The Parlor Trick That's Burning Down the House

What we call AI is a subset of a subset of AI that’s really just a parlor trick—complex maths cleverly wrapped up in a nice package, trained on enormous amounts of stolen data, that gives the illusion of intelligence. Technically, an LLM is a specific architectural application of deep learning that uses the transformer’s attention mechanism to perform stat inference via language tokens, scaled to a level where emergent properties—such as reasoning and in-context learning—begin to manifest. The thing humming in your pocket, autocompleting your emails, summarizing meetings nobody wanted to attend in the first place—that’s not intelligence. It’s not even close. It’s an LLM: a large language model. A stochastic parrot in a tuxedo. A next-token predictor that has read the entire internet and can tell you, with a straight probabilistic face, that the word following “the mitochondria is the powerhouse of the” is “cell.” That’s not reasoning. That’s a thermodynamically expensive autocomplete, and we’ve built entire data centers to worship it. It’s ELIZA all fucking over again.

The loudest people in the room promoting this technology aren’t the experts. They’re not mathematicians. They’re not engineers. They’re not the PhDs who spent decades studying the theoretical underpinnings of machine learning, carefully caveating their claims with confidence intervals and limitation sections. Sure, there are a few randomly interspersed within the ranks of the executive class—token brains wheeled out at investor days to lend a patina of credibility—but largely it’s tech bros and oligarchs. The same people who brought you the metaverse, the crypto crash, the SPAC bubble, NFTs, and the conviction that nap pods in an open-plan office are a substitute for healthcare. They don’t understand the technology in any meaningful sense, but they understand its utility as a narrative device, a leverage point, a cudgel. The AI story isn’t about building things. It’s about extracting things. It’s about justifying the elimination of headcount, the crushing of labor, the consolidation of power into ever-fewer hands, all while sounding like you’re riding the rocket ship of history rather than throwing gasoline on a bonfire of human capital.

Whenever anyone questions the macro-level narrative about AI—any question, not even a cynical one, just any question—it’s treated like contempt or hostility. How much are we spending on AI? What is our ROI? Are we sure this is a better investment than conferences? Have we done due diligence on how our data is being used by the vendor? You ask that in a meeting and the silence that follows is the particular silence of a room full of people who have staked their entire career on not asking that question. The temperature drops. Someone from the AI strategy team—invariably a person whose job title contains the word “evangelist,” a term that should send a cold shiver down the spine of anyone who’s read a history book—shifts in their chair and says, “We’re really excited about the potential,” which is not an answer. It’s a coping mechanism. Funny story, I grew up in an evangelical cult so I know a lot about this space and it operates.

And to be clear, I do have open contempt for the religiously devoted usage of these tools, and I do have much more striking questions. I can’t even get my foot in the door of asking a single one. Questions like: how long until you replace us all with this tool that doesn’t ask questions and responds to everything like a salesman saying “you look like a really intelligent guy”? How long until the summarization model that can’t reliably distinguish between a critical bug and a feature request is making staffing decisions? How long until the hallucination becomes the policy, and the policy becomes the layoff, and the layoff is executed by an algorithm that’s never once been held accountable for anything in its existence because it’s not a person, it’s a math problem wearing a chat interface?

Given the fact that the United States has the absolute worst track record of labor representation in the industrialized world, and our mentally addled president is too busy building a ballroom—erecting an arch to stand in front of the “eyesore” that is our heroically fallen veterans’ headstones—and handing out no-bid contracts to his friends to get paid more than the budget of the Apollo missions to fix a few fountains and paint over the problems of the reflecting pool, leaving all the real plumbing problems wholly unaddressed: well, you’ll have to pardon my total lack of Pollyannaish belief. Meanwhile, in fantasy land, he’s golfed 25% of the time he’s been in office, watched the Knicks lose at Madison Square Garden, and spent millions of taxpayer dollars to destroy the White House, wreck the Capitol building to watch a homoerotic display of oiled-up men slip and slide all over each other for his birthday, selling $1 million VIP-per-person tickets, millions in broadcast deals, millions in tickets, and all manner of other privatized profits—but I’m sure the cleanup of this shit show will be footed by the taxpayer. He’s been patting himself on the back so hard it’s a wonder he hasn’t knocked the wind out of himself, screaming on Truth Social about bringing back manufacturing jobs that don’t exist, won’t exist, and even if they come back will just be automated. Meanwhile, real jobs—tech jobs—are being sent abroad in the tens of thousands. Trucking jobs are increasingly getting automated. AI spending is causing huge cuts to U.S. labor. And no policy has been drafted to do shit about any of the real problems being faced by any of these challenges. To name just a few of the challenges facing labor and there are more than I can possibly enumerate here.

All the while, the tech oligarchs are invited to fancy dress parties and $50,000-a-plate crypto launches. You’ll have to pardon my total lack of belief that anyone in any branch of government in any party, any member of any executive board, or anyone on the capital investment side of AI has anything but contempt for labor. They can’t wait to see it crush and destroy the livelihood of every single white-collar job in America while hoping, begging, and pleading that it lets them get 1% richer. Because the destruction of millions of lives, at any and all human toll, is worth that 1% increase to these deranged lunatics. That’s not hyperbole. That’s the revealed preference of the executive class, displayed in quarterly earnings calls and stock buybacks and the quiet, systematic elimination of any job that can be outsourced, automated, or reclassified as a gig.

These tools were foisted on me and my team without any meaningful support. There was an expectation that, via magic, our output would exponentially increase. None of the tools were well understood. We were expected to build them into our software and magically understand them well enough to do so without any training, resources, investment, or support. Deadlines for all of this were set—ambitious, arbitrary, untethered from anything resembling engineering reality—and the team doing all of this wasn’t even staffed. It would be one week before the launch of our first AI-powered app that the team would have all its members. We had one quarter to build an app using technology that none of us, and no one at the time, had ever used. One quarter. To build something production-grade with a technology that hallucinates, degrades under load, and whose behavior can change without warning when the vendor pushes a model update.

The pressure was enormous. Team members left the organization because of it, and I applaud their willingness to defend their boundaries. They understood something that the invisible man—the same invisible man from the boardroom, the same shadow that doesn’t know your name but knows your headcount cost—refuses to understand: that their finite, mortal, irreplaceable lives are not fuel for a hype cycle. They walked. Some of them are thriving now at companies that still build real things. Some of them are still recovering. All of them were right.

But one thing remains a mystery to me. The experts are mysteriously absent from the room. They’re there when it comes to building out tooling, explaining complex topics, discussing fascinating machine learning classification technology that’s hot off the press. The real experts—the mathematicians, the engineers, the people who actually understand the loss curves and the attention mechanisms and the catastrophic forgetting problem—are incredibly generous with their knowledge when you approach them with genuine curiosity. But when it comes to strident claims about labor replacement, about AGI arriving in eighteen months, about AI being the most important invention since fire, the real experts aren’t weighing in. They’re a lot more cautious. A lot more measured. A lot less reckless with their rhetoric.

It’s almost like knowing things gives you two different perspectives that idiots don’t have. First, it gives you the intelligence to see the consequence of what you’re saying before you speak it. You understand the cascade effects. You understand that when you tell a room full of CEOs that their customer support staff will be obsolete in two years, those CEOs will start firing people now—not in two years, not when the technology is ready, but immediately, because the narrative is the permission structure, and the layoffs will juice the stock price whether the technology works or not. Second, expertise enables you to understand that even an infinitesimally small chance that you’re wrong is still a career-ending thing to be wrong about. The real expert knows the limits of the model. The real expert has seen the edge cases. The real expert understands that the gap between a demo and a production system is not a gap of optimization but a gap of fundamental capability, and that gap may not close on the timeline the investors want—if it closes at all. So the real expert stays quiet, or speaks in careful probabilities, and the vacuum is filled by the loudest, least qualified voices in the room.

The executive class doesn’t have this liability. They can spout whatever horse shit they want, free from any accountability, because they’re America’s darlings. They’re exempt from the normal laws of consequence because “sometimes visionaries are just too genius for the laws of physics,” or whatever bullshit Elon Musk’s PR person tells you when he announces that self-driving Teslas are “just another year away” for—what?—the twentieth year in a row now? For anyone else, that’s stock manipulation. Lying. Misleading investors. Demonstration of incompetence. Failure of understanding the underlying technology. Failure to learn from one’s mistakes—like twenty times. Pathological stupidity. Psychosis that should be disqualifying from holding the position of washing vegetables in an Outback Steakhouse kitchen. But for some reason, we’ve hegemonized this class of roaringly unintelligent, uninspired, moody, pathetic, thin-skinned, wholly unqualified, big-mouthed nepo-babies, elevated them to a status immune from any and all accountability.

Fucking why, though? Shouldn’t accountability be amplified as you ascend? When I got my captain’s license, the bitch of having that credential is that any collision on the water means I am partially at fault, because there is always something I could have done better to avoid the accident. That’s the standard. That’s what responsibility means. As the tech lead of my team, when we make a decision that results in a technical failure, a security incident, or an outage, I am first in line for accountability. We say it’s a blameless culture, but that’s theater. I am accountable, and I should be. I take very seriously my duty to represent my team well, to own the failures publicly and praise the successes collectively. That’s the barest minimum of my job’s duties.

But the executive class? Drive the company’s financials over a fucking cliff and have to lay off 10% of your staff? That’s just business nowadays. When did we decide that this isn’t a catastrophic failure of budgeting, strategy, direction, and every single thing that the executives are literally responsible for? What the fuck are they even for, if not to prevent these sorts of catastrophic events from destroying the business, or destroying the most valuable asset the business has—its people? The answer, I think, is that we either never decided or we just stopped asking. The accountability gap widened so gradually, over so many decades, that nobody noticed the moment it became a canyon. And now the people who crash companies are rewarded with golden parachutes, while the people who built those companies are given severance packages that look like a typo.

As the hype train continues to be screamed down from on high from this same class of unaccountable people—the same ones who will certainly use AI to justify trimming headcount because of unforeseen budget overruns—they will be cutting the experts. And the experts were here all along. Just like the experts were there when the Space Shuttle Challenger took off, warning about the O-rings. Just like the experts were there preventing the screwworm from traveling north into the United States, a quiet, unglamorous miracle of public health that saved billions in agricultural damage, until Elon Musk’s DOGE deemed them… which was it? Waste, fraud, or abuse? Could someone interview that fuckwit’s gaggle of clueless, naive, lazy, arrogant little pricks to find which category the NWS Monitoring program fell into? Maybe since Musk is such a genius he’ll have a solution for the problem. Or, since he’s responsible, he can pay for the remediation, or reimburse the tax payer for the increases in beef prices? Oh wait, there’s no accountability for this class of idiots. I forgot. Similarly, the experts were there warning people to vaccinate and look where measles cases are—if you can even get an accurate picture because we’ve cut the measuring programs needed to properly understand and contain the spread. Bird flu? Back on the menu, boys! Ebola? We’ll be lucky to dodge that bullet. Lyme disease? NY has some interesting cases of a new variant popping up this year that are severe enough to beg the question: should we rename what we’re seeing a new disease because it’s so much more severe? The experts are always there, standing just off to the side, speaking in careful, measured tones, getting drowned out by the roar of the people who mistake confidence for competence and volume for validity.

Hold on while I take a sip of raw milk.

It always takes something catastrophic for people to realize they should have listened. To realize that the people who didn’t learn what they learned to make a buck—they learned what they learned because they were curious, because it fascinated them, because they were nerds—were the ones who actually knew what they were talking about. They became experts out of their passionate conviction to give something back, not to extract something. And somehow, we’ve villainized the curious and hegemonized the exploiters. We’ve built a system where the people who understand the technology are silenced by the people who stand to profit from its weaponization. We’ve allowed the invisible man to replace expertise with trendslop, accountability with impunity, and curiosity with the cold, dead logic of cost reduction.

The parlor trick is impressive. I’m not denying that. The maths are elegant. The scale is staggering. But the parlor trick is being sold as a god, and the people selling it aren’t the ones who built it. They’re the ones who saw it and thought: how can I use this to fire people? How can I use this to avoid paying for labor? How can I use this to remove the last remaining friction between me and total extraction?

The experts are right outside the door. They’ve been there the whole time. They’re not hard to find—they’re on arXiv, on research forums, in the quiet corners of the internet where curiosity still lives. But we don’t listen to them. Probably because they don’t take a hammer to their jawline, say racist shit, sell supplements, preach prosperity gospel, or whatever the algorithm’s incentive structure and profit motive runs on. Instead, they speak in probabilities and caveats and “further research is needed,” and that’s not sexy. That doesn’t move the stock price. That doesn’t justify a $500 billion capital expenditure on data centers. So we listen to the oligarchs instead, and we let them burn the house down with a parlor trick. Or, destroy the living conditions of entire neighborhoods with a data center approved by 6 council persons who sold out their constituents for a signed photo of Musk.

The fire is spreading. The experts are still talking. Maybe—just maybe—we could try listening before the roof collapses on all of us.