Introduction: The Overhyped AI Dividend Has Nothing to Do with Ordinary People
Everyone online is trumpeting how AI boosts efficiency, how this is a new wave of technological dividends, and how it’s a chance for ordinary people to turn their lives around. But only those of us truly in the thick of it—the ordinary workers—know the truth: this “dividend” was never ours. Take me, for example. I once made tens of thousands a month, and at my peak I could pull in several months’ worth of year-end bonus plus hundreds of thousands in stock options. But this AI revolution has erased every advantage I’d accumulated before. My income dropped sharply, my year-end bonus went to zero, and though I survived several rounds of big layoffs, there were simply no more opportunities for promotion. By the end of last year I was laid off anyway—something I knew would happen sooner or later. After losing my job this time, I just started looking for work, and let me tell you: not only can’t I find a job at the same salary, I’m having to fight tooth and nail just to survive on a pay cut. I jumped on the bandwagon of deep learning and using AI—in my previous job I’d even built a lot of RAG Q&A bots—yet I still couldn’t escape the fate of being laid off. During this unemployed stretch I’ve done vibe coding and built lots of products, and the wildly popular OpenClaw—I even installed it on my own little mini-PC and used it intensively for a while. And in the end all I got was one realization: AI’s magic belongs to other people; the exhaustion and the struggle are mine alone. And this exhaustion has long since escalated into a career burnout that’s hard to shake off. This isn’t just my personal experience—it’s a shared predicament of countless white-collar workers, as many recent reports have revealed: AI hasn’t lightened our burden at all; it’s made career burnout worse.
The Dilemma: Don’t Learn AI and You’re Unemployed; Learn It and You Still Can’t Break Through
We’re pushed by the times into being dragged into this AI rat race. Companies shout “AI boosts efficiency,” which in essence means replacing manpower with AI and squeezing costs—where one position used to need three people, now one person plus a single AI can get it done, and the other two are simply labeled “redundant” and become victims of the layoff wave. So “don’t learn AI and you’re unemployed” becomes a sword hanging over every worker’s head. We have no choice but to spend time and energy learning AI tools and adapting to an AI-dominated work model—as if not keeping up means being completely abandoned by this era.
But what’s most helpless is this: even if we throw everything into keeping pace, we still can’t squeeze into the circle of AI dividends. I naively thought that learning to use AI would let me keep my job, or even get promoted and a raise. But reality hit me hard: I learned AI and still couldn’t find a job at my old salary, and had to accept a pay cut. Even when I delivered AI’s value at work—helping the company boost efficiency and cut costs—what I got in the end could still be a layoff notice. Because I gradually understood: the dividend of AI efficiency was never meant for ordinary workers.
The Truth of the Dividend: Capital Harvests, Ordinary People Are Just “Tool People”
Those who truly profit from AI’s development are the big capitalists and industry oligarchs who control AI productivity—Nvidia, which makes the computing power; Google, Microsoft, OpenAI, which develop the models; and the big tech giants that control internet traffic. They sell computing power, sell models, charge service fees, and make a killing; they use AI to lay off staff and cut headcount, converting every bit of the saved labor cost into their own profits and stock price. And we ordinary workers are nothing but “tool people” on their dividend chain—consumables pushed to adapt to AI and squeezed for surplus value beyond measure. We learn AI not to feast on the dividend, but to pay a “life-preservation fee,” barely clinging to our positions, surviving on borrowed breath.
The Truth About AI Coding: Beneath the Filter, Nothing but Potholes and Exhaustion
Even the AI coding that’s been hyped as godlike, plus the recently viral OpenClaw, turn out to be full of gripes in actual use—nothing like the magic advertised. I used AI vibe coding intensively for as long as half a year, with all kinds of tools—Trae, Cursor, Kimi CLI, qrcode—and I also deployed OpenClaw on a mini host. The former is, in essence, just an advanced probabilistic collage machine, completing requirements by randomly generating code. Every generation’s quality varies wildly, full of holes. Even if you define coding standards and requirements, once the context gets too long it still makes things up and reinvents the wheel; you have to communicate and retry over and over just to barely get the result, like drawing a lottery every time. The latter is even more absurd: aside from burning through tokens like crazy, the things it can do are far less efficient and reliable than just calling the API directly—pure added hassle.
AI vibe coding really can quickly implement basic functionality and save us a bit of time on boilerplate code. But it’s worlds apart from production-grade products—no fault tolerance, no logging or alerting, no security validation, no consideration of performance or concurrency. Ship it and you’re planting landmines everywhere, and the time spent troubleshooting later far exceeds the time saved. What’s more torturous is that AI often turns correct code into broken code, and then you have to communicate and correct it over and over, dragging back and forth. Originally I could hand-write a piece of code with a coherent train of thought and clear logic that ran stably—but with a mindset of “optimizing efficiency,” I let AI help me revise it, and the more it revised, the messier and more broken it got; asking it to fix the bug only spawned more new vulnerabilities.
This back-and-forth communication cost and verification cost has long since canceled out whatever so-called “efficiency gain,” and in the end it’s more tiring than thinking and writing the code myself. More crucially, this communication cost never disappears—it used to happen in meetings, aligning requirements across departments; now it’s just shifted onto the user, into the repeated back-and-forth with AI. From a neuroscience perspective, this tiredness is essentially the brain’s “cognitive entropy increase”—the core of entropy increase is a system moving from order to disorder, requiring a continuous expenditure of energy to maintain order. The brain’s cognitive activity is exactly a process of fighting cognitive entropy increase. When I write code myself, my thoughts are coherent and the logic loops shut; the brain is in an ordered cognitive state, and it actually consumes less cognitive energy. But when using AI to write code, repeated communication, correction, and verification throw the brain’s cognitive system into disordered chaos, and to fight that chaos the brain needs to burn more energy, so the fatigue surges along with it. The brain’s working memory has limited capacity—it can only handle 4 to 7 information units at once—yet I have to simultaneously write instructions for the AI, verify holes, and repeatedly communicate and correct, juggling multiple fragmented tasks; this jumbled information worsens cognitive entropy increase. Add to that the AI’s flip-flopping between getting it right and getting it wrong, forcing me to switch cognitive states frequently and further scrambling an orderly cognitive process, and pile on the emotional entropy increase from the expectation gap of “AI should boost efficiency but instead adds chaos”—under this double drain, of course it’s more tiring than writing code myself. I’ve never denied that AI genuinely boosts efficiency for basic work to a degree—things like writing basic templates, simple scripts, and repetitive CRUD, low-skill tasks with strong repetition, AI really can save a lot of trouble. But if you say AI can completely replace an engineer, that’s still not achievable as far as I can see—an engineer’s logical thinking, problem-troubleshooting ability, deep understanding of the business, and ability to backstop complex scenarios are all things AI cannot replicate. Even so, the efficiency gains AI brings will still leave many people facing unemployment risk. Companies were already eager to squeeze labor costs, and AI’s arrival gives them even more of an excuse to cut headcount—work that used to need many people gets handled by a few people plus AI, and the rest are forced into unemployment and a job-hunting predicament.
The whole internet’s hyped-up AI coding, and things like OpenClaw, are nothing but a scam beneath a filter. Those who tout AI as omnipotent are either marketing accounts and course-sellers who don’t understand production deployment—showing off only the simplest demos and never mentioning the potholes behind them—or they’re corporate higher-ups who don’t bear frontline responsibility, who look only at code output volume and tool usage rates, and don’t care how much time we engineers have to spend cleaning up the mess AI leaves behind.
Those of us who actually do production-grade delivery and have to backstop online bugs—none of us blindly worship these AI tools. They can help us write some scaffolding and template code, help us check syntax and fill in comments, but they’re absolutely not up to the key work of core business logic and high-availability architecture. I felt this deeply in an interview the day before yesterday—I went to interview for an AI agent developer role at an e-commerce company at EFC in Future Sci-Tech City. The person in charge kept talking about “empowerment” and “innovation,” one moment saying they wanted to do automated livestream clip generation and AI-automated video editing, the next saying they wanted to build a scraper to pull data from Douyin and Kuaishou livestream rooms for data analysis, and also make digital humans. It sounded elaborate and high-end—but then I learned the whole company’s R&D resources were just 4 people. That kind of operation, compressing manpower to the extreme, is simply jaw-dropping. What’s more absurd, I later found out the previous developer had quit less than 2 weeks after joining—I guess he was crushed by that absurd staffing and overloaded demands.
AI Worsens Career Burnout: It Looks Like Efficiency, But It Actually Adds Burden
The exhaustion AI brings is no longer simple work fatigue—it’s career burnout that seeps into your bones. On this point, many recent reports and surveys give the truest corroboration. Take PR firm founder Anurag Garg, for example: after ChatGPT launched at the end of 2022, he was eager for his 11-person team to all use AI to handle daily tasks like brainstorming client stories, media outreach, and meeting transcription. He thought it would boost efficiency and keep them competitive, but the result backfired: not only did the employees fail to improve efficiency, they actually doubled their working hours because of having to write prompts for AI and verify erroneous output, and stress and tension skyrocketed. On top of that, AI platforms updated frequently, so they had to spend extra time learning new features, and the team complained endlessly. In the end Garg had to give up forcing everyone to use AI and return to the team working autonomously, and only then did everyone regain their working state.
How similar this is to my experience—using AI vibe coding means repeated error correction, using OpenClaw means suffering crazy token consumption and low efficiency, and the company I interviewed with used AI as an excuse to compress manpower and make a few people shoulder overloaded work. All of this constantly drains our energy and breeds burnout. And the survey data speak even more clearly: an Upwork survey of 2,500 knowledge workers across four countries showed that 96% of executives expect AI to boost productivity, 81% therefore increased demands on employees, but 77% of employees said AI actually reduced productivity and increased their workload, and 47% had no idea how to achieve the efficiency gains their employers expected.
What’s more worrying is that AI is sharply increasing the risk of career burnout. A Resume Now survey of 1,150 Americans showed 61% believed using AI at work would increase the likelihood of career burnout—and among those under 25, the figure was as high as 87%—while 43% believed AI would damage work-life balance. Behind this is a double over-extraction brought by AI: on one hand, companies use AI’s name to compress manpower, making a few people do the work of many—like that e-commerce company I interviewed with, where 4 developers had to take on a pile of high-end projects, and the previous employee quit 2 weeks after joining, essentially because he was burned out beyond endurance; on the other hand, AI tools proliferate, and we not only have to learn all kinds of tools but switch among them, which actually increases time cost and mental burden.
As research by the work-management platform Asana shows, among employees who use 6 to 15 work apps, 15% miss messages because of too many tools; among those using 16 or more, 23% see reduced efficiency and scattered attention due to frequent tool-switching. UCLA management professor Cassie Holmes put it perfectly: “Using multiple apps requires extra time to learn and switch, and that lost time is painful, because we’re very sensitive to wasted time.” And Leah Steele, a coach who transitioned from law, has seen far too many peers fall into burnout because of AI—after AI was introduced, lawyers’ caseloads exploded from 50 to 250, and they had to learn new tools, shoulder double the workload, and worry about being replaced by AI. This high-load, low-control environment makes many people lose their passion for work and sink into deep burnout.
Even more ironic: AI is supposed to streamline workflows, yet it traps us in a cycle of “endless tools to learn, endless validation to do, endless pressure to bear.” Garg’s predicament is the best example: he not only had his team use ChatGPT but also introduced multiple AI tools like Zapier and Perplexity, and the result was a proliferation of tools and utter chaos—even he felt breathless and utterly frustrated, and in the end could only return to a simple working model. This is exactly our situation as ordinary people: to avoid being eliminated, we’re forced to learn all kinds of AI tools, but the more tools, the heavier the burden and the stronger the burnout—and over time, even our passion for work and sense of value get worn away.
The Fig Leaf: “One-Person Company” Can’t Conceal the Cruelty of the Unemployment Wave
These days we’re trapped in a dilemma: don’t learn AI and we’ll be laid off; learn it and we not only don’t get a share of the dividend but also endure higher work intensity, lower pay, and maybe still can’t escape replacement. We’re like puppets on invisible strings—competing over AI on one hand, over-extracted by AI on the other, exhausted and yet unable to escape. What’s more contemptible is that “one-person company” is now being touted everywhere, as if after losing your job everyone can go solo and become their own boss. But that’s nothing more than a self-deceiving fig leaf under the unemployment wave.
The so-called “one-person company” seems to point the unemployed to a way out, but it ignores every real obstacle an ordinary person faces in starting a business: no startup capital, no resource channels, no stable customer base. With just one person, one computer, and so-called AI tools, you can barely survive, let alone profit. Those who tout the “one-person company” will never tell you that most unemployed people can’t even find a stable job—so how could they single-handedly prop up a company? They merely use the pretty illusion of “going into business for yourself” to cover up the cruel reality of the unemployment wave, so that those forced out of work stop complaining about the environment and instead blame themselves for not trying hard enough, for not daring to try.
Conclusion: Poke Holes in the Lies, and May Workers Be Treated Gently
I don’t want to deny AI’s value—it truly is changing the world and boosting social efficiency. But I want even more to poke holes in those lies about the AI dividend: AI was never a chance for ordinary workers to turn things around; it’s a new tool for capital to harvest profits and over-extract manpower. We work hard to learn AI and all the tools that go viral—not to seize the dividend, but just to gain one more chance at survival in this rat race. And what this rat race brings is not only the risk of unemployment but increasingly severe career burnout, leaving us struggling in exhaustion and burning ourselves up under pressure.
I write this not to complain, but just to voice what countless ordinary workers feel: we’re not afraid of hard work, not afraid of learning. What we fear is working hard and learning, and still getting no reward we deserve; what we fear is being pushed along by the times while never finding our place; what we fear is that after this AI wave passes, we’re still just the dust that gets discarded; and what we fear even more is that in this endless AI rat race, we burn through all our passion and are finally crushed by career burnout.
Beneath the AI dividend, may every ordinary worker who runs hard be treated gently; may our efforts no longer be just for “staying alive,” but truly for grasping the opportunities that are ours; and may AI no longer be a tool that over-extracts from us, no longer a driver of burnout, but truly a force that lightens our load and helps us regain a sense of value in our work.
评论
0 条评论