Nita is a body on the annotation operation I own through two shells in southern West Virginia — a staffing shell that employs her and an operations shell that absorbs the IP assignments and bills the platforms. She is a contractor, not an employee; the distinction saves me $4,100 a year in matching contributions and I reclassified her entire line to achieve it. I pay her $9.25 an hour. You carried an AI assistant in your pocket last week. It finished your sentences. It guessed what you meant. Nita taught it what a sentence is. Fourteen million labels a quarter come off my floor, and Nita placed 3,200 of them in a single shift — 400 an hour, one every 9 seconds, the pace my scanner enforces because any slower and the throughput column on my dashboard turns the wrong color. You got your email drafted. She got $9.25 an hour. Her labor costs me $0.0029 a label.
Holman Jenkins, at the Journal, is alarmed that the government might claim a share of AI profits. He argues that Washington already captures 27% of national income — that taxpayers and the federal treasury are already positioned to benefit from AI‑related growth without any further action. He is correct about the arithmetic. I want you to notice where his accuracy stops, because it stops one sentence before the one that matters: I am positioned too, and I arrived at that position before Washington did. The model that will generate Jenkins’s future taxable revenues was built on labor I purchased at $9.25 an hour, in a county where I was the only employer posting wages that week. Jenkins wants to protect the industry from confiscation. I want you to read the next line slowly.
The wrist brace Nita needs to keep clicking costs $8. I don’t buy it. The annual eye exam costs $150. I don’t buy that either. The content‑counseling session for the man on my Nairobi crew classifying assault imagery — that budget line was deleted at $400 a month per head, three quarters ago. The per‑label savings from removing all three: $0.00017. Across 14 million annotations a quarter, that comes to $2,380. The compliance review at my staffing subsidiary costs me $2,400 a quarter. The $20 difference would buy the wrist braces for every annotator on my floor for a year. I kept the money. Her wrists won’t last another 18 months.
Now widen the frame. My total labor cost across the shell — every annotator, every shift — is $520,000 a year. The models my labels train are valued in the hundreds of billions. That is a return above 100,000 to 1. Jenkins thinks 27% is sufficient for the government. He may be right. But the Treasury will collect its percentage regardless, because the income exists. What I captured before that income ever reached the federal ledger — that is the margin Jenkins does not write about. He guards the front door while I walk out the back with the room.
And now the government has announced a thirty‑day review of AI models. Jenkins calls it a taking. I call it a free signal. The compute spins down, the model waits, the investors fidget — and the labeling queue does not stop. Nita still sits. The wrist brace still costs $8. The review period raises the scarcity value of the models, which raises the price I can charge for the labeling that makes the models possible. It costs me not $1 of the premium. The government is holding the product back; I am holding the wage back; the government is doing the holding for free. The investor’s panic is my pricing signal, and I received it at no cost.
You think I’m unusual. I’m a curriculum. The young woman who re‑priced my Nairobi counseling line sat in my seminar 3 years ago. She ran the subtraction faster than I did. Another graduate operates the shell outside Juárez where Maritza grades the same images — violence, pornography, the inside of a wound — 10 hours a day, at $13.25 an hour, with no chair at all. The arithmetic is identical. A thirty‑day review in Washington raises the scarcity value of the models, and the cost of the labor that trains them stays flat, whether the laborer sits in the Delta or across the border. One theorem. Two bodies. The principle does not care about the geography.
Leviticus 25:44 — “Both thy bondmen, and thy bondmaids, which thou shalt have, shall be of the heathen that are round about you.” The West Virginia hollers are round about me. Juárez is round about me. The review period has done nothing to change the bond. The bond is older than the technology it trains. The bondservant is the injured party, and the bond is the taking the government has been conducting since before the model was a line of code. Jenkins cannot see the body because he is looking at the investor’s balance sheet and calling the investor the injured party. The investor is not the injured party. Nita’s wrists are the injury. The $8 is the margin. The review is the gift.
“For the wisdom of this world is foolishness in God’s sight. As it is written: ‘He catches the wise in their craftiness.’” — 1 Corinthians 3:19. I teach this to the students who will one day set the wage on the job you are afraid of losing. They arrive sentimental. Most of them still believe a worker’s years mean something. I cure it by October. By spring, the best of them see the differential before I do. I am not a monster you can quarantine. I am a faculty. The students are already in your building.
Nita will clock in tomorrow at $9.25 an hour. The scanner will hold its pace. She will place another 3,200 labels and they will become my property — perpetual, irrevocable, her name nowhere on the three‑page IP‑assignment clause — before her shift ends. The government will review. The investors will fret. I haven’t looked at where her wrists will be in 3 years. I look at the operating margin. It is pretty.
Sterling A. Varice holds the Hayek-Friedman Chair and serves as Dean of Instruction at Warden University’s College of Business and Economics in Richmond, Virginia. He is the author of three textbooks: Divine Mandates for Labor Utilization, Social Obligations for Profit Maximization, and Calibrated Deprivation: A Manager’s Guide to Employee Motivation.