“I know what it’s like to lose. To feel so desperately that you’re right, yet to fail nonetheless. It’s frightening. Turns the legs to jelly. I ask you, to what end? Dread it. Run from it. Destiny arrives all the same. And now it’s here. Or should I say… I am.”

— Thanos

I’ve been turning that quote over in my mind a bit recently. I think it describes the process I’m going through with AI: the process of making peace with what seems to be inevitable. Over the last few years I’ve used these tools heavily, though not enthusiastically. Revealed preference says that if you use something, you must want it. But the tidiness of the theory falls apart on contact with reality. The real world is messier. A lot of us adopted AI because we felt that we had no other choice. Two things can be true at once: you can choose a tool of your own free will (even get genuinely good at it) and still have your doubts and reservations. I think I’ve finally made peace with that.

The Phantom Menace

Making peace implies there was a fight, and perhaps there was: for a while, what I felt was closer to resentment. So I think the fair thing to do is lay out the biggest issues I have with AI before going into how my thinking has evolved over time. This way, as a reader, you can decide whether the rest deserves your consideration. For what it’s worth, my objections are more socioeconomic than technological.

There are a lot of reasons to dislike AI, of course (not least of which is that the people making it keep telling us there’s a non-trivial chance it could cause the extinction of our entire species). But even if it doesn’t send us the way of the dinosaurs like some giant asteroid made out of linear algebra and probabilities, there are plenty of valid reasons to have, shall we say, mixed feelings about this particular technological revolution.

My number one issue with the current iteration of the AI boom is how it is warping the economy. The biggest way is by crowding out other borrowers. The Wall Street Journal, not exactly a hotbed of AI skepticism, calls it the biggest economic bet in U.S. history. And that’s before you count the $1.65 trillion in leases and purchase commitments that five tech giants have built up off their balance sheets in four years. For the first time, the country is spending more on the chips and servers that fill data centers than it spends on housing.

Often you’ll hear that the AI boom is the only thing propping up US economic growth. But that’s a misreading of the situation. What the boom is doing is competing with everyone else for money and materials. Steve Eisman puts this year’s new AI debt at roughly $500 billion, issued into the same market where the Treasury borrows, and argues the 10-year yield would be much lower without it. That reaches the rest of the country as more expensive mortgages and business loans. Goldman’s economists think the effect on borrowing costs is small so far, but they also find data centers pulling construction resources away from other projects.

10% of households own roughly 90% of the stock market, and only a sliver of Americans work for or invest directly in the companies building AI. Everyone else is on the other side of the ledger, worried about keeping their job.

AI is the leading reason companies have given for layoffs this year, cited in roughly one of every five cuts. Unemployment is 4.2%, which sounds fine, and it’s the number markets watch. But look underneath it: long-term unemployment is up 55% since 2023, and the Richmond Fed notes the way it’s rising is a pattern it has only ever seen in recessions. Recent college graduates are being welcomed into the workforce with a hiring market that is the coldest it’s been for new grads since the depths of the Great Recession. I can’t tell you how much of this is AI and how much is high interest rates or the hangover from pandemic over-hiring, and some companies are surely citing AI as cover. But AI is the reason they are choosing to give. And the consumer spending holding it all up is coming from the richest fifth of households, who account for 57% of it. The boom is disproportionately enriching a small group of already wealthy people at a moment when inflation is outpacing wages and labor’s share of national income has hit a new low.

Another way it distorts the economy is through inflation. Even the Fed has been forced to acknowledge that the buildout is already raising prices; the productivity payoff is still in the mail. The inflation brought by the buildout is another dose of pain for the parts of America that were already being squeezed.

The “AI boom” is also almost certainly a massive (and surprisingly fragile) bubble, whose bursting will maim the economy. Believing it is a bubble doesn’t deny that it is a revolutionary technological change that will endure. The internet was a bubble before it became a revolution. The power and endurance of a technological change can’t negate the timing mismatch between when the astonishing amount of investment will begin to pay off and when investors expect to be paid back. And when that mismatch is realized, there will be hell to pay.

But fear not: there is likely to be insult added to financial injury. Downturns are when businesses tend to adopt automation to cut costs, so a crash will likely accelerate the “AI jobs apocalypse.” Until quite recently, the CEOs of AI labs had been telling anyone who would listen that AI could eliminate half of all entry-level white-collar jobs within one to five years. On the short end, that’s plainly a fairy tale. But five to ten years from now? That time horizon is more ominous. Again, there is a timing mismatch. AI undoubtedly will create many new types of jobs in the economy eventually, but it’s not clear that it will be in time to employ the people who lose their jobs due to automation.

In a more human sense, the runaway arrogance of the people who stand to gain unfathomable amounts of wealth is most clearly seen in the overweening contempt with which they treat the rest of the country. “The year of efficiency”, “bullshit jobs”, “laziness being rewarded with success” or “the employee base going feral”, call it what you will. The people running the biggest tech companies have spent the last few years telling the world, in increasingly open terms, what they think of the people who work for them, and then cutting them down to size with ruthless and seemingly endless rounds of layoffs, happening in parallel as the AI “boom” ramped up.

It sent a loud message to multiple audiences. To their tech labor force it signaled that they had fallen in status and faced new economic precarity. To investors it signaled efficiency and cost-cutting that sent their stock soaring. To the rest of corporate America it signaled the near-term automation of the knowledge economy workforce. But it also sent a signal to an unintended audience, the rest of America: AI means job losses and falling social status. It took longer than it should have for tech leaders to understand the gravity of that mistake.

If people working on AI truly believe that what they are working on has a decent chance of wiping out all people, then they have gotten to the point of believing they are entitled to make that bet on everyone’s behalf. The rest of the species is just along for the ride (nobody bothered to check in with us).

AI has created a crisis of agency for people. The constant messages of inevitability that its backers blare out are an attempt to paper over its fragility and to obscure the massive sums that have been gambled on uncertain outcomes. When workers are told that their performance reviews, and by extension their livelihoods, hinge on how much they use AI, they’re going to tokenmaxx whether or not they want to. All this comes together as a feeling that AI is something being done to us rather than any choice that we made. And that’s triggering feelings of apathy, anxiety and drift, as your future gets written without you, along with a sense of profound anger.

There’s something sinister, cynical, and faintly ridiculous in the amount of effort that has gone into forcing a sense of inevitability onto the AI “boom.” Inevitability should speak for itself. Nobody needs a marketing campaign to convince the West Coast that the Big One is coming.

It is inevitable that eventually the Cascadia subduction zone and the San Andreas fault will rupture and unleash one of the most cataclysmic natural disasters in US history. Local and state governments that lie within the danger zone are in an unending yet agonizingly slow process of trying to prepare for that day. The constant preparation is not because of loud, ominous rumbling coming from these faults. It’s because the geologic record shows these faults have ruptured again and again, and nothing about the underlying physics has changed. The geologists warning about it gain nothing personally from being believed.

The biggest proponents of AI’s inevitability, on the other hand, get rich if you believe it. That doesn’t make them wrong, but it does mean their insistence counts for nothing as evidence. If you have to insist on it, and you get paid when people buy it, you’re not a prophet, you’re in sales.

“When truth leaves us, when we let it slip away, when it is ripped from our hands, we become vulnerable to the appetite of whatever monster screams the loudest.”

— Mon Mothma

It’s been interesting to watch a wave of antipathy toward AI build over the last several months, just as I’ve been trying to make my peace with it. The politicians who were so eager to court the data center frenzy last year have turned on a dime, now claiming they had their doubts all along. So it goes. What has been surprising is just how potent the backlash is politically. It doesn’t follow the traditional lines of polarization that most political issues hew to. It’s a coalition broad enough to have Bernie Sanders and Steve Bannon take the same stage at an anti-AI rally and preach for the same cause. A truly mind-boggling development in a country with a partisan divide as calcified as ours. And it didn’t come out of nowhere: Bannon was warning that entry-level jobs for people under 30 were about to be eviscerated back in May 2025, in the same Axios column where an AI lab CEO predicted half of them would disappear. Bernie was into data center moratoriums before it was “cool”.

But most telling has been the panicky, light-speed turnarounds we’re seeing from governors and local politicians. These politicians are closer to their constituents than Congress, which spends a good chunk of its year in Washington, DC, far away from the home states and districts that they were elected to represent. The speed at which the blue-state governor of New York and the red-state governor of Texas rolled out what amounted to bans on data center construction within weeks of each other shows that the pressure is broad-based. As much as venture capitalists, CEOs, and administration officials would like to pin the backlash on a Chinese psyop, it is a rational reaction from large segments of the country to something that, from their point of view, carries more peril than it does promise.

Nor do I buy the national security imperative as the reason large sections of the American public must bear the cost of the AI boom. The competition with China is real, that’s not in dispute. What I’m disputing is the sincerity of using it as a battering ram for the industry’s AI agenda. For years, Silicon Valley treated China as a market to be won, not a threat to be contained. Meta reportedly built a censorship tool to get back into China, while its CEO kept Xi Jinping’s book on his desk and bought copies for his colleagues. Even now, the same industry warning that America must win the AI race at any cost has lobbied to keep selling advanced AI chips to China when export controls threatened its revenue. National security, it turns out, is negotiable when it cuts into earnings.

What Silicon Valley seems to have on offer for our future is the economic caste system of the permanent underclass. They had better hope no one weighs that offer against the thing they keep telling us to fear. You can’t ask people to sacrifice to stop a future that, from where they sit, looks about as unfree as the one you’re selling them. The China threat only works as a threat if the alternative is worth defending. For a lot of Americans watching their job prospects shrink and their electric bills climb, that case is getting harder to make.

Through their own words and actions, the bosses of tech and AI created the public opposition to AI. The throughline of the AI backlash is a belief that “these guys are going to get filthy rich, and I’m going to lose out.” This belief is the core of the movement, whether it manifests as growing anger over rising electricity rates or as suspicion that local politicians are selling out their constituents to data center developers.

A New Hope

But I’ll admit that’s bleak. I think it is a virtue to have strong opinions that are loosely held, in the sense that when new information conflicts with what you hold to be true, you are flexible enough to face and incorporate the new reality. It’s important to face the facts even if they conflict with what you think you know. That’s something that I’ve had to do in regard to AI especially in light of recent events.

It would be fair to categorize me as more in the AI skeptic camp than the true believers. Recent history had conditioned me to be skeptical of the claims about the power of AI coming out of Silicon Valley. Venture capitalists, start-ups, and tech giants had spent a lot of credibility on hyping crypto and the metaverse. It seemed like there was a bit of desperation for finding the next big thing to keep the valuations up.

On top of that there was a huge gulf between what they were claiming and my own real-world experience using AI tools. The first time I used Claude Code, right after it was released, I asked it to translate Swift code from the SceneKit API to RealityKit. It promptly went to work and bricked my project’s .pbxproj file so badly that Xcode crashed every time I tried to open it. Git saved the project, but not my opinion of the tool. Stacked on top of that were the ever-present AI hallucinations that made using these tools unreliable at best. The enormous claims of AGI and the eventual creation of “the God machine” seemed like insane hyperbole if not outright hucksterism. It didn’t add up.

In the last six months that has changed. The tools have become more reliable, and more capable at completing complex tasks particularly in coding. It’s not flawless by any means, but there has been an enormous leap in its utility. Last week I rebuilt my personal site, which I hadn’t touched since 2023, with the same tool that bricked my Xcode project.

The transformation of AI’s utility happened first for programmers and has now engulfed mathematicians. I expect it will ripple out to other sections of the knowledge work economy in due course. Suffice to say, this makes it harder to dismiss out of hand the claims of AGI and super-intelligence coming in the not-too-distant future.

I had largely bought into the notion that every fantastical claim or warning from people at the heart of the AI economy was a twisted marketing ploy to juice their funding rounds and inflate the bubble. But the disclosure of incidents with “rogue AI swarms” over the summer along with the public pronouncements of the potential for human extinction by researchers at the AI labs has made me reevaluate. If anything these were disclosures and claims that endangered these companies and the bubble rather than pumping them up.

That’s not to say some of the claims weren’t made with an eye to pumping air and hype into the bubble. But it’s harder to take solace in the idea that it’s all a charade. It’s at the very least a mixed bag of motivations rather than outright desperation or a scam.

“There’s no fate but what we make.”

— Sarah Connor

So where does that leave me? Well, you could say that is also a mixed bag. If I accept that all the warnings of catastrophe are at least partially legitimate (not in the sense that they are going to come true, but that they come from a genuine belief rather than being purely constructed to serve an agenda) then I also have to do the same for the promised benefits. And if that is true, then it’s not all doom and gloom. Moderna’s AI-designed cancer vaccine just passed a Phase 3 trial, so the miracle cures are real (whether you can afford one is a separate question). AI holds the promise of unlocking economic and scientific advancements capable of helping us overcome problems we’ve been mired in for decades, paving the way for a future I would actually want to live in.

There is also hope for us in a more human sense. As our digital world becomes overrun by artificial content and bots, the value of genuine human connection will become clearer. Its value will increase as it becomes scarcer relative to all the other content out there. This is something I can feel hopeful about. The story of technology in the 21st century has largely been one of gradual isolation from other people, as digital communication put up barriers where real-world contact once thrived. But in a world where people can’t tell whether there is another person at the other end typing or a bot, the value derived from virtual connection will plummet.

As for AI being a crisis of human agency, the potency of the backlash as a political force gives hope that the damage can be undone. Americans overwhelmingly feel that their government doesn’t represent their interests, that the wealthy have bought off the politicians. Nothing ever changes, nothing ever gets solved, and there’s nothing you can do about it. But that hasn’t held true for the AI backlash. This unlikely coalition is making the politically impossible possible.

The American public largely seems to be turning on AI, with widespread loathing and fear of how fast it is being forced on us. The technology itself is inevitable. That earthquake is coming whether anyone likes it or not. But the arrangement that’s being built around it is not geology. It is a set of decisions, and the governors of New York and Texas unmade some of them within weeks of each other. The campaign to make the average American feel disempowered has been largely successful, but it is nonetheless a lie.

AI is here, and it’s not going anywhere (even when the bubble pops). The dread and the promise are both real. But the technology being here is the only inevitable part. Who pays for it, who profits from it, and how fast it gets shoved onto the rest of us are decisions, and they’re being made right now by people who get rich if you believe it was never up to you. I’ve made peace with the first part. Making peace with the second is what they’re counting on, and I haven’t. Thanos had the better line. Sarah Connor had the better point, and she was the one still standing at the end.