No Clanker Ever Called Me a Resource

I know I’m really late to this party (like, over a whole year late), but I still think calling chatbots and robots clankers is really funny.

As a vehicle of expressing hate, “clanker” has got a lot going for it. It has not one but two /k/ sounds, which we all know is the funniest sound. It reminds you of broken devices driving you insane with their incessant clanking. It’s easy to use in anger, with that nasaly central vowel able to carry a sharply acrid and venomous tone. Goddamn clankers again, you can imagine saying, with all the vitriol you can muster. You really clanked it up this time.

It’s not the only anti-robot slur that came out of this tide of robophobia on the internet. I’m not going to list them all here, because some veer too close to real-world hate speech, but I will mention tinskin and cogsucker as personal favorites. Does that make me a robophobe? Probably. Who cares? These machines are not aware, and Roko’s basilisk is an absurd notion devised by some proto-tech-bro armchair philosopher that I refuse to consider seriously.¹

That’s not to say I don’t feel a bit uneasy about this. I don’t condone denigrating classes of people, but robots are definitely not people, so it’s okay, right? We’re just borrowing the language of out-group hate to express the deep unease so many of us feel at the rapid emergence of generative AI as an unavoidable economic, technical, and social reality. Still, the odd combination of glee and spite that using slurs makes you feel is disquieting, even if you know it’s a joke and the targets are acceptable. It’s just too easy to enjoy it.

And as much as this all makes me chuckle, I feel this anger at clankers is playing too hard into the AI hype playbook, and trending too close to historical misdirections of anger. Are we just punching down in jest because we don’t know how to punch up in earnest? If we wink hard enough, does it make it subversive? Maybe we should just be mad at the folks foisting bots upon us, rather than the bots themselves. But what form could that even take? AI mania is so pervasive it’s hard to know where to direct our criticism. Perhaps meta-ironic hatred of robot-kind is better than just quietly accepting our fate, but I hope we can find a way to do better.

Growth

One of the reasons resisting can seem so hopeless is that tech companies are just so poweful. An overwhelming majority of Americans believe that technology companies, and social media companies in particular, have too much power and influence. What can you do to stop them? Collectively, these companies control the platforms we use to discuss them. The largest of them comprise a whole third of the S&P 500. And their influence in business and politics seems to only ever grow.

But much of that power rests on the tech industry’s perceived ability to act as a continual engine of market disruption that, through technological advancement, has transformed the way we live and work, while making a lot of people a lot of money in the process. And while the industry used to produce revolutions on a schedule, that pace of innovation has slowed in recent years, with VR, AR, crypto, and many other well-funded trends failing to ever become truly mainstream. B2B SaaS startups, who once promised to make our work lives easier, and occupied an enviable position as the darling business model of venture capital, now feel tired and dystopian. An smartphone app is now something your local supermarket chain obliges you to download to claim discounts. Big Tech may be powerful, but if coupons is all it has left to disrupt, the clock is ticking for its hypergrowth era.

Enter Generative AI. In November 2022, ChatGPT was released, and immediately captured the imagination of the general public and technology investors. It felt like a massive leap forward in technology, that as of a few years prior, was only capable of producing language that quickly veered into the surreal or nonsensical. It was a genuinely good chatbot! Like any new technology, it was raw, and it wasn’t fully clear what the limitations or applications would be. But with time and resources, that could be worked out, right?

But Big Tech did not have time. They had resources, and they flung them at LLM technology with aplomb. The industry needed a revolution. LLMs were not merely chatbot models. When you hook them to an execution harness that can do things, they become agents that can be made to perform arbitrary tasks. And these agents, well they’re a technological panacea, a cure for what ails you, a solution to all the worlds problems, at least if we throw enough compute power at them. This will be as big as—no, bigger than the iPhone or the Cloud. And anybody who doesn’t see this is a fool and a Luddite.

Thus began the current AI boom. LLMs were not to be a fascinating new technology that needed time to find its place and proper use the world. It’s Silicon Valley’s next hypergrowth idea. Turn on the hype machine and let ’er rip. This one is going to be big! It’s going to revolution how we work. It might just end human labor as we know it. It might kill us all lol XD. But it will definitely revolutionize your business if you try to cram it in to every nook and cranny you can and use AI as much as humanly possible.

For regular people, this rapid shift in attitudes around AI wasn’t so pleasant. It meant getting threatened constantly that AI would take their jobs. It meant having AI features shoehorned into the software they use. It meant having AI slop shoved in front of them and their confused loved ones on every platform. And that grocery store coupon app now had a useless chatbot included. This was the future? It kinda sucks. Putting down robots as clankers could easily seem like a harmless way to let off steam.

But even the obviously comical notion that these bots have become sufficiently human-like to warrant their own slur is in itself a backhanded compliment to those who are marketing these technologies. It’s promotion by way of scaremongering, in the same vein as Sam Altman advocating for regulation to prevent harm from AI, or Dario Amodei refusing to release Mythos to the general public since it’s too powerful. Does this sort of conspicuous caution smell a bit fishy? It’s because it’s a ruse, a mind trick borrowed from horror movies. They’re showing you the shadow of the monster and delaying the big reveal, so your imagination can run wild picturing how dangerous it is.

The fear that it evokes is the point. Fear is one of the most important factors driving the hype. Ordinarily, the main fear AI marketers want to induce is the Fear of Missing Out, but really any flavor of anxious suspicion will do. Fear is a very powerful and primitive emotion. Fear of one thing—the robots taking your jobs, China using DeepSeek to destroy Liberal Democracy, autonomous killbots run amok—can easily transfer to fear of related things. And these worries train us to keep AI in our heads by hijacking a simple survival mechanism that has served us well for a long time: track the thing you fear. Observe it, watch out for it, make sure it’s not going to sneak up on you and get you.

But you should not fear AI. I mean, in the off chance that it becomes Skynet, I guess we should all be scared, but I generally regard that as so outlandish that it’s not worth considering. Otherwise, AI itself poses very little threat to you. After all, despite the disappointingly wide adoption of the odious word agentic to describe automated processes connected to LLMs, chatbots have no agency in any meaningful sense. They are not responsible for anything.

No, it is those who set them in motion and have incentivized (or demanded) their use, whether fit-to-task or not, that warrant our fear and loathing. And as much as I appreciate a good anti-clanker anthem, beating up robots is not sending our frustration to the right places. Directing our emotions towards machines and models only distracts from the real problem: an out-of-touch, unelected, and largely unaccountable technology elite are attempting to conjure mythic levels of business Growth out of an impressive but ultimately unproven technology, to the detriment of every other possible goal that matters to humans.

Much of the modern economy has evolved to view Growth as the singular and eternal goal of any profit-seeking enterprise. I will not make the argument here that growth or profit are Bad Things, but that seeking Growth above all else is a malignant and destructive philosophy that affords power to some truly nasty people and hollows out our economic structures from the inside out.

“Growth” in this sense is a nebulous concept, but generally speaking, it means an increase or anticipated increase in any measurable business outcome. The primary way Growth is measured is through company valuation, since that shows a market’s faith in the capacity of an organization to continue to grow. Or at least to convince others that it will continue to grow. Tech stocks have high price-to-earnings ratios precisely because the broader market has become convinced, through years of successes and good returns, that new and better things are just around the corner, and those new things mean new revenue streams.

That’s all well and good, but this kind of success wave has caused businesses to focus on Growth to the near exclusion of anything else that could matter. And when any single value becomes the thing we optimize for, it becomes much more manipulable. It’s why I think it’s quaint when some folks talk about how greedy business folks just want profits. They’re living in 1992. Greedy business folks want you to believe there’s a huge potential to control a majority of a $200 billion total addressable market, and that you need to get in on the ground floor now! It doesn’t matter that an operation is bleeding cash, because that can be fixed later once we have a captive market, right?

Technology, and in particular Silicon Valley, has become the envy of the business world because of its track record of producing these kinds of Growth stories. Business people salivate over it. Every company wanted to be a software company. Many companies in traditional industries declared that they were, in fact, software companies, or that they were “like a startup” because they wanted to associate themselves with the shine of Growth that the tech industry had. Signaling that your business is just like a tech company became a sort of lifehack that executives can use to push Growth narratives without having to make substantial changes to their business.

And It’s this laser-focus on Growth and the use of superficial optics to simulate it that is at the core of the problem. It doesn’t matter if your products are good, innovative, well-made, helpful, financially sensible, socially responsible, or environmentally sound. As long as they can drive the narratives that pump the stock, that’s what counts. This has always been true, to some extent. Stories about companies or people not caring about what results from their actions as long as they make money are as old as money itself. But we’re at a point where the obsession with Growth has become so widespread and so ingrained, that we’re staring down one of the largest asset bubbles in history.

And the people we should be angry at are company executives, board members, and influential investors. I argue that most of these folks comprise an executive–investor caste, a sort of modern, semi-porous business aristocracy that keeps its power by hoarding the rights to allocate capital resources and narrate commercial stories. But I’m getting ahead of myself a bit here.

This is a systems problem, so it’s hard to completely isolate blame, but executives and investors are the ones who make the business strategy and funding decisions. Someone has to be responsible. We can’t let the accountability for this simply dissipate in the ether because it’s complicated. Not every executive or investor is a cynical ghoul or an AI true believer, and all of them are subject to immense pressures to conform to the dominant narratives in their own space. Isolated defiant members of this class can be easily replaced, drawing from a large pool of compliant acolytes clamoring at a chance to join the club. But executive–investors also are far better positioned than the vast majority of us to benefit financially from whatever outcomes arise from these situations, including access to golden parachutes and similar contingencies. So I don’t feel so bad about calling them out.

There’s lots of other groups that I think deserve blame in all this: politicians that enable these industries, business journalists who fail to even apply the most basic skepticism to the claims of companies, tech-bros and techno-optimists that lap this stuff up and blithely assume that anything with the smack of high technology must be progress, and AI researchers and eager-beaver engineers that couldn’t be bothered to take a step back and apply critical thinking or moral reasoning to their field of work. I won’t discuss those in this essay. This is about executives and investors, particularly in technology, and the rotten system they operate in.

Let’s start with tech investors, by which I generally mean venture capitalists, but could also apply to board members of Big Tech firms, and fund managers heavily invested in this ecosystem. This is a group of people who have spent the last 10-15 years watching the line go up on a lot of businesses without really ever understanding why. And when the line doesn’t go up, it fills them, as a class, with the sort of unacknowledged existential dread needed to seed the mass delusions that sustain the AI bubble. It may sound hyperbolic and frankly a bit presumptuous to ascribe this to the psychology of investors, but remember this is a social circle that adulates the kind of person who can proudly claim they don’t engage in introspection.² Someone like this has literally no other meaning in their life than Growth.

Executives generally exist downstream of these philosophies, in that they need to manifest the Growth that investors crave, but like investors, they lack the ability to do anything directly relevant to producing it. So they are locked into a cycle of crafting narratives that investors like using the raw material and powers available to them. At its best, this means navigating a business towards revenue streams, orienting its production to those ends, and providing enough story to fill in the gaps and keep everyone motivated and aligned towards that mission.

If that sounds underwhelming, that’s because it’s just a nice way of saying “pointing at a goal and getting out of the way so folks can do their jobs.” The best executives embrace this. This is a hard job. It requires that you know enough about what is going on in every aspect of your business to explain it, while aiming to please (or at least avoid displeasing) various audiences, always projecting confidence despite your lack of expertise, all while having very little direct control of the results you are accountable for. It is maddening. Indeed, every executive I have spoken candidly to about this has admitted as much.

In the right circumstances, this relatively laissez-faire management method has produced incredible results. But it leaves a lot to chance. What if the goal is wrong? What if you hired bad people? What if the timing isn’t right? Executives, particularly chief executives, are under tremendous pressure to generate Growth, and the vaunted hire-experts-and-let-them-cook method may not be enough. So these folks are constantly incentivized, especially in larger companies and mature industries, to be on the lookout for things that could change the equation in their favor.

Thus it’s no surprise why LLM-based AI became so popular so quickly among this set. LLM chatbots, despite their many flaws, are a genuinely stunning leap forward in technology, and their ability to produce convincing results to a mind-boggling variety of prompts is practically tailor-made to wow the generalist mind of a chief executive. If you’re not buried in the details, everything an LLM produces looks like they managed to stuff an everything-expert into a computer. Clearly this is the technology of the future.

We just need to make sure everyone is using it as much as possible.

Management

The one thing that always felt fishy to me about the AI hype wave was that it was the first tech advancement that felt decidedly top-down. That is so unlike the other waves of tech transformation that have taken over workplaces.

I remember having to beg to use cloud services rather than self-hosting on-prem. I remember bring-your-own-device roll-outs happening as a security response to the reality that employees were bringing their own devices anyways, because company-provided tech was not good enough. Hell, I remember fighting with layers of management to have the rigth to install automated configuration management software, so we didn’t waste hours and hours manually administering thousands of distributed machines. You’d think the benefits of these kinds of things would have been so obvious that management demand we do them, but in virtually every case I can think of before AI, the pitch was always bottom-up.

I spent the about 5 years of my career trying to evangelize the importance of development and release velocity. Simple changes like limiting work-in-progress, having team members swarm on tasks, and devoting a certain percentage of time to continuous improvement are really obvious and proven ways to improve throughput, but higher-ups often hesitated to commit to and uphold those policies because they would mean they might have to systematically say no to influential people. It was better to make the non-decision of just giving one feature task to each person at all times, even if it meant slower work.

The obvious conclusion is that managers and executives, generally speaking, haven’t got a clue how work works. Or if they do, they can’t insist on doing what it takes to improve it without pissing off someone powerful who doesn’t get it and doesn’t care. Sure, sure, #NotAllExecutives are so clueless and callow.³ But the skills required for being a manager or executive are vastly different than the skills required for being an individual contributor. I can be cynical about what these skills are (and I will, just you wait!) but I want to be clear that I’m not saying all or even most managers or executives are incompetent fools. I am saying that they often don’t know what they are doing, because businesses are extremely complex dynamic systems with sophisticated emergent behavior themselves embedded in a complex dynamic system with sophisticated emergent behavior. Nobody knows what they are doing, beyond some basic-level instinct following, and a broad idea how to direct some others towards some common goals. It’s a miracle of human sociology that businesses work at all.

I also want to be clear that about what I mean by “work” in this sense. I mean the delivery of valuable products and services to customers on a regular basis. I mean the routine functioning of an organization and how it’s able to direct activity and pay its employees and keep the lights on an keep enough people invested enough to do the actions necessarily to maintain it as a going concern in a fundamental sense. I mean the completing the tasks required to please customers and manage vendors such that they continue to purchase and provide and keep the streams of cash and goods and services flowing. This is line-level work, the meat of any business. The stuff needed to keep these operations ongoing is management.

I think most people can intuitively develop some basic management competence, although there are definitely people who are demonstrably bad at managing stuff, and a few that even have some natural talent. The ones who are the best at it seem to set a few key principles, aim the business in the right direction, and let people get to work. And companies that end up growing successfully do so through a combination of a small set of decent practices, a few critical big breaks, and relentless opportunism. It’s what Nassim Taleb might call “convex tinkering”. There are actual business strategies that aim to exploit this mode of operation, notably “Lean” methodologies (when properly applied). And the whole start-up concept—at least original version seen mostly before 2012 or so—was built around the idea of making a relatively large number small bets in hopes of finding a few big winners.⁴

Within the business world, we’re not usually permitted to talk about luck like this in the open. The success of a venture must be narratized; that is, a story must be told that identifies the success as the Effect of a well-known Cause. This obligation even gets embedded into buzzword lingo like the onerous “right-to-win”, which exists solely to lay a stronger claim to causal relationships between choices and outcomes than can be supported by real-world information. You tried something and it worked, so you want to try it again? You didn’t uncover a “right-to-win”; you experienced operant conditioning, and just like my 1 year-old, you learned to repeat the actions that led to a result you like.

Talking in terms of luck is deeply uncomfortable for most folks. Humans have a deep psychological need to believe there’s a reason things work out the way they do, especially when we have something we value (pride, reputation, money) on the line. So we demand that these murky, randomness-infused situations be cast into narratives that we can use to regain the psychological safety that comes from believing we are part of a story with good guys and bad guys and an overarching plot with causal through-lines. Investors, board members, and other external stakeholders demand that someone be responsible for the actions of a company and their results. And it’s hard to blame them.

Thus business executives find themselves in a weird place where they are responsible for the outcomes of a business, but have very little direct control over or even awareness of the details relevant to success. Operating at a high level usually necessitates operating at a low resolution. There’s not much you can do about this. Even if you spot-check the ground-level from time to time—hello, fellow laborers!—you won’t get the full picture you need to make sense of what is going on. There’s too much detail, and too many reasons you either can’t or won’t see it. In many cases, folks lower on the ladder are incentivized to obscure these things from you. Nikhil Suresh describes it better than I can:

To make matters worse, reality that is accessible is usually not accessible from a high vantage point. From a bird’s eye view, you have no way of knowing that 80% of a specific team’s output is from Sarah, and Sarah’s son just broke his arm playing soccer so that project is about to collapse as she scrambles to cope. This is totally visible to some people at the business, but is not going to be shared with the person making promises to the board. We could build a complex systems-thinking approach about our work, but that is very hard and will have obvious fuzziness.

The tools and models that higher-ups work with tend to assume a lot more standardization on the ground than actually exists. I’ve always found it funny that, at most companies I’ve worked at, maximum headcount is frequently set at the board level. I assume this is only because it’s easy to measure, not because it’s an item that can be easily correlated with the output of a company. It’s not even in direct proportion with budget, since the pay for an employee can vary so much from role to role.

But the language of executive management always operates at this hyperreal level where the managed values that are treated as Very Important lack a clear relationship to the much messier reality of the day-to-day. Person-hours, story points, project estimates, status colors: these are all reductions of information meant to make the world of the business’ operations palatable to those who direct it. And using this information, it is assumed, management can push buttons and shuffle resources around to address issues as the business context and priorities change and new information comes to light.

It’s understandable why you’d want this model to work. Interchangeability of parts has been an enormous boon to industrial productivity practically everywhere it’s been applied. The challenge has always been figuring when and how to apply it. Clever business leaders will conduct lean experiments in new areas to see what gains can be unlocked, but they won’t just assume you can talk about creative work like it’s assembly line produciton. But too often, it’s not clever business leaders that dominate, and the organizational culture will demand to discuss all projects in wildly oversimplified terms: Taylor-esque reductions of industrial complexity that were barely passable for factory work and deeply unsuitable for modern knowledge work.

If asked directly, few sane managers or executives would admit to thinking about their employees like this. But their actions belie their true beliefs. If you think you can move folks across teams frequently, you’re treating them as impersonal tools. If you assume you can replace workers with AI or underpaid overseas labor, you’re thinking of them as cogs. The organizational dynamics demand human interchangeability, and it is up to management to paper over the cognitive dissonance between that and the human reality of labor, usually with trite jargon and contorted metaphors.

The fact that they bother to resolve the dissonance at all is evidence that managers and executives intuitively understand that skilled workers cannot be reduced to fungible resources. What a drag that must be! It’s one thing that you have to pay people for their labor, but you have to take on all of the murky illegibility of dealing with real humans. Real humans have values and goals and preferences that have nothing to do with the primary company goal of ever-increasing valuations. This can mean giving workers competitive compensation, good working conditions, and a decent amount of freedom to pursue and balance a variety of outcomes according to their own judgement.

If this sounds like the baseline expectation for being a good employer, congratulations! You’re not a stonk-pilled moron. If you don’t regard Growth as being the single goal of an organization, it’s blindingly obvious. But the incentives in the modern economy tilt against this. While an especially good manager may regard this as part of the challenge, and work to balance these pressures, it is still a distraction from reliably delivering the outcomes your superiors or investors will demand. And the higher up you go, the more abstracted from reality those goals will be.

Put simply, the more distant the stakeholder, the less they care about anything other than number go up. Executives are at a point in the control system where they have little incentive to care about anything other than pumping valuations, but they also have little direct control over the kinds of outcomes that can reliable show investors what they want to see to drive those numbers. So often they are just left to find whatever levers they can to induce Growth and pull on them repeatedly until they stop working. Everything else is a side effect.

It’s not hard to see how these incentives arise. After all, we all want the same thing when it comes to retirement savings. There we’re even further abstracted from the day-to-day, and usually diversified in such a way that no single individual outcome can dominate. In contrast, a chief executive is tasked with the continual appreciation of the market value of the company they administer. Failure to achieve that usually means dismissal, which is a setback even if they have a cushy landing pad. But they do get the big bucks, the cushy landing pad, the benefit-of-the-doubt from the media, easy access to a social network of other power people, and, crucially, the power to reorganize or terminate employment conditions for those that earn their living under their organization. Forgive me if I don’t feel terrible about assigning them a good chunk of the blame for this state of affairs.

Still, given their situation, you can see why executives would be eager to find what control surfaces they can to manifest bigger, better, and faster outputs without the messy uncertainty of human behavior getting in the way. The levers executives traditionally have don’t always induce unambiguous effects on company performance. Sure, you can set goals and establish high-level metrics by which whole departments will be measured. But you have to delegate the responsibility of hitting those numbers to underlings, who in turn will set more goals and further delegate. You don’t really know what will work or fail, nor do you know who gets the credit or blame. And even if that still feels like leaving a lot to chance, it has been the state of affairs for executives since skilled knowledge work has become the dominant form of labor in the modern organization.

Narrative

So if you’re a manager or executive, what kind of general strategies could you apply to succeed in a world like this? One obvious one is to get really good at corporate storytelling. You can become able to quickly and convincingly explain successes and failures as being part of some wider narrative. Folks who can do this can weather a much broader range outcomes happening under their tenure, because they can weave any number of things into a tale that represents the world the way they want you to see it.

Doing this well requires a deep sensitivity to business tropes, and the ever-shifting dynamics around them. Lacking this sensibility, incompetent managers will clumsily deploy dated buzzwords and incongruous expositions in an artless attempt to signal membership in the echelons of upper management. These folks are aware that the storytelling game is being played, and that the use of certain words is part of how the players recognize one another, but they can’t make it look natural, and so they are marked as pretenders, both to run-of-the-mill employees and higher-ups with greater mastery of these skills. In fact, I’d argue that that’s the purpose of cringey buzzwords: as a shibboleth to quickly identify members of an elite tribe.⁵ But I digress.

Storytelling is the skills that washes over a thousand uncertainties. Everyone past a certain level in business knows you can’t guarantee certain outcomes, but you can probably engineer and few, and if you build your story around that, it can be hard to argue with. Have you ever been on a project where the executive sponsor insisted on a particular outcome? They were building their story around that outcome. They wanted to be able to tell the board that they had a achieved X and it meant Y. So you can bet the pressure was on to deliver the outcome, and the thumb was on the scale to ensure it was achieved. Reality was bent to ensure it happened.

I can’t name the number of projects I’ve been on that I thought were total failures. Death marches. Assured impossibilities. And I’ve struggled and fought up to the end and come up short. I’d come in the next day expecting to be told off, only to learn that the project had been announced and was a smashing success! You see, the quality of deliverables was only relevant insofar as it was necessary for all the most important people to agree that Thing Done. Usually all parties and counterparties to a project have a vested interest in making sure that Thing Done. Thus, they need only declare Thing Done in a bold enough way to make it likely no one will look close enough to notice the the wheels are falling off.

Understanding this explains so much about the absurdity of the modern workplace that it feels like forbidden knowledge. Like it should be some kinda conspiracy. But it’s not. It requires little coordination other than the tacit realization that two or more groups of people want the same thing, and that nobody important is going to be checking the detailed results. The sponsor at the other company wants the deal to go through. Your boss wants to hear that the project is done. Investors want to hear that the total addressable market is in the billions and growth is inevitable. Of course people might be skeptical that you will underdeliver, but when dealing with prediction and uncertainty it’s the narrative that does the convincing.

This is the skill that must be developed: how to tell your audience stories they want to hear, in almost any situation. Notice I say “stories they want to hear” not “what they want to hear.” Lesser practitioners of this game try make every bit of narrative palatable and satisfying, which rings hollow for all but the most credulous listeners. We know the world simply isn’t like this. Simply cherry-picking facts and putting positive spin on things can only get you so far. More seasoned narrative experts learn that good stories need negative emotions to heighten the stakes. They need tension and resolution. Someone needs to be in danger or distress, only to be rescued by the hero at the last minute. Stories like this feel more real, if only because the emotional investment makes them feel more lived-in.

There are limits to this, of course. Even an expert storyteller can have their bluff called if the reality isn’t behind it and the audience tires of suspending disbelief. And you can’t fool all the people all the time. Storytellers must know their audiences, and not all audiences will be amused. And most of the time you can’t target an audience broad enough to induce a multi-trillion dollar hype-wave, because an audience that broad is bound to draw enough enough skeptical, antagonistic voices and bring the whole narrative down on your head.

But what if you could? What if you created a machine that hijacks the human gestalt reasoning pathways? What if it did a good-enough facsimile of intelligence that people’s tendencies to fill in the gaps, extrapolate, and anthropomophize were fully activated? And what if you told every executive that it was a magic way to increase productivity across the board, and it might result in massive unemployment because it will replace a staggering fraction of the workforce? What if powerful people, abstracted from day-to-day realities of work but always under pressure to show growth and primed to follow business narratives, were convinced en masse that a single new technology was the Future?

You’ve got the recipe for a historic hype bubble. And this hype bubble a given a lot of powerful people a lot of narrative cover to devalue labor across and incredibly wide swath of the economy. It’s easy (and fun) to assume that executives like to engage in machiavellian gambits to get what they want, and while there are certainly 4D chess players out there, it really seems a lot of them will just try shit and see what works. No, really. It’s a way to expose themselves to the upside of chance that most of us proles are usually too scared to try. As long as there’s a story that can help them shed blame and criticism when things go awry, or a way to back out of a decision gracefully, they’ll be fine. And if it goes well? Then you claim the credit. There’s plenty of relevant examples.

Are you annoyed at paying support staff? Try what usurious predator pay-as-you-go provider Klarna did. Eliminate support as a concept, because the chatbots will do it. Oh that didn’t work? It’s okay, you can bring them back as contract workers, so you can skirt employment laws, avoid unionization risk, and crack the digital whip more effectively through gamified, asymmetric gig app interactions, all while claiming it’s better for them because they are their own boss.

In fact, just like any change that purports to increase productivity, AI provides great cover for shedding labor cost. Did you overhire during the ZIRP-y haze of the Covid pandemic? Never fear! You can claim that, as an AI-native organization, you have found new efficiencies that make a significant proportion of your staff redundant, allowing you to reframe your moves as a innovative rather than an error correction. And don’t worry, the increased pressure on your remaining employees will help push them to pick up the slack left by unfilled roles. Or at least they will suffer in silence for fear of being “randomly” selected in the next round of layoffs.

What if you already have some already-marginalized gig workers that need another good kick? Like, say, copy translators. With more products produced for global markets than at any point in history, you could be forgiven for thinking the modern world has been a boon for translators. You would be wrong. The existence of machine-translation post-editing methods, the constant pressure to reduce costs, and the extent to which consumers are inured to bad translations mean you don’t have to pay these folks all that much money. But even though LLM translations are not significantly better than earlier iterations of machine translations, the narrative surrounding them has given almost everyone permission to use them and claim success because AI, so you can get translators to take post-LLM-editing contracts at 1/4 the rate of genuine translation, even if polishing the LLM output takes as much time as translation from scratch.

It’s almost immaterial whether LLM chatbot technology is actually providing any of the benefits. It is the widespread perception of AI as transformative and inevitable that does most of the heavy lifting here. Individual workers (or managers or executives) may well doubt the general or particular claims about AI’s efficacy, but they cannot risk challenging such a pervasive narrative without it being perceived as an attack on an organization that is presumably supposed to be “all in” on AI.

Power

Very few executives even have the power to change any of this. I’d love for this essay to end with a call to anyone with power who is willing to start openly decrying this bubble for what it is, but I’m not that naive. Narratives may be the source of executive power but they can’t just bend society-wide beliefs to their will. What I am calling out is the use of these trends by narcissists to cast off even the pretense that something matters to them but never-ending asset Growth.

Ironically, the dim, reductive way in which these jerks view workers as fungible resources is where the AI narrative falls apart. The more you think of workers as resources—black-box machines that transform inputs into outputs—the easier it is to be an AI True Believer, because you have no idea of what details matter and why.

Fortunately for us, the details of most jobs do actually matter⁶, and an LLM that operates in a perpetual state of summarizing and probabilistic pathfinding cannot fully replace the work of thoughtful humans with skin in the game and a drive to create and adapt. I state this confidently because this is what we observe when we look at the current, actual state of LLM technology. There are too many examples of companies rehiring laid-off workers to ignore. And despite the numerous headlines about Big Tech layoffs, the total headcount at the largest companies has remained relatively stable. If Generative AI were as revolutionary as its proponents claim, we’d be seeing a far more stark effect on these figures.

Thus, barring a genuinely novel development arising in how these models work, the net societal effect of LLMs is not (and probably will never be) a mass replacement of human labor with AI agents. But the narrative that this threat is real and coming has further undermined the bargaining power of workers to earn their share of rewards and advocate for better conditions for themselves. Memes and jokes that accept this inevitability as a threat only serve to support this power shift.

The clankers are not our enemies. They are mathematical constructs and tools that do not merit that level of anthropomorphization. I get the joke: we’re not denigrating an underclass of androids, we’re lashing out at the tools of a powerful elite. But this enemy is happy to wield both technophobia and technophilia as weapons. Whether the underlying technology will actually bring about the prophesied changes is irrelevant. Whether you welcome or fear the technology is beside the point. What matters is the perception of AI inevitability. That is what allows these folks to further concentrate wealth and power among themselves at the expense of those who actually get things done.

Once classic moral defense of capitalism has been that we can tolerate a few assholes getting mega rich if methods they use to obtain that wealth lead to a net betterment of conditions for all of society. A rising tide raises all ships, or so the argument goes. But that isn’t how this story is playing out. The average person, while facing ever-increasing costs-of-living, is being asked to do more work for the same money, and is held captive in that job by often precarious financial situations, and the likelihood that the next job would be little better.

Yes, there are “regular” people getting rich from this trend—a full three-quarters of Nvidia staff are millionaires—but these are paper gains, and it’s not always clear how frequently or effectively these folks can cash out, since these figures don’t break down vested versus restricted equity units. This is pretty much limited to the tech industry though, and even then the cracks in the old system are showing. Big Tech may not be permanently shedding much headcount with their layoffs, but they have been able to shed some of the RSU obligations they once had, if only to funnel the cash from the stock buybacks those necessitated to fund the insane CapEx of the AI datacenter buildout. It turns out these seemingly obvious paths to wealth are in fact very manipulable.

Even the Silicon Valley startups themselves, which at least afford a lottery ticket to a privileged few, are starting to manipulate the state of affairs to ensure the value only goes to a tiny set of individuals. Take Windsurf for example, who pulled a fast one on their employees by selling Google a perpetual license for $2.4 billion, paying out investors and a core set of employees, who would join Google. They then sold the rest of the company, including the rest of its employees, to Cogntion at a (then undisclosed) fire-sale rate of $250 million. Although the deal was structured to provide “full equity vesting,” having a valuation fall from a peak of $1.25 B undoubtedly hurt any holders of options granted after that high-water mark. Three weeks later, as a delightful fuck you to the former Windsurf employees, Cognition CEO Scott Wu offered a choice: work like a dog or accept a buyout, saying:

“We don’t believe in work-life balance—building the future of software engineering is a mission we all care so deeply about that we couldn’t possibly separate the two”

Of course it’s the mission that they care about. We’re building the Future here, so don’t expect to see your children if you stick around. Anyone familiar with modern Silicon Valley will understand the subtext here: get to work if you don’t want to be part of the permanent underclass. The mission is to disenfranchise workers everywhere with machines that may or may not do half as good as job as actual people, but who can at least further the agenda of business elites by scaring the rest of us into shutting up and grinding it out.

For those outside of tech, the pressure has ratcheted up as well, with the same layoff threats and demanded productivity increases have been made across every industry. Sure, there are plenty of folks who seem really excited to use them. A shocking number of small business owners—ever on the lookout for ways to improve their undoubtably thin margins—seem to love AI flyer generation, even if it undermines consumer trust in their brands⁷. But most people seem more skeptical that it can really do the work they have, you know, spent years developing the skills to do. Of course it’s impressive that the machines can now do a reasonable facsimile of many jobs, even one that may be good enough for mockups, proofs-of-concept, and idea generation. But if you’re close to the work, you know it’s not really good enough for your core duties. Despite this, even skeptical people end up using chatbots to get their work done, not because they find them genuinely helpful, but to keep up with the expectation for increased output that has come from AI mania. And the quality inevitably suffers.

You’d think that noticeable drop-offs in quality would raise more eyebrows, but businesses are seldom instrumented well-enough to notice that. When all KPIs are focused on AI adoption, it’s easy to ignore the negative effects for a while. What gets measured, gets managed after all. And the pressure is on to make sure we are all Doing AI, so the stats will be juked to show that we are very much Doing AI and it is Working.

Of course, stopping this will take at least the bubble popping, and it’s not clear how or when that will happen. The quasi-religious belief in the transformative power of Generative AI spreads like a contagion through the business world, and once it sets in it is very hard to root out. Some of it will surely outlive the heights of the bubble, just as there’s still crypto and VR holdouts today. Too many people have publicly committed to being “all in on AI” to easily back out now without taking a very conspicuous L, or without undermining vendor and customers who have also voiced their belief in a glorious LLM-powered future. I don’t really recommend fighting this at work more than you strictly have to. It’s probably not going to have any real effect.

Outside of business, though, we stand a chance to change minds. These narratives must be actively questioned and deconstructed at every turn, and at every level in society. Always check the scale, and whose thumbs are on it. More often than not, you’ll find them attached to folks that are heavily invested in this charade, hoping the infernal machine stays running running long enough to cash out. Treating new technology with healthy skepticism isn’t going to make us lose anything, and the people who say otherwise cannot be trusted because they cannot be disentangled from the massive web of vested interests that refuses to let this bubble truly pop.

No clanker ever called me a resource, but plenty of executives did, and without even knowing or caring who I was. Our dehumanization comes from other humans, ones that have been inducted into a cult that only values Growth, a wobbly notion of potential to capitalize on resources, as signaled by asset valuations. Petty concerns like reality, humanity, and genuine progress are ancillary considerations that only become important when they are necessary to fuel narratives or fund exit liquidity for insiders.

I suspect LLM technology is here to stay in some form, and that’s not a bad thing. We won’t really know where the sweet spot for their application is until the religious fervor around AI use dies down enough for businesses to take a critical eye to their costs, uses, and effects. Use in coding environments seems likely. But the promotion of these tools as a general-purpose solution to every conceivable human problem is stupid, wasteful, and short-sighted.

The robots are not taking our jobs, at least not very effectively. But the idea that they might has forced us all to accept worse conditions and keep our mouths shut out of fear that we might lose our livelihoods. As much as we can, we should all resist these silly notions whenever possible. I know you can’t always openly voice these opinions at work. But at home, at church, at the barbershop, at the community center, at the kids’ sports games… you can call out the bullshit for what it is. You can head off arguments about how society will deal with everyone being unemployed by questioning whether that’s actually happening. You can ask for actual examples of real software products that people pay money for that were vibe-coded by non-technical people. You can ask whether it’s really a good thing to use a machine to write your emails for you. You can ask if there is any genuine investment for return on AI projects.

And you can give the chatbots and clankers a break. It’s the executive–investor wankers that deserve all the hate.