Information is free now. Type a question into Claude and you get a competent answer in seconds. ChatGPT can write your essay, solve your calculus problem, draft a legal brief, produce a marketing plan, and debug your code. The knowledge that used to cost four years and $100,000 is now available at zero marginal cost.

So why isn't everyone succeeding?

Because the degree was never about knowledge. It was about discipline. And nobody can AI their way out of needing that.

The Thing the University Never Admitted It Was Selling

Think about what college actually required of you. Show up at 9 AM. Turn in the paper by Friday. Study for the exam on Tuesday. Sit through the lecture even when it was boring. Do the reading even when nobody was checking. Submit to a structure somebody else designed for four years.

That is not knowledge transmission. That is discipline scaffolding. The university rented the time, the place, and the structure. You supplied the submission. The credential at the end was a receipt that said: this person can finish things.

The content was almost incidental. You could have learned the same material from books, from YouTube, from conversations with smart people. You didn't need a $40,000 annual tuition bill for access to information. You needed the structure that forced you to actually engage with it.

The university disguised this by bundling everything into one product. The credential. The lectures. The social network. The maturation experience. The alumni connections. One price, one package, one receipt at the end. The credential function generated the revenue. Everything else was a cost center riding on the credential's back.

The GI Bill scaled this bundle to the middle class. Eight million veterans got tuition paid. Employers adopted the degree as a cheap screening filter. Over 30 years, the degree became the default expectation for middle-class employment. Nobody designed this. It was an accident of history. a specific configuration of government subsidy, employer demand, university capacity, and cultural script that happened to align.

That configuration is dissolving. And when the knowledge part goes to zero, what's left is the thing that was always there, now visible for the first time.

The Gym Membership That Nobody Used

A university is like a gym. Some people get the personal trainer. Small classes, faculty who know their name, real expectations with consequences, oral exams, the Harkness table. That's a real discipline scaffold. Someone is watching. Someone is holding you to a standard. The cost is high but the product is real.

Most people got the empty gym membership. The big lecture hall, the adjunct professor who doesn't know your name, the TA who grades your paper against a rubric, the automated plagiarism checker, the GPA that's an average of AI-generatable outputs. You paid for the card. Nobody checked whether you exercised.

The elite school is the personal trainer. People will keep paying $80,000 a year for that. not for the knowledge (which is free) but for the structure, the peer pressure, the mentorship, the being held to a standard. The elite credential becomes more valuable post-AI because it certifies something AI cannot produce: having actually submitted to real discipline applied by real people.

The middle-market school was the empty gym membership. The credential was the membership card, not proof of a workout. AI just makes it obvious that nobody was actually in the gym.

The Scale Problem

This isn't a moral failure. It's a math problem.

The university scaled from 4 million students in 1945 to over 20 million today. You cannot have a personal trainer relationship with 20 million people. The ratio of students to faculty who actually know your name went from roughly 10 to 1 in the small-college era to 300 to 1 in lecture halls and 30 to 1 in classes where TAs do the actual grading.

To function at that scale, the university replaced relationship-discipline with bureaucratic proxies.

A teacher who knows your work and adjusts to you was replaced by a syllabus. A fixed document any AI can follow. Real feedback on your thinking was replaced by a rubric. A checklist any AI can match. A mentor who would know if you cheated was replaced by a proctor or an honor system. The proctor cannot tell AI work from student work. A master's judgment of your competence was replaced by a GPA. An average of AI-generatable outputs.

The credential was never certifying "you know things." It was never even certifying "you submitted to real discipline." It was certifying "you were processed through these bureaucratic proxies without breaking them."

What the Princeton Professor Discovered

Rory Truex is a tenured professor of politics at Princeton. In May 2026, he published a piece in the Chronicle of Higher Education called "The Semester That Shook My Faith as a Professor." It went viral.

Truex described teaching in Princeton's Einstein classroom. the kind of room that symbolizes everything elite higher education is supposed to be. And he found himself asking a question that would have been unthinkable a year earlier.

"I kept thinking to myself: Is this really the best way for my students to learn? At this point, wouldn't they be better off just conversing with Claude for an hour or two a week?"

He wrote: "What was I anymore? Just a tired, 41-year-old man, with Claude Haiku levels of understanding about China, armed with some dad jokes and outdated cultural references."

This is not a man worried about students learning less. This is a man discovering that what he thought he was providing. knowledge transmission. was never the real product. And the real product. the discipline scaffold. was something his institution had systematically underinvested in for decades while overinvesting in everything else.

The data backs him up. Half of Princeton's incoming class of 2029 used AI to write their college application essays. In a senior survey, 25 to 37 percent admitted using AI when not allowed. Sixty five percent knew of a peer cheating and chose not to report it. A startup called Companion.AI built a "homework agent" named Einstein that logged into Canvas, watched lectures, did readings, and uploaded assignments. The CEO, age 22, called it "freeing horses from their carts."

Princeton's response was telling. Take-home exams dropped from 168 to 49 in one semester. Oral exams returned to smaller classes. The institution is scrambling to rebuild the discipline scaffold it outsourced to bureaucratic proxies decades ago. because the proxies no longer work when the cost of producing acceptable outputs drops to zero.

Truex's crisis is the credential collapse thesis arriving inside the credential class's own consciousness. If a Princeton professor can't tell whether his students did the work, the degree he signs his name to has no meaning. Not because students are worse. Not because the content changed. Because the discipline scaffold that gave the credential its value was replaced by a verification system that AI can game trivially.

The Guild Was Always a Discipline Machine

Before universities, there were guilds. A master took apprentices. The master provided structure, set deadlines, inspected work, demanded repetition until mastery, and withheld sign-off until the standard was met.

The master's function was never "knows more things." It was: applies discipline pressure. The scarcity was always the master's attention. their willingness to supervise, correct, demand better, and refuse to sign off until the work met the standard.

The guild model didn't die because it was worse at producing competence. It died because it couldn't scale. One master can only supervise so many apprentices in a lifetime. The university solved the scaling problem by replacing the master's judgment with bureaucratic proxies. syllabi, rubrics, GPAs, standardized tests. The proxies worked as long as producing acceptable outputs required real effort. AI removed the effort requirement overnight.

But here's what changes everything. AI also makes the guild model scalable again. The master can oversee more apprentices because AI handles the routine oversight. form-checking, compliance, arithmetic, basic correctness. The master's attention gets reserved for what only the master can do: validate judgment, inspect quality, withhold sign-off.

A tax preparer with an EA credential operates inside this model. The IRS is the ultimate master. Three exams, passed under proctored conditions. Continuing education every year. Personal liability for every return. The discipline scaffold is built into the regulatory structure. You cannot AI your way through an IRS audit. You cannot ChatGPT your way past the moment a client asks you a question and you either know the answer or you don't.

The credential that survives is not the one that proves you know things. It's the one that proves you submitted to real discipline. The guild returns not because guilds were romantic. They return because discipline was always the binding constraint, and AI forces us to finally see it.

Three Kinds of Discipline

The AI labor transition is usually described as two curves: displacement and benefit. But there are actually three, and they map to three kinds of discipline.

The first kind is externally imposed discipline. The job tells you what to do, when to do it, how to do it. The structure comes from outside. Junior analysts, entry-level accountants, copywriters, customer service. these roles ran on discipline provided by the job structure. AI automates the tasks that required that discipline. The jobs dissolve. About 15 to 25 percent of the workforce lives here.

The second kind is internal discipline. The discipline to check, to care, to hold a standard. Verification work. Judgment work. The senior engineer who reviews AI-generated code. The experienced lawyer who signs off on AI-drafted briefs. The tax preparer who represents a client in an audit. This discipline cannot be automated because it requires a human to care about the outcome. About 5 to 10 percent of the workforce lives here, and the number is growing.

The third kind is physical discipline. Showing up when and where required. Being present. The plumber who arrives at 8 AM. The nurse who works the night shift. The electrician who climbs into the attic. AI can design a building but cannot wire it. AI can draft a tax memo but cannot sit across from a client. About 60 to 70 percent of the workforce lives here.

The disruption isn't happening because AI is smarter than people. It's happening because AI removes the need for the first kind of discipline. The jobs that required externally imposed structure dissolve. The people who had those jobs have to find new discipline scaffolds. either internal discipline (Curve 2) or physical discipline (Curve 3). The people who can't find either become the stuck.

If You're Trying to Figure Out What to Do

The people who win in the next decade are not the people who know the most things. Knowledge is free. Claude knows more than any human will ever know. The winners are the people who can sustain disciplined action over time without an external structure forcing them to do it. Or the people who find a new external structure. a guild, an apprenticeship, a licensed profession. that provides the discipline the old system used to provide.

The institutions that survive are not the ones with the best information. They're the ones that provide real discipline scaffolding. The elite college with oral exams, small seminars, and faculty who actually know your name. The employer program with structured progression, mentor oversight, and performance standards. The professional license with continuing education requirements, liability exposure, and a regulatory body that can revoke your right to practice.

The institutions that die are the ones that were only ever content delivery systems dressed up as discipline scaffolds. The lecture hall with 300 students and a TA doing the grading. The online degree with no human interaction. The credential that says you sat through something without any evidence you actually did the work.

If you are 18 and deciding what to do, the question is not "which school has the best program." The question is "which path will actually force me to do the work." If you are 30 and reconsidering your career, the question is not "what should I learn." The question is "what structure will keep me showing up."

The EA credential costs $627 in exam fees and four to twelve months of study. It does not ask about your SAT scores, your parents' income, or whether you rowed crew. It asks whether you can pass three exams under proctored conditions and represent a taxpayer before the IRS. That's a discipline scaffold, not a knowledge proxy. It's cheaper, faster, and more honest than a four-year degree. And it's one of the few credentials where AI cannot fake the part that matters.

*Sources: Rory Truex, Chronicle of Higher Education (May 2026). Anthropic educator Claude analysis (2025-2026). Princeton University surveys (2025-2026). Burning Glass Institute degree requirement data (2019-2024). St. Louis Fed, Serdar Ozkan (Feb 2026). BCG AI workforce projections (Apr 2026). Wall Street Journal labor market data (Q1 2023-Q1 2025).*

See which credentials actually provide real discipline scaffolding →