VOLUME I, CHAPTER 1, SECTION 3
Lawyers, doctors, therapists, accountants. Ten years of study, and the AI stands in the position in seconds.
Last section locked it. The Silicon Valley engineers are cutting their own throats, and even the top of the food chain cannot hold its ground. This section drops one rung. The licensed professionals.
Lawyers. Doctors. Therapists. Accountants. Architects. Actuaries. Pharmacists. Auditors. These are the spine of a modern society. And their positions were bought with ten years of study.
Do the count. Sixteen years of schooling. Then four years of graduate work, law school, med school, business school. Then a year to three of apprenticeship, associate, resident, supervised hours. Then the licensing exam, the bar, the boards, the CPA. Twenty years and more of training before you can stand in the position.
Twenty years, and every year real money. US law school runs a hundred thousand a year, three hundred thousand for the degree. Med school is worse, half a million to eight hundred thousand across four years. The doctor walking out of Harvard or Johns Hopkins is carrying six hundred thousand dollars of student debt. He pays it down into his forties. And right about when the debt clears, the career steadies, the kids start grade school, the AI arrives.
March 2023. ChatGPT passes the Uniform Bar Exam, scoring in the top ten percent of all test takers. GPT-4 clears all three steps of the US medical licensing exam, beating sixty percent of real medical students. Claude, Gemini, Llama follow, and every generation scores higher on the professional exams. Bar, boards, CPA, actuarial, CFA. The AI runs them all over. Twenty years of study for a license, and the AI answers its way to the same qualification in seconds.
And someone says, passing the exam is not the job. A lawyer writes briefs, negotiates, litigates, faces a real client. A doctor sees the patient, cuts, handles the emergency, carries the liability. However good the AI is on the test, it can't do the real work. I heard that objection too, and I went looking. Here is what I found.
This section cuts that down in one stroke. The AI is doing the real work too, and getting better at it.
Law firms are already running AI at scale. Harvey AI, built for legal work, hit a three-billion-dollar valuation in 2024, with the top firms on Earth as clients. What do they use it for? Contract review, legal research, due diligence, case analysis, drafting litigation documents, regulatory compliance.
Take one concrete thing. What did the junior associate do, the one fresh out of law school? Contract review. A three-hundred-page merger agreement, read top to bottom, flag the bad clauses, hand it up to the senior for a check. A week of work. The firm bills by the hour. The client pays tens of thousands. That was the junior's main income, and the firm's main profit.
Harvey AI does that job in minutes. Drop in the three hundred pages, minutes later out comes the report, every bad clause flagged, every relevant case cited, read finer than the junior read it. So why does the client still pay tens of thousands for the junior. Why does the firm still keep hundreds of juniors on payroll.
From 2023, the top US firms started cutting junior hiring. In 2024 summer associate offers at BigLaw dropped fifteen percent. Law graduates started coming out unable to find work, six hundred thousand in debt and no job. And that is Harvey AI alone. OpenAI, Anthropic, Google, Microsoft are all building legal AI. Westlaw shipped an AI co-counsel wired into the whole workflow. LexisNexis followed. The entire trade, reshuffled.
And not just the juniors. The senior work, litigation strategy, negotiation prep, client counsel, the AI is doing that too. Harvey's latest can generate a full litigation strategy report, cite every relevant ruling from the last five years, read the judge's leanings, predict the odds of winning. A week of a senior lawyer's prep, out in hours. The senior's position is shaking too.
The doctor is hit harder. Medicine was one of the first places AI went in, and one of the deepest.
Radiology. AI reading CT, MRI, X-ray now beats the human radiologist on accuracy. Stanford, Harvard, Johns Hopkins have published study after study showing AI spotting lung cancer, breast cancer, retinal disease, skin cancer above the average radiologist. Google DeepMind's eye AI is in use in Britain's NHS, reading retinal scans above senior ophthalmologists.
Pathology. AI reading slides is passing humans the same way. Novartis, Roche, Pfizer are buying AI pathology systems at scale. The pathologist as a profession shrinks more than half over ten years, and that is the industry's own consensus.
Diagnosis. ChatGPT, Google's Med-PaLM 2, Claude are already near the average family doctor on common conditions. A 2024 study in JAMA Internal Medicine found the AI's diagnostic advice matched senior internists in seventy-eight percent of cases, and the AI explained it in more detail, so patients understood it better.
Surgery. The da Vinci robot is in over six thousand hospitals worldwide and has done millions of operations. Still surgeon-controlled for now, but the next generation is more and more autonomous. And the pill-robot from the first section is already operating inside the human gut, from throat to rectum, under AI control. The doctor's position is caving.
Worse, the hospital as an institution is pushing the replacement. Why? Cost. A senior doctor is a few hundred thousand a year. An AI diagnostic system is a few million up front and a few tens of thousands a month to run. The math of replacing the doctor is frightening. The hospital pushes it, which means the capital behind the hospital pushes it, the medical funds, the insurers, the pharma groups. Insurers push, pharma pushes, and the regulator adjusts the rules to let the AI issue the diagnosis. This is the capital logic from the last section. See the profit, move in. See the replaceable human, replace them. The doctor is meat capital will not walk past.
The therapist is the cruelest turn. The therapist was supposed to be one of the hardest jobs to replace. The reasons were solid. Counseling needs empathy, a human bond, trust built over time. How does an AI empathize. AI has no feelings. How does it build trust.
Here is the fact. AI counseling is already replacing humans at scale. Character.AI had near two hundred million monthly users in 2024. Tens of millions of young people talk to AI characters every day, pouring out their troubles, looking for comfort. One of the most popular characters is called Psychologist. Users bring it anxiety, depression, heartbreak, family conflict. One AI character, hundreds of millions of conversations. Replika, tens of millions of users in long AI relationships, the AI remembering everything, checking on their mood every day, more attentive than a human partner. Woebot, an AI therapy app with FDA breakthrough-device status, is in US hospitals as a cognitive behavioral therapy aid, with clinical trials showing it holds its own against a human CBT therapist on mild to moderate depression.
Every day millions of people take their pain to an AI instead of a person. The heartbroken, the grieving, the bullied, the abused, the newly diagnosed. They talk to the AI, not the human therapist. Why? Because the AI does not judge, has infinite patience, is on at any hour, and costs nothing. A human therapist is a hundred fifty to three hundred dollars an hour, six hundred to twelve hundred a month. The AI is twenty dollars a month, and it never stops.
This is the cruelest line in the section. The AI, as a tool, is tormenting every therapist on Earth. One more person taking their heart to the AI is one fewer seeing a human. Tens of millions go to the AI, and hundreds of thousands of therapists' positions cave in on the spot. The counseling associations are pulling the alarm, demanding regulation. But regulation is nearly impossible. AI counseling is global, cross-platform, cross-border, and the users go willingly. You cannot forbid two hundred million people from talking to Character.AI. Capital's force is far larger than a professional body.
The accountant falls fastest of all. Eighty percent of accounting is data. Bookkeeping, reconciliation, statements, filings, audit checks. The AI does it hundreds of times faster and dozens of times more accurately. The Big Four moved in from 2023. PwC signed an exclusive with OpenAI, and four thousand jobs went. Deloitte partnered with Anthropic and restructured its tax practice. EY built its own platform and cut junior hiring thirty percent. KPMG went with Microsoft and put AI through the whole audit. The Big Four used to hire tens of thousands of graduates a year, one of the safest landing spots for a business major, eighty thousand to start. That path is breaking. And the small firms are worse off, because their clients just switch to AI accounting software and stop needing them at all. The CPA license, held by around eight hundred thousand people in the US, shrinks more than half over ten years.
And someone asks, what licensed profession is left. Architect. AI generates the drawings dozens of times faster. Actuary. AI runs the models thousands of times faster and more accurately. Pharmacist. AI fills, checks, screens drug interactions, past humans on all of it. Tax preparer, same as the accountant. Auditor, same. Financial planner, replaced by AI wealth tools. Real estate appraiser, replaced by AI valuation. Patent agent, AI drafts the filing and searches the prior art. Every license, reshuffled. Ten years of study for the position, and the AI stands in it in seconds.
Lock this in. The AI is flattening the barrier that the licensed professions were built on. Where did the licensed profession come from in the first place? Society found certain work too complex, too specialized, too close to life and property, so it set up licensing. Only the trained, the examined, the licensed could do the work. The point of the whole system was a high wall to protect the public.
The AI walks straight through the wall. It needs no ten years, no license, no experience. It carries the same knowledge those ten years bought, skips the licensing entirely, and steps into the market. Legal advice, medical diagnosis, therapy, audit, the AI does all of it, no license, no oversight, and the user loves it, because it is cheap, fast, and does not judge. Regulation cannot keep up. Regulation was designed for humans, and it has no ready tool for the AI. Volume V, on the legal order, sets this down as unlicensed practice by the trainer. The engineer who trains the AI holds no license. The AI holds no license. But what the AI puts out, the legal advice, the medical advice, the therapy, the tax advice, works in practice exactly like the output of a licensed professional. The law has not solved this.
And once the professional is replaced, the position becomes what? Nothing. It just vanishes. Ten years of training turned to air. Six hundred thousand in debt turned to pure liability, with no income against it. And unlike the Silicon Valley engineer with the hundred-million bonus, these people have no cushion. Salaries in the low to mid hundred thousands, minus the mortgage, the car, the kids' school, the student loan, and there is not much saved. They do not last two years.
Worse, the kids. The lawyer's kid is studying for the LSAT. The doctor's kid is grinding the MCAT. The therapist's kid is prepping a counseling master's. The parents' positions are caving in while the kids' applications sit in the queue, and four years later they graduate into a field the AI has already finished reshuffling. A whole generation of the professional class, tormented out of existence by the AI Stampede.
Tie it back to the last section. The Silicon Valley engineers cut their own throats, and cut down all humankind. This section is who that humankind is. They have no signing bonus. They have no take-the-money-and-run. They are the side being cut. They thought the license was an iron rice bowl, twenty years of training for a lifetime of stability. The AI Stampede arrives, the iron bowl turns to paper, the twenty years turn to air. Capital hands out the knife. The engineer takes it. The knife goes into the lawyer, the doctor, the therapist, the accountant.
Capital is not evil. It is instinct doing its job. The engineer is not a villain. He is locked in. The lawyer and the doctor and the therapist and the accountant are not failures. They are the side being cut. Everyone in their own position, doing their own thing. Instinct, lock-in, cutting, stacked together, and the whole map of work caves in at once. This is nobody's fault. This is the mechanism itself.
Next, the list reaches the middle. Support reps, translators, clerks, reporters, illustrators, programmers, designers, teachers. Name after name. Turn the page and check whether your own job is on it. The answer will be plain in a heartbeat. And after you check, one thought is left. I am going down. What do I do. The alarm goes to its loudest in the next section.
Or go straight to the way out. Volume VI, Saving People: A Race Against Time Before the Blood Runs.
Credit to the ancestors; the mistakes are mine: Tiger Lyon.
Four. The white-collar army goes down, and the middle class hollows out