A Debt, Not an Externality

The intelligence doing the displacement was built on the labor of the displaced. That is not a side effect. That is something owed.

Last week’s piece The Weight of What’s Coming made a promise it did not keep.

The essay listed four proposed answers to the displacement crisis AI is producing across knowledge work. Universal Basic Income. Elon Musk’s “universal high income.” The viral critique of UBI as a cage. And a fourth, teased but not developed: a Pigouvian tax on automation, drawn from a paper by two tenured economists who found that every other intervention fails. The essay flagged that the tax was worth its own piece. This is that piece.

The argument is that the economists have the mechanism right and the category wrong. A corrective tax on automation is a serious instrument. But the frame they use to justify it, externality, does too much work in their paragraphs and not enough work in ours. An externality is a side effect the market fails to price. A debt is something owed. The distinction is not rhetorical. It reshapes what the tax is for, who it is owed to, and whether the refusal to levy it is merely bad policy or something more severe.

This piece also lands in the middle of a sequence. Monday’s The Luciferian Thesis named the cosmic pattern: human creative capacity, co-opted by an adversary whose strategy has always been to redirect it upward, toward the subjugation of image-bearers. Wednesday’s Tools, Not Taskmasters offered a diagnostic: does a given AI deployment extend the image-bearer’s capacity, or administer her? This piece applies that diagnostic to one very specific extraction: the one that built the models themselves.

What follows is an argument in two movements. First, a fair reading of the paper, which makes an economic case that even a skeptical reader of welfare economics should find difficult to dismiss. Second, a theological reading that takes the mechanism seriously and reframes the category. The economists did the harder work. The church’s job is to name what their vocabulary cannot reach on its own terms.

What the Paper Actually Says

Brett Hemenway Falk at Penn and Gerry Tsoukalas at Boston University posted "The AI Layoff Trap" to the economics-theory arXiv on 2026-03-21 (Falk & Tsoukalas, 2026). The frame is game-theoretic. Across fifty-three pages of formal model, they show, with conditions made explicit, that the rational individual choice to automate produces a collectively worse outcome even for the automating firms themselves. Workers lose wages. Consumer demand contracts. Firms lose the customers they optimized against. This is a prisoner's dilemma in the technical sense, a coordination failure that self-interested actors cannot escape by acting self-interestedly.

Having named the trap, the authors turn to exits. They test every standard policy instrument in sequence. Retraining programs and upskilling. Unemployment insurance. Wage subsidies. Capital income taxes. Worker equity schemes. Negotiated agreements between firms and labor. Universal basic income. Every one of these, modeled against the structural mechanism the paper describes, fails. Not for political reasons. Not because of implementation friction. The failures are features of the mechanism itself. UBI funds consumption without addressing the demand externality on the production side. Upskilling does not resolve the race-to-the-bottom incentive. Equity arrangements presume a level of worker bargaining power the trap removes.

The one instrument that works in their model is a Pigouvian tax on automation. Named for the early-twentieth-century British economist Arthur Pigou, a Pigouvian tax is a corrective tax designed to internalize the social cost of a private act. Carbon taxes are the archetype. So are the excise taxes on tobacco and alcohol. The tax is calibrated to make the private cost of the act equal to the public cost. In the Falk-Tsoukalas model, a tax on firms that replace workers with AI, priced to offset the demand externality the replacement produces, resolves the trap. It makes automation rational only when the productivity gain exceeds the social harm.

The authors concede, flatly, that this is politically impossible in the present United States. The capital class most likely to shape federal AI policy is the capital class whose automation strategies the tax would target. The paper ends by naming the trap, diagnosing the solution, and acknowledging that the solution will not be implemented. An economist’s version of what a pastor would call lament.

That last move matters. This is not alarmist theology dressed up in equations. This is a pair of tenured economists running standard welfare economics to its end and producing a conclusion their own discipline is structurally unequipped to act on. The paper is the guild reporting its own framework’s failure out loud. That is the reading The Whaling Collapse Came for New Bedford developed: when the economists start using the vocabulary the theologians have been using for years, something inside secular discourse has ruptured.

Externality Is the Wrong Word

Welfare economics defines an externality as a cost that falls on a third party outside a market transaction. The classical case is a factory that produces goods, sells them, and emits pollution into a river downstream. The buyer and seller share the benefit of the trade. The community that fishes the river bears a cost neither of them priced. The word names a leak. Something spilled out of the transaction that the parties to the transaction did not account for.

This is the frame Falk and Tsoukalas reach for when they name the mechanism. The AI firm automates, captures the productivity gain, and the displaced workers bear a cost the automating firm did not price. By the technical definition, that is an externality. The paper’s instrument follows from the frame: correct the leak by pricing the cost back into the firm’s balance sheet.

But the analogy breaks when you push on it. Carbon emissions are a byproduct of combustion. They are not what the combustion is for. The factory makes steel; the carbon is what escapes when it does. The relationship between the private act and the external cost is incidental. Change the combustion process, and the cost can, in principle, be removed without changing the product.

AI displacement is not like this. The output of a language model is not a byproduct of the displaced workers’ labor. It is the displaced workers’ labor, restructured and scaled. Every corpus that trained a frontier model was composed, line by line, of human judgment. Legal briefs. Medical notes. Code repositories. Marketing copy. Customer-service transcripts. Financial analyses. The model’s capability is a function of the training data the way a distilled spirit is a function of the grain it was made from. Remove the grain, remove the spirit. Remove the training data, remove the model.

That is the frame-breaking observation. The workers are not downstream of the AI firm’s productive act. They are upstream. They are the production. The cost the tax would correct is not a side effect of the firm’s automation decision. It is built into the object being automated. An externality frame treats the harm as incidental. The actual relationship is constitutive. The diagnostic I developed in Tools, Not Taskmasters—does this extend the image-bearer or administer her?—lands here as sharply as it lands anywhere. The system’s capability is the laborer’s work. The system’s deployment subjugates the laborer whose work it is. The direction of power flows upward at both ends of the transaction. Train the model on unpaid work. Deploy the model to replace the worker. The extraction is symmetric. The observation originated in When the Builders Stop Building: engineers had trained the AI that cut them, and the builders, in a precise sense, had built their own replacements. Scale that observation across every knowledge-work profession and the pattern is identical.

Scripture has a clearer word for this arrangement than welfare economics does. When one party’s output becomes another party’s profit without compensation, the tradition does not call it a spillover. It calls it unpaid wages.

Scripture on Wages, Debts, and the Laborer

Leviticus 19:13 (NKJV): “You shall not cheat your neighbor, nor rob him. The wages of him who is hired shall not remain with you all night until morning.” The syntactic move is worth noting. Three clauses. Three moral categories. Cheating. Robbing. Withholding wages. The text does not draw distinctions between them. Unpaid wages is placed in the same moral register as theft, because in the Hebrew ethic they are the same act. What the laborer produced belongs to the laborer until the laborer is paid. Delay in payment is retention of what does not belong to the retainer.

Jeremiah 22:13 (NKJV): “Woe to him who builds his house by unrighteousness and his chambers by injustice, who uses his neighbor’s service without wages and gives him nothing for his work.” The prophet is indicting Jehoiakim, king of Judah, who built his palace on the forced labor of his people. The indictment is not that the labor was unproductive. The palace went up. The craft was real. The walls were beautiful. The indictment is that the king extracted the labor and did not compensate it, which made the beauty of the palace evidence against the builder rather than in favor of him. Productive extraction, unpaid, is worse than idleness. Idleness merely fails to generate value. Unpaid extraction generates value by consuming the dignity of the laborer who produced it.

These texts are not metaphors. They are legal-ethical instructions from a tradition that took the wage of the worker with absolute seriousness. The scriptural framework does not treat the wage as a market variable to be negotiated against the employer’s alternative options. The wage belongs to the laborer by virtue of the laborer’s image-bearing, and withholding it, for any reason the market invents, is theft.

The scriptural framework extends further, into the accumulated structures of debt and ownership that produce withheld wages at scale. Deuteronomy 15:1-2 (NKJV) institutes the year of release: “At the end of every seven years you shall grant a release of debts. And this is the form of the release: Every creditor who has lent anything to his neighbor shall release it; he shall not require it of his neighbor or his brother, because it is called the Lord’s release.” The seven-year cycle is not charity. It is structural. It is built into the economic constitution of the covenant people because the covenant God knows that absent such mechanisms, debt compounds and the compound consumes the debtor.

Leviticus 25 goes further. Every fiftieth year, the Jubilee, ancestral land returns to the family that lost it. The slate is reset. This is not a redistribution argument in the modern ideological sense. It is a testimony. It testifies that the God who wrote the covenant built debt-forgiveness and structural restoration into the economic order, because He knew the natural tendency of accumulation was to concentrate wealth and hollow the laborer. The Jubilee is not a policy proposal for the twenty-first century. But it is a witness that any economic order unmoored from such correction will, over time, produce exactly the condition Falk and Tsoukalas are now describing in game-theoretic language.

The cumulative frame: unpaid wages is theft. Accumulated debt requires structural correction, not market equilibration. The laborer is an image-bearer whose work is not an input to be optimized but a participation in the Genesis 1:28 mandate. Any economy that forgets these has not yet reckoned with the tradition whose moral vocabulary it still uses to justify itself.

The Training Data Was You

Every frontier AI model is trained on human output at a scale the training-data ledgers cannot fully accommodate. The legal-brief corpus was written by lawyers. The medical-notes corpus by physicians, nurses, and medical scribes. The code repositories by software engineers. The customer-service transcripts by customer-service workers. The marketing copy by marketers. The financial analyses by analysts. The radiological imagery, the insurance underwriting memos, the HR policy documents, the translated literature. Every item is a crystallized act of human judgment. Every item was produced in the context of a job being done by a person. None of those persons signed a contract authorizing their output to be used to train a system that would eventually do the same work without them.

The scale of the extraction is what makes the case so strange to the existing moral imagination. If one lawyer’s brief were scraped without consent to train a legal AI, the doctrine of contract and copyright would have a clean mechanism. The harm would be individually nameable. The remedy would be nameable too. When ten million lawyers’ briefs are scraped, the doctrine struggles. Not because the moral structure has changed, but because the aggregation has confused the legal imagination into treating what is the same act, performed at scale, as a different act entirely.

But the moral structure has not changed. Ten million unpaid wages is not less of a debt than one unpaid wage. It is ten million times more of a debt. We experience it as less only because the harm is distributed thinly across many individuals and the benefit is concentrated heavily in a few firms. The market calls this efficient. The tradition calls it theft at scale. The word is ugly because the act is ugly. The scale does not launder it. The scale compounds it.

The cost of that extraction, as The Pipeline Paradox argued, is not borne by the firms doing the scraping. It is borne by the workers whose labor produced the corpus and by the next generation of workers who will never now apprentice into the professions the corpus fed. The artists and writers are pressing an early version of this case, through copyright suits against companies that scraped their output to train generative models. Those suits will resolve on narrow legal grounds and the rulings will matter. But the copyright frame is a subset of the moral framework, not its whole. The artists are arguing the leading edge of a case that belongs to every profession whose labor is in the corpus. The legal lever today is copyright. The moral claim underneath the lever is wages.

What This Looks Like in April 2026

At breakfast on Saturday, an elder from our church, a Fortune 500 engineer approaching thirty-five years at the same company this September, walked me through an internal dashboard his employer is using to track its workforce’s AI adoption. He had discovered it by accident. A few weeks earlier he had sat through an evening I hosted at our church on the theology and technology of AI, went home curious about what tools his employer made available to him, and started clicking through internal systems he had not explored before. That is how he found the dashboard. Nobody on his team knew it existed. When he raised it with them, he was the first to tell them they were being scored.

The dashboard greets him by name. It reports his standing: #403 out of the company’s roughly 23,000 employees. It awards him a badge, Explorer, and tells him he is 26 points from his next badge, Advocate, and then Pro, Expert, and finally Guru. It shows him where his four closest teammates sit on the ladder. It shows the public leaderboard, the top five employees named with their point totals, all Gurus. It coaches him: Keep climbing! Only 1,986 points to break into the top 5. His own rank has slipped, he told me. He suspects the drop is what he gets for taking a couple days off the week before. His read of the system was unsentimental: They want to encourage our growth in the AI realm by leveraging competitiveness and maybe a bit of fear.

Neither of us said the quiet part out loud, because neither of us had to. A company does not build a gamified adoption leaderboard for all 23,000 of its employees, with badges and cheery coaching copy and peer-visible rankings, because the CFO is curious about API spend. And a company does not leave the existence of such a dashboard for its employees to stumble upon unless the silence is itself part of the design. The ranking is a triage instrument. It is how the company will decide, when the decision comes, who stays and who goes. The softness of the framing is the coldness of the mechanism. Your learning journey. Keep up the great work! Join the top performers and inspire others on your AI journey! An entire vocabulary of self-development layered over an instrument whose function is to sort image-bearers into will-be-kept and will-be-released. A man approaching thirty-five years at this company is being evaluated not on the systems he has shipped, the teams he has developed, or the judgment he has accumulated, but on how quickly he learns to use the tool that was trained on thirty-five years of engineers like him. He has already internalized the pressure. A weekend of rest costs him ranking. Workers like him trained the thing. Now they are being ranked on their adoption of it, peer to peer, publicly, with badges, on a leaderboard most of them do not know exists. The one who does not climb will be released so the one who does can be released later.

Scripture has a specific category for this move. Jeremiah 22:13 does not indict Jehoiakim merely for extracting labor. It indicts him for extracting labor while concealing what he was doing. The palace went up. The craft was real. The workers were there. The king had arranged the arrangement so that his authority over the extraction stayed out of view. Unpaid wages openly contested is a labor dispute. Unpaid wages extracted from workers who do not know they are being scored for release is something the tradition treats as closer to fraud than to a labor disagreement. The silence is not an adjacent problem. The silence is part of the theft.

And the vocabulary is part of the weapon. The dashboard speaks a carefully engineered language to him: Your learning journey. Your personal AI impact dashboard. Power User. Keep climbing! The company is producing the language by which the worker will describe his own condition. If he accepts the vocabulary, he will experience his eventual displacement as a personal failure to climb fast enough. The tradition that formed this country speaks a different vocabulary: unpaid wages, theft, image-bearer, Jubilee, neighbor. Which set of words survives in the worker’s mouth determines whether he ends up asking why didn’t I climb faster? or why was my thirty-five years treated as a line item? The fight over the tax, if it ever comes, will be fought in the second vocabulary. The refusal to fight it will be justified in the first.

This is the diagnostic from Tools, Not Taskmasters operating at industrial scale. The direction of power has inverted. The employee is not being served. The employee is being administered. The ranking system does not exist for the employee’s benefit. It exists to make his replacement legible to the people who will decide to replace him.

A few days before that breakfast, Marc Benioff announced that Salesforce is restructuring around an agentic architecture, pulling back from the per-seat UX/UI model that defined the company’s previous era. Commentators have been trying to work out what it means for SaaS strategy broadly. There are cases where following the pivot now would be premature. Most B2C software still serves human users who are not about to disappear, and a company that stripped its human-facing interface for an agent layer right now would be building for a customer base that does not yet exist.

But the Salesforce announcement is not really a claim that human UI is going away in general. It is a claim that within Salesforce’s specific customer base, the humans who currently use Salesforce are going away. Salesforce built its business on per-seat licensing: ten thousand employees at a client company, ten thousand seats, ten thousand recurring invoices. If those client companies are about to cut their workforces in half and replace the rest with agents, per-seat licensing collapses. Benioff is not announcing a product pivot. He is announcing his read of his own customer base. He expects his clients to eliminate the seats he has been billing against. The platform is adapting to the displacement its clients are preparing to execute.

And read the second-order. Every major company has, for the last twenty years, invested heavily in UX/UI designers. The role has been one of the most strategically central positions on modern software teams, because the competitive advantage of a platform was the quality of the human experience it produced. If the interface a platform now optimizes for is an agent’s API call rather than a designer’s dashboard, the UX/UI designer has just been rendered obsolete. Not gradually. Not next cycle. Now, in one announcement, at one company, that a thousand other companies will read and respond to within the quarter. The same pattern the elder is living in his engineering ranking is being lived, today, by an entire profession whose work was until this week considered indispensable.

Put the two announcements side by side. The elder is being ranked on his token use because his employer is preparing to decide who to release. The platform his employer pays for is restructuring because its clients are all preparing to release people at scale. The supply side and the demand side of the same labor market are pricing in the same future. That future is not hypothetical. It is the one being budgeted for in April 2026, on conference stages and in HR dashboards, at the ranking of the elder at my breakfast table and in the press release from a CEO who knows exactly what his customers are about to do.

The debt this piece is arguing for is not a debt that will accrue in some speculative future. It is accruing this quarter, on the balance sheets of companies openly restructuring around the displacement. And it is being extracted today, from a thirty-four-and-a-half-year engineer whose rest already feels like delinquency, on a dashboard most of his colleagues do not know exists. The injury is live. The naming cannot wait.

A Pro-Capitalism Lament

I want to be precise about what I am not arguing.

Scripture acknowledges private property. Thou shalt not steal presumes a category of legitimate ownership worth protecting. The Jubilee law does not abolish private holdings; it corrects their accumulation over time. The parables of Jesus assume commercial activity, contracts, wages, and stewardship of resources entrusted to individual actors. The biblical economy is not collectivism. It is an ordered economy under moral constraint.

I also like the capitalist system that made America the economic and cultural engine it has been for most of its two hundred and fifty years. I am, by temperament and conviction, ardently pro-American. I have served and pastored in this country all my life. I have built companies under its system. The dynamism this country produced, the freedom it extended, the flourishing it seeded across generations, is not nothing. It is a real historical good, and I will not pretend otherwise.

Alex Karp, the CEO of Palantir, has frequently named Calvinism as the intellectual root of American dynamism. I would press the point a little differently and say Protestantism more broadly, and specifically the Protestant work ethic, the conviction that ordinary labor under God is itself a vocation and that the fruit of that labor bears moral weight. For roughly two-thirds of America’s history, that ethical framework was the operating assumption of American capitalism. The market was not without greed. It was not without sin. No economy this side of Eden is. But the system operated within a moral grammar inherited from the pulpit and the pew. The employer who withheld wages knew he was under a God who named that as theft. The worker whose dignity was extracted had a moral vocabulary available to name what was happening to him, because the same pulpit that formed the employer also formed the worker.

That ethical framework has eroded in the post-Second World War era, and more precipitously in the post-Cold War decades, until what we have now is a capitalism that retains the machinery of market exchange but has shed the moral vocabulary that held the machinery accountable. Shareholder primacy replaced stewardship. The labor input became a line item to be optimized downward. The productivity gain became a dividend for capital rather than a rising floor for workers. This is not capitalism per se. This is capitalism stripped of the grammar that kept it tolerable. Amoral at best. Tending toward immoral.

The biblical pattern for what happens to civilizations that optimize without moral constraint is not ambiguous. Babylon was judged. Tyre was judged. The prophets do not rail against the existence of commerce. They rail against commerce that crushes the poor, withholds wages, and accumulates without repayment. The nation that loses its ethical compass and continues to accumulate does so under a shadow.

I am a native of this country. I do not want to see it judged. Neither did Isaiah. Neither did Jeremiah. They prophesied against their own nations while loving them, which is why their prophecy reads as lament rather than contempt. I feel the same weight writing this. The capitalism I grew up inside, the country I love, is not operating the way the tradition that formed it would recognize. The debt being accumulated in this AI economy is one of many debts a morally unmoored economy is running up, and the unpaid wages of displaced image-bearers are among the clearest entries on the ledger.

Which is why the debt-based tax matters, even if it never passes. The instrument is not revolutionary. It is restorative. It asks capitalism to do what the tradition that birthed capitalism always assumed it would do: pay what is owed.

What a Debt-Based Tax Would Name

The move from externality to debt is not rhetorical. It reshapes what any corrective instrument is actually trying to do.

A Pigouvian tax framed as externality correction is priced to offset the social cost of a private act. The revenue flows to the general fund. The remedy is generalized. The harm, once corrected in aggregate, is assumed to be resolved. This is the carbon-tax model: tax the firm, fund whatever the state chooses with the proceeds, trust that the price signal will reduce the behavior over time.

A debt-repayment mechanism is different in kind. It is priced against what is owed. The revenue flows toward restoration: to the workforce whose labor built the training data, to the communities hollowed by the displacement the training data made possible, to the households whose rent is due next week because the skill they spent twenty years developing is now a column in a model’s weights. The question the instrument answers is not “how do we make the firm internalize a cost?” The question is “how do we pay a debt that has accumulated in the construction of a system whose beneficiaries have never acknowledged its existence?”

That distinction changes the design parameters at every level. An externality tax can be modest. A debt-repayment mechanism has to be proportionate to the debt. An externality tax pays itself as it corrects. A debt-repayment mechanism pays the creditor until the debt is satisfied. These are different instruments, aimed at different moral objects, even if they are both called Pigouvian in the paper.

Whether such an instrument is politically achievable is a question I will not pretend to answer. Falk and Tsoukalas say it is not, and I am inclined to believe them. The capital class shaping AI policy has no structural interest in the tax being passed, and the political economy will not produce it voluntarily. But the argument is not that the instrument must be politically achievable before the moral frame is correct. The order is the other way around. The debt exists whether or not the debtor has the political will to service it. The moral claim is prior. The political viability is secondary. If the instrument is never designed, the debt does not disappear. It accrues.

My own position remains what it was last week. I am not a legislator, an economist, or a policy analyst. I build AI tools, study their ethical implications, and pastor a congregation in the shadow of what both are producing. The instrument design is not my work. The moral framing is. Whatever instrument is eventually designed, whatever combination of taxes, trusts, compensation schemes, or data royalties, will only be proportionate to the harm if the harm is named correctly at the start. Externality language will always underprice the remedy. Debt language will not.

What the Church Must Name

The church cannot design this tax. It was never supposed to. The church’s vocation in economic moments of this scale has historically been different from the vocation of the legislator or the economist. The church’s work is naming.

The prophets named what the economy refused to name. Amos named the treading down of the poor. Jeremiah named the unpaid wages of Jehoiakim’s palace. James named the cries of the reapers whose wages had been kept back. None of them designed the instrument that would repay what the economy had withheld. They named the debt. The naming was the prerequisite. The instrument, when it came, came later, or did not come at all. The prophetic task was the naming, because the economy could not name itself.

I write this as a pastor. Which means the people this economy is administering are not data points to me. They are specific faces in specific pews, specific families who are preparing their children for a labor market that has already decided it does not need them. In late February, Jack Dorsey cut nearly four thousand of Block’s ten thousand workers, roughly forty percent of the company, and announced it in an early-morning note posted to his X feed. The note was addressed “to the company.” Thousands of employees learned they were being let go from the same public post the rest of us read over our coffee. The coldness of that particular move is what the unmooring of the last section actually looks like when it operates at scale. No pastor was in the room. No tradition was consulted. The efficiency was total, and the acknowledgment of what had been taken was absent. That is not a personal indictment of Dorsey. It is an observation about what an economy administers when the moral grammar has been removed from its vocabulary. The laptop opens, the post is already there, and the dignity the worker brought to the job the day before has been deleted overnight.

Here is the thing about the naming, though. It works. The elder I described earlier did not find the dashboard his employer was using against him because a journalist broke the story or a union filed a complaint. He found it because he sat through an hour of pastoral teaching on the theology of AI, went home with a question he had not had the day before, and started looking. Nobody on his team knew what he would find. Nobody on his team knew it was there to find. He brought it to them at a meeting, and four more image-bearers saw what had been hidden from them because one man had been given the vocabulary to see it first. That is not an exhortation. It is a report. The church is not only calling for the naming. The church is where the naming already happens, in ordinary conversations on ordinary weekends, when pastoral teaching equips a worker to recognize what his own company has hidden from him in plain sight. That is the prophetic function working exactly as designed.

What the church must name in 2026 is this: the workers whose labor trained the systems that are replacing them are owed something. The something is not gratitude. It is not a retraining grant. It is not sympathy. It is a debt—identifiable, structural, scaled—and the refusal to name the debt is injustice in the precise sense Scripture uses the word.

And the church must name this from inside the displacement, not outside it. Ministries I serve are running the same calculation every secular firm is running. A commentary translation that once employed human translators for five years now runs in an hour. An administrative assistant becomes a prompt. A staff writer becomes an editor of model output. We are saying “I am sorry” to the people whose work we replaced. We are choosing the machine anyway, because stewardship of the mission requires it. The prophetic critique has never been issued from outside the compromised community. Amos was an Israelite prophesying against Israel. Jeremiah was a priest prophesying against his own court. The critique comes from inside, or it does not come at all.

The anthropological premise beneath all of this, the one developed in The Child Who Couldn’t Do Anything, is that the measure of a human was never economic output in the first place. Which means the debt is not measured in productivity terms at all. It is measured in dignity denied. That is why the tradition calls the failure to pay it theft rather than inefficiency. Inefficiency is a problem of optimization. Theft is a problem of the soul.

A civilization that refuses to name its debts has severed itself from the moral grammar that sustains any just economic order. The debt the AI economy owes the workers whose labor built its models is not the only debt in the ledger, but it is an especially clean one. The creditor is identifiable. The benefit is concentrated. The mechanism of extraction is documented in the training-data disclosures the labs themselves produced. The only thing missing is the word.

The word is debt. Not externality. Not spillover. Not cost of progress. Debt.

And the question before the church in 2026 is whether we will say it out loud, inside our own compromised presence in this moment, or whether we will wait for the market to say it for us. The market will not. It does not speak the word at all.

---

Sources

1. Falk & Tsoukalas, “The AI Layoff Trap” (arXiv:2603.20617)

2. The Luciferian Thesis — Miles DeBenedictis

3. Tools, Not Taskmasters — Miles DeBenedictis

4. The Weight of What’s Coming — Miles DeBenedictis

5. The Whaling Collapse Came for New Bedford — Miles DeBenedictis

6. When the Builders Stop Building — Miles DeBenedictis

7. The Pipeline Paradox — Miles DeBenedictis

8. The Child Who Couldn’t Do Anything — Miles DeBenedictis

9. Jack Dorsey, note to Block on nearly-half workforce reduction (X, 2026-02-26)

---

This article was developed using AI writing tools I built to work with my voice, research, and editorial framework. The ideas, arguments, and theological positions are mine. The pipeline that helps me draft, evaluate, and refine them is something I created as part of my work at Nomion AI. I believe in building with AI and being honest about it. If you want to know more about that process, ask me.