think in terms of what agents can deliver within the technical ecosystem they occupy. Where is the agent's toolset? Where is the agent's memory? Where is the agent's workflow? When I direct the agent to do something, is it going to look software-shaped? As in, is it going to be an interface that adequately reads and writes data so that I can solve the problem? Software is leveraged, expressed in silicon. Fundamentally, if you know how software works, and so much of software is just reading and writing data and presenting it in a way that's useful, if you start to think in those terms, you're going to be able to apply the specific domain knowledge you have in design, in finance, in customer success, and you're going to be able to use AI agents more effectively.
Even if your job isn't building software, this used to be a product-only thing or an engineering-only thing. The idea that we now work with agents is becoming universal. And the idea that we have to think in software terms is coming out of the technical box. It's coming out of engineering. It's coming out of product. It's coming for all of us. And I want to be clear, your expertise doesn't disappear here. It just becomes foundational rather than differentiating by itself. You need to have great domain knowledge to direct AI effectively. It's part of how seniors compete in a world where everyone has access to the same AI tools. But you have to be able to leverage that through AI. And I think most people think of that still in terms of their specific domain. We have this sort of single-lane focus. And what I'm calling out is that we have a giant bottleneck on skilling. Like, all of our skills are starting to converge around this one gigantic meta-skill of driving AI agents.
The second collapse I want to talk about, I mentioned temporal collapse. This is really important, and we keep missing it. Career leverage is compressing into the present moment because AI is accelerating time. Consider even just the Sweetbench coding benchmark. AI systems could solve 4% of problems in 2023, and they've essentially solved the entire benchmark two years later. I, I don't know exactly what it's going to be when you see this video, but it's around 90-95%. Sweetbench is saturated. And the fact that we saturated it is not even the most important thing. The doubling time to get that number up is shrinking. AI progress is accelerating.
Traditional career planning assumed you had the time. Learn a skill, apply it for years, build expertise, get promoted, eventually learn that expertise and figure out how to leverage it in leadership. That timeline gave you a sense that you could plot out your growth over time and get some breathing room. You could be strategic about when to invest your learning energy. And that assumption, if you take it at face value like you could in the 2000s and 2010s, that's now catastrophically wrong because you have to assume a career path where AI is gaining speed ever more rapidly. And this creates a really tough dynamic for career planning. I don't want to sugarcoat that.
The skills that will matter in 2027 are being defined now by people engaging now. If you wait until the tech settles down, you're going to find that the early adopters have already built the workflows, established the norms, and captured the opportunities that you were waiting for. They'll have two years of compound learning while you're still figuring out the basics. So, there is, I cannot promise you a way to settle down. This is a chaotic period. There is no mature state to wait for. There is only a continuously steepening curve, and it's going to reward folks who can climb in early and go faster.
I compare AI to riding a bike. If you are going slow on a bike, it's really hard to balance, and you feel like you're never going to catch up. But experientially, when you go faster on a bike, the steadiness increases. The way to balance gets easier. And kids have so much trouble learning this. They think if they go slower, they'll be safer. But they're actually safer going faster. And that is what you have to learn with AI. You're actually safer leaning in and going faster than you are going slower because slower forces you to constantly think about breaking and stopping and slowing down and figuring out how you can adjust and work this into your existing workflow. And I see so many of us acting like kids on a bike for the first time. We're just trying to figure out how to go very slowly. I got to say, AI is going too fast for that. You got to get on the bike and go as quick as you can because that's the easiest way to balance. Like, people ask how I keep up. It's because I'm going pretty fast on the bike, and it feels really steady.
The old career model assumed your expertise appreciated over time. You would learn something valuable. It would stay valuable gradually. It would compound. The new model is really different. Your expertise atrophies. It depreciates unless you continuously update it. And the depreciation rate is accelerating because AI progress is going faster. I'm not trying to argue for panic here. It's an argument for continuous engagement. The people who are thriving now are not the ones who just go to an AI class and master it once and then coast. They're the ones who develop the meta-skill of continuously learning and adapting as the tech evolves. The half-life of any given piece of specific AI knowledge is short, and it's getting shorter. The half-life of the learning habit around AI is getting longer and more durable.
If you doubt the magnitude of what's happening, follow the money. This is the biggest CAPEX project in human history. Big tech's combined AI capital expenditure was close to half a trillion dollars in 2025. It's going to be well over half a trillion in 2026. And in total, the big five: Amazon, Microsoft, Google, Meta, Oracle, plan to add more or less at least $2 trillion in AI-related assets in the next four years. This is a tremendous amount of operational investment in what these companies believe is the future. The money is committed. AI is happening and is going to define the next era of computing so thoroughly that we've got to understand that there is no other way there. The only way out and through is AI. And that's what I mean when I say it's collapsing timelines and compressing career trajectories. There is no other way through the career path that does not include AI.
And things get uncomfortable at that point. Many people are resisting, and you don't have time to resist. If you tried ChatGPT in 2022 and said it hallucinated and you just left it, you don't have time for that anymore. You don't have time to say, "I'll wait until it matures." Like, that's an, that's like sitting by the bike and saying, "I'll wait till it gets steadier." It's not going to get steadier. You don't have time to say, "My job is immune." I got news for you. It's not immune. Anytime you are touching a computer, you are touching AI. That's how pervasive it's going to be in the next year.
And to be clear, for people who are saying, "I want to exit the ride. I want to stop a tech career." I have respect for that. And I know folks who have said, "I had my career. I think I'm done. I want to open a bookshop. I want to go and start doing carpentry." That's fine. That's great. That is something that you can choose to do intentionally. I think that's a much more productive choice than trying to stay in the industry that is converging on AI and trying to resist that. That's just not going to go well, and it's going to make everybody, including you, kind of miserable. And so, if you really think AI is not for you, I think the best thing you can do is pick that alternate career path that takes you away from the screen because if you're going to stay in fields touched by AI, which is increasingly everything to do with a computer, you're going to have to engage.
I want to close by giving you some encouragement. It is easy to look at this and to be doom and gloom. It is easy to say, "I did not make the choice for AI." I would argue none of us did. The industry as a whole made that choice, and we are all living through this moment together. We did not choose to compress timelines. We did not choose to compress career paths. It's happening for us. And I have seen over and over again that when people recognize that, and when they choose to say, "Even though I didn't get to pick this, I am going to choose to engage with AI with curiosity. I'm going to choose to engage AI and learn to ride the bike. I'm going to lean in as far as I can lean in, even if I'm not quite sure." That is going to get you so much farther. It is going to get you an accelerated rate of learning. You're going to be less overwhelmed. Curiosity literally opens up your brain. And we need openness to this AI world if we want to be able to shape it in a way that works for us. And I've seen numerous dozens and dozens and dozens of examples where people have chosen that positive path. They've chosen to lean in in widely differing fields: healthcare, tech, finance, engineering, product. I've even seen folks lean in on like small-town community building in AI. And without exception, that choice to positively lean in has taken them farther.
And so, if I can leave you with anything in the middle of a timeline that feels like it's increasingly wild and unpredictable, it's just an invitation to get on the bike with AI. You got to go faster. And you've got to be able to believe that if you lean in and pry, if you jump in and you say, "All right, I'm going to, I'm going to try something new. I'm going to try Claude Code." Whatever it is that's new to you, right? I'm going to try lovable. I'm going to try a different way of working with my chatbot. Great. And then do the next thing. And then lean in a little farther. And then lean a little farther. And you're going to go faster and faster and faster and faster. And it's going to feel steadier over time because you're going to pick up how AI works across all of these systems in your unconscious brain. The patterns will start to solidify, and you're basically learning to work with this new piece of technology in a way that feels very stable over time. And so, if I can encourage you with anything, is that going faster is safer and less scarier with AI than going slower. Cheers.