The Week the Human Became Curriculum

One man, his brother, and one point eight billion dollars. Matthew Gallagher invested twenty thousand dollars in September 2024 with no office, no investors, no employees. He built Medvi, a telehealth platform for GLP-1 weight-loss drugs, using ChatGPT for code, Midjourney for ads, Runway for video, and an AI chatbot for customer service. One year later, the New York Times verified his financials: four hundred and one million dollars in sales, two hundred and fifty thousand customers, a net margin of 16.2 percent. The 2026 projection: $1.8 billion. The only employee is Elliot, his younger brother, age thirty-six, working from Cincinnati filtering communications. Sam Altman had predicted the billion-dollar one-person company. Tyler Cowen titled his Marginal Revolution post accordingly. What no one titled is the comparison: Hims & Hers needs 2,442 employees to generate comparable revenue. Medvi generates two hundred million per worker. Hims, nine hundred thousand. The ratio is not a competitive advantage. It is an extinction event.

Four thousand freelancers teach their trades to the machine that will replace them. OpenAI calls it Project Stagecraft internally, a name that already contains the confession: it is about setting a stage. Three to four thousand contractors, paid fifty dollars an hour, simulate 439 distinct professions: commercial pilots, emergency physicians, sculptors, pharmacists, geoscientists, music directors. Each builds task simulations replicating real work: context, goals, deliverables, references. Wired reported that a parallel initiative asks contractors to upload actual files from past jobs: Word documents, PDFs, code repositories. The responsibility for stripping confidential information falls on the contractor. IP lawyer Evan Brown of Neal & McDevitt called it putting the lab at great risk. One contractor told Business Insider the quiet part: "We all were aware that we were basically training AI to replace us." What is being extracted is not a mineral. It is a competence. And the miner is paying for his own pickaxe.

The forecast says growth. The fine print says disappearance. The Forecasting Research Institute assembled sixty-nine economists, fifty-two AI experts, and thirty-eight superforecasters for the most comprehensive study yet on the economic impact of this technology. The baseline is moderate: 2.5 percent GDP growth, 61 percent labor participation by 2030. But the rapid-progress scenario, assigned a 14 percent probability by economists, tells another story: 3.5 percent GDP growth, labor participation falling to 55 percent, ten million fewer jobs, 80 percent of wealth held by the top 10 percent. Wealth rises. Employment falls. Between both curves fit exactly the people no one will mention in the press release. The Wall Street Journal reported 640,000 AI-related jobs created in the US between 2023 and 2025. That is a large number. So is the number of jobs those 640,000 are designed to eliminate.

Harvard gave the adviser's job to the chatbot. First they canceled the faculty pre-concentration advising pilot: six hired professors, fifty students each, eliminated in March 2025 after two years of operation. Then, in April 2026, they announced the replacement: a chatbot built on ChatGPT Edu that answers questions about concentrations, graduation requirements, and course selection for the Class of 2030. Brooks Lambert-Sluder, Assistant Director of Advising Programs, assured that the tool is not meant to replace human advising. The sequence says otherwise. You eliminate the human, install the machine, declare the machine does not replace the human. Peer adviser Georgia Bohney explained the logic: over 50 percent of student questions have definitive answers available online. What she did not say, because she did not need to, is that the other half cannot be answered by any algorithm. That half just had its door closed.

Jack Dorsey turned his company into a circle with AI at the center. On February 26, 2026, Block eliminated approximately four thousand positions, reducing headcount from over ten thousand to just under six thousand. Shares rose 16 percent that Friday. Dorsey published an essay with Sequoia's Roelof Botha titled "From Hierarchy to Intelligence" describing the new structure: an AI "world model" at the center, humans at the edge, three roles only: individual contributors, directly responsible individuals, and player-coaches who still write code. No middle management. The machine coordinates. Dorsey tripled headcount during the pandemic and now cuts it calling it vision. A data scientist named Naoko Takeda survived the cuts, refused a 90 percent pay bump, and quit: she would rather her peers kept their jobs. Forrester Research estimated 55 percent of employers regretted AI-linked layoffs, and a third spent more rehiring than they saved. But the stock had already risen. And in this operation, that is the only thing that counts.