In early 2025, a Polish startup called Vidoc Security was interviewing candidates for a backend engineering role. Midway through one video call, the interviewer asked the candidate to do something odd: wave his hand in front of his face.

The candidate refused. Interview over.

Here's what was happening. The interviewer had noticed micro-glitches around the candidate's jawline, lighting that didn't quite track with head movements. He suspected a deepfake. Passing a hand over your face breaks the facial-landmark tracking that real-time face-swapping software relies on. The candidate knew this too. So he wouldn't do it.

This is the world now. Job interviews where you have to prove you're not a digital puppet.

A few weeks later, cybersecurity firm KnowBe4 announced they'd hired a North Korean operative. This is a company whose entire business is teaching organizations to spot fraud. The guy passed four video interviews and a full background check. He started installing malware on his first day. You can't make this up.

Gartner now predicts that by 2028, one in four job candidates globally will be fake.

Meanwhile, a Columbia student named Roy Lee built a tool to help software engineers cheat on technical interviews. He used it himself, landed offers from Amazon, Meta, and TikTok, got expelled when Columbia found out, and turned the scandal into a company called Cluely. It raised $20 million from Andreessen Horowitz. The thing that got him kicked out of school got funded by one of the most prestigious VCs in Silicon Valley. Let that sit for a second.

What's happening here isn't just "AI enables cheating." That's the surface read. What's actually happening is that AI has degraded signal reliability across the entire verification pipeline, simultaneously.

Degrees? "Did you do this work, or did AI?"

Interviews? "Is this you, or something feeding you answers?"

Take-home projects, writing samples, even in-person coding tests with people showing up wearing earpieces and live AI feeds. All of it. Compromised.

When one verification mechanism breaks, systems adapt. You route around the damage. But when all of them break at once, the system doesn't gracefully degrade. It snaps to a new equilibrium.

What we're watching is a 150-year experiment in credential-based competence signaling start to unwind in real-time. And what replaces it will look, on the surface, a lot like what preceded it: demonstrated competence over certified competence.

Universities have existed for almost a thousand years, but for most of that history, they had basically nothing to do with getting a job.

University of Al-Quarawiyyin, the first in the world, was founded in 859. The University of Bologna, generally cited as the oldest in Europe and the first in the Western world to award degrees, was founded in 1088. Oxford in 1167. These are ancient institutions. But they weren't job training centers. They existed, in the West anyway, to do three things:

First, train clergy. The Church needed administrators, canon lawyers, theologians. Pope Gregory VII issued a decree in 1079 requiring cathedral schools for exactly this purpose.

Second, produce lawyers and doctors for elites. Many early university students were actually older citizens looking to advance careers they already had, not young people preparing for careers they might have.

Third, preserve knowledge. The Church was, for centuries, the only institution in Europe with consistent interest in preserving and transmitting classical learning. Universities were essentially libraries with teaching functions attached.

You could study four subjects: law, medicine, theology, or arts. That's it. A degree in theology qualified you for administrative positions in the clergy or in the university itself. It was a closed loop.

In 1910, 3% of American young adults had college degrees. By 1940, still only 7%. For eight hundred years, universities and economic activity ran on basically parallel tracks. If you wanted to learn how to actually do something: make things, build things, trade things. You didn't go to university.

You went to a guild.

The Guild Logic

The medieval guild system operated on a hierarchy everyone understood: apprentice, then journeyman, then master. You learned by doing, under direct supervision, for years. In England, the Statute of Artificers in 1563 codified this. Apprenticeships lasted seven years, and masters could take no more than three apprentices.

Your credential was your "masterpiece," literally the piece of work that proved you'd mastered your craft. Not a certificate about work. The work itself.

This is the key thing: the signal couldn't decouple from the substance because they were the same thing. You couldn't fake a masterpiece. You couldn't buy one. You had to actually make it, and your master and the other guild members watched you make it.

Of course people tried to cheat. People always try to cheat. But cheating a system where you're observed for years is a different problem than cheating one where you mail in a transcript.

Trust ran on personal knowledge, not institutional proxies. Everyone in the guild knew everyone else. Your reputation was built over years of direct interaction, not a line on a resume that someone in a distant city would take at face value.

The Limits of the Guild System

I'm not romanticizing this. The guild system had serious problems.

It was local and slow. Seven-year apprenticeships. Geographic constraints. You couldn't move to a new city and instantly have your competence recognized. You'd have to rebuild relationships from scratch.

Guilds were also gatekeepers. They controlled entry, limited competition, extracted rents. If the local blacksmith guild decided they didn't want more blacksmiths, you weren't becoming a blacksmith, regardless of your talent. Maybe you'd insulted the guild master's son. Maybe your father had undercut his prices twenty years ago. Maybe you just had the wrong surname. Tough luck.

And critically, the guild system couldn't scale. When you need hundreds of workers doing fragmented tasks, the intimate master-apprentice model becomes economically unviable. But that's a story for the next act.

What matters here is understanding what the guild system actually was: a verification mechanism built on direct observation and long-term relationships. It worked because communities were small and mobility was low. The question "can this person do the job?" was answered by people who had watched them work for years.

That world is gone. But the problem it solved, how do you know if someone can actually do what they claim, never went away. The credential system was an attempt to solve the same problem at scale.

The credential society wasn't designed. Nobody sat down and said "let's make degrees the universal requirement for employment." It emerged from the collision of several forces, each of which made sense on its own terms.

Industrialization Broke the Guild Model

The fall of American apprenticeships began as factories needed bodies, not craftsmen. "From the introduction of the first labor-saving machine dates the decline of the apprentice," noted a Massachusetts Bureau of Statistics report in 1907. They could already see it happening.

But nobody illustrates the shift more clearly than Henry Ford.

Before 1913, building a car was essentially craft production. Skilled mechanics assembled entire vehicles, moving around the car, using judgment, adapting to variation. You needed years of training to be useful. Then Ford and his engineers at Highland Park did something that changed not just manufacturing but the entire relationship between skill and work.

They broke car assembly into 7,882 distinct operations. Then they asked: how many of these actually require skill?

Their answer was brutal. 949 operations required "strong, able-bodied men." 3,338 required "ordinary physical strength." The rest, over half, could be performed by "women or older children." And 670 could be done by legless men, 2,637 by one-legged men, 715 by armless men, 10 by blind men.

This wasn't efficiency theater. It was a systematic extraction of skill from workers and embedding it into the process itself. The knowledge that used to live in the craftsman's hands now lived in the sequence of the line, the design of the jigs, the standardization of the parts. Ford didn't need people who knew how to build cars. He needed people who could repeat a motion.

The question shifted. It was no longer "can you build this?" It was "can you be trained, show up, and repeat?"

This is why the high school diploma became the first industrial-era credential. Not because you need algebra to attach door handles. But because completing high school proved you could follow instructions, tolerate tedium, and submit to institutional authority for years at a time. The credential stopped measuring capability and started measuring trainability and reliability. It became a sorting mechanism for who could be slotted into which level of the machine.

The $5 Day in 1914 reveals the social contract underneath all this. Ford doubled wages not out of generosity but because turnover was 370% annually. The work was so soul-crushing that people kept walking off. The high wage was compensation for accepting deskilled work: give up craft autonomy, get consumer purchasing power.

The intimate master-apprentice relationship that transmitted judgment, the ability to handle variation, to know when something was "right," became economically impossible when you needed to staff a 500-person facility with interchangeable workers doing interchangeable tasks. The guild model didn't die because it was bad. It died because it couldn't scale.

Professionalization Demanded Standardization

Here's something wild: before the 20th century, "professional" didn't mean much.

In the 1800s, becoming a doctor in America required almost nothing. Many medical schools were small proprietary operations, owned by one or two doctors, unaffiliated with any university, run to turn a profit. You could get a medical degree after two years of study. Laboratory work was optional. Dissection was optional. If you could pay the tuition and sit through some lectures, you were a doctor.

The result was predictable. Quacks everywhere. Patent medicines laced with opium and cocaine sold as cure-alls. Surgeons who didn't wash their hands. Maternal mortality rates that seem medieval by modern standards.

Then came Abraham Flexner.

In 1910, the Carnegie Foundation sent Flexner, an educator with no medical training, to visit all 155 medical schools in the United States and Canada. His report was devastating. He described schools operating out of basements with no laboratories, no clinical training, no admissions standards. He named names. He published everything.

Within a decade, half of American medical schools had closed. The ones that survived affiliated with universities, added years of training, required laboratory work, demanded undergraduate degrees for admission. The profession consolidated, standardized, and credentialed.

This was originally about quality. The "anyone can be a doctor" situation was genuinely killing people. But quality control has a way of becoming gatekeeping over time. Abraham Lincoln studied law by working in local attorney offices and reading borrowed books by candlelight. That path closed. The professions used credentials to establish legitimacy first, then to control entry.

Mass Higher Education Happened Globally

The GI Bill gets the American credit, and rightly so.

When Congress was considering it in 1944, Harvard's president warned that admitting working-class veterans would turn top colleges into "educational hobo jungles." He was wrong.

The Bill funded the educations of 22,000 dentists, 67,000 doctors, 91,000 scientists, 238,000 teachers, 450,000 engineers. Three presidents. A dozen senators. Fourteen Nobel Prize winners.

But this wasn't just American. The second half of the twentieth century saw a global transition from "elite" to "mass" higher education. Japan rebuilt its university system after the war. Europe expanded access dramatically. Developing nations built universities as fast as they could pour concrete. The same basic deal everywhere: government funds education, degrees signal employability.

In 1940, 7% of American young adults had college degrees. By 2000, it was over 30%. The credential that once marked you as exceptional now marked you as normal.

Legal Incentives Made Credentials the Safe Choice

Griggs v. Duke Power (1971) is underappreciated in this story.

Willie Griggs was a Black employee at Duke Power's Dan River plant in North Carolina. Before the Civil Rights Act, Black workers were restricted to the Labor Department: the worst jobs, the lowest pay. In 1955, Duke Power started requiring high school diplomas for any other department. When the Civil Rights Act passed in 1965, instead of opening opportunity, Duke Power added IQ tests on top of the diploma requirement.

The numbers tell you everything. 34% of white males in North Carolina had high school diplomas. 18% of Black males did. On the aptitude tests, 58% of whites passed. 6% of Black applicants passed.

Here's the thing: white employees hired before these requirements existed performed their jobs just fine. The tests weren't measuring job ability. They were barriers dressed up as standards.

The Supreme Court ruled unanimously against Duke Power. Chief Justice Burger wrote that "the facts of this case demonstrate the inadequacy of broad and general testing devices, as well as the infirmity of using diplomas or degrees as fixed measures of capability."

But the ruling had an unintended consequence that shaped the next fifty years of hiring.

After Griggs, any employer who used aptitude tests had to prove they were job-related. That meant validation studies, documentation, legal exposure. Expensive. Risky. You know what wasn't legally risky? Requiring a college degree. Degrees were considered neutral. Defensible. If you got sued, "we required a bachelor's degree" was a much safer answer than "we gave them our own test."

So employers stopped testing for competence and started requiring credentials instead. Not because they believed degrees proved anything about job performance. Because degrees were legally safe.

A civil rights victory accidentally turbocharged credentialism.

And the liability logic extended beyond discrimination law. "We hired someone with the right degree" became a defense whenever things went wrong. If your new hire fails spectacularly, "they had a Stanford MBA" distributes the blame. "We gave them our own aptitude test and they scored well" does not. HR departments don't necessarily believe credentials identify competence. They believe credentials provide cover.

Status Competition Created Self-Reinforcing Demand

Once degrees became associated with middle-class status, demand became self-reinforcing regardless of whether the education was necessary for the job.

This isn't abstract theory. Watch what happens when a good job starts requiring a degree: other jobs in the same tier start requiring degrees too, because employers assume the best candidates will have them. Then jobs in the tier below start requiring degrees, because everyone above them does. The floor keeps rising. The high school diploma that got your grandfather a management job now gets you an interview at McDonalds.

America's upwardly mobile families flat-out rejected any second-tier vocational track that presumed their children shouldn't even attempt Yale or Harvard. The question wasn't "what does my kid need to learn to do this job?" It was "what credentials does my kid need to compete for status?"

This is education as signaling game, not human capital accumulation. And elite universities understood this perfectly. They're not primarily selling education. They're selling admission. Getting into Harvard signals that you could get into Harvard, which says something about raw ability regardless of what you actually learn there. The credential itself is the product, separate from what it supposedly represents.

The Golden Age: When It Actually Worked

I don't want to suggest credentials were always theater. There was a period, roughly the 1950s through the early 2000s, when the system actually worked reasonably well.

Degrees were selective enough to signal something. The curriculum was demanding enough that completion meant competence. Employers could trust the signal because the pipeline was rigorous. The correlation between credential and capability was real.

This was genuine democratization. Kids from factory towns could compete with old money. The playing field wasn't level, but it was more level than what came before. The social deal held: society invested in education, education delivered employability.

The Decay: Too Much of a Good Thing

The credential system didn't fail because it was a bad idea. It failed because it was a good idea that succeeded so completely it collapsed under its own weight.

If 7% of people have degrees, a degree tells you something. If 40% have degrees, it tells you less. So employers start requiring better degrees. Not just college, but which college. Not just bachelor's, but master's. The credential treadmill accelerates.

Just look at physical therapy: bachelor's degree in the 1980s, master's in the 1990s, doctorate today. The actual job didn't change that much. The credential requirements tripled.

A 2017 study looked at 26 million job postings from 2015 and found degree inflation affecting 6.2 million positions. Here's a stat that should bother you: 67% of production supervisor postings required a bachelor's degree, but only 16% of people actually doing that job had one. The requirement and the reality had completely decoupled.

There's another problem nobody talked about much: credentials are backward-looking. A degree tells you what someone could do when they graduated. Competence is present-tense. In a slow-moving field, this gap stays small. But the faster things change, the wider the gap between what the credential measured and what the job actually requires. A computer science degree from 2015 prepared you for a world before transformer models existed. The credential doesn't update. The job does.

Meanwhile, universities respond to demand by expanding enrollment. Expanded enrollment means less selectivity. Less selectivity weakens the signal. Which creates more demand for graduate degrees. Which creates more supply. Which dilutes further.

The ratchet only moves one direction.

The Theater

By the late stage, the system is mostly performance.

Degree requirements appear on job postings that have nothing to do with the actual work. The credential doesn't measure competence anymore. It measures ability to acquire credentials. Which correlates with family resources, time availability, willingness to play the game.

The original purpose was democratizing access based on ability. By the end, it had inverted into gatekeeping based on who could afford to play.

And everyone kind of knew it was theater, but no one could unilaterally stop. That's the collective action problem. If every other candidate has a degree and you don't, you're disadvantaged, even if the degree is meaningless for the actual job. So everyone keeps playing a game they know is broken.

The system was structurally unsound before COVID. Before AI. Hollowed out from within. Held together by inertia, legal defensibility, and the absence of alternatives.

COVID cracked the facade.

Elite institutions delivered Zoom lectures for full tuition while students sat in childhood bedrooms. Cheating became rampant with remote exams. Not because students suddenly became dishonest, but because the honor system had always relied on physical surveillance, and nobody had noticed. The "college experience" was exposed as largely social and credentialing, not educational. Parents who'd been writing checks for decades started asking uncomfortable ROI questions they'd never thought to ask before.

But the system might have survived that. Institutions are resilient. Memory is short. Tuition could have come back down, in-person instruction could have resumed, and everyone could have agreed to forget the awkward years when the emperor's nudity was briefly visible.

Then AI shattered it completely.

The thing about AI isn't just that it enables cheating. It's that it made fabrication so cheap and so good that the entire verification apparatus became unreliable at once. Degrees? "Did you learn this or did your model?" Interviews? "Is this you or are you being fed answers through an earpiece?" Take-home assignments? "Did you make this?" Writing samples? Coding tests? Research papers?

Every mechanism employers used to check if someone was actually competent was compromised simultaneously.

This is why the system is breaking rather than bending. You can route around one broken verification method. You can design new interview formats when the old ones stop working. But you can't route around all of them failing together. When every signal becomes noise, you don't get gradual degradation. You get a snap to a different equilibrium entirely.

The Canary in the Coal Mine

If you want to see where credential collapse leads, look at India.

Not because India is unusual. Because India got there first. When you have 1.4 billion people competing for a limited number of positions, credential inflation hits terminal velocity faster than anywhere else. The pressure to game the system becomes immense. And game it they did.

A single university in Rajasthan sold 43,000 fake degrees. Another in Himachal Pradesh moved 36,000. These aren't outliers. They're the ones that got caught. The University Grants Commission maintains a running list of fake institutions; it currently names 21, but everyone knows that's a fraction of the actual number. "Programmer factories" churn out graduates in six months, not because they've revolutionized education, but because the credential is the product, not the competence. Certification-sharing rings let one person take exams for dozens of others.

The result? Indian IT companies learned decades ago that degrees mean nothing. TCS, Wipro, Cognizant, Infosys, all of them. They hire in bulk through campus recruitment, yes. But then they test. They train. And they fire.

Infosys brought over 700 trainees to their Mysore campus in early 2025. These were kids who had done everything right. They passed the recruitment rounds, received offer letters, waited through the pandemic backlog. Some waited two and a half years. They relocated. They took loans. They committed based on the signal the credential sent.

Then came the internal assessment. Failure rates that had historically been under 10% jumped to 30–40%. Over 700 terminated. Then another 240 a few months later. Kids crying in Mysore, asking for one more night before they have to go home and explain to their parents what happened.

This isn't an Infosys story. Every major Indian IT company does some version of this, because they have no choice. When the credential means nothing, you have to build your own verification. You hire provisionally and then you check if the person can actually do the work.

The West is just catching up to what India figured out by necessity: the only reliable signal is demonstrated performance. Everything else is theater.

The Stranded Generation

There's a generation caught in the middle of this transition, and they have every right to be furious.

They took the AP classes. They padded the extracurriculars. They wrote the essays, took on the debt, completed the degrees. They followed a playbook that had worked for their parents, their older siblings, every authority figure who'd ever advised them. Get the credential. The credential opens doors.

Over half of recent Gen Z graduates now say their degree was a waste of money. The unemployment rate for young men with bachelor's degrees is now roughly the same as for those without. More than half of 2024 graduates are underemployed, working jobs that don't require the degree they spent four years and six figures acquiring. They're sending out 45, 60 applications a month into a void of ghost jobs and AI screening tools and "entry-level positions requiring 3+ years experience."

They're stranded between two worlds. Trained for a system that's collapsing, not equipped for what's emerging. Too credentialed for the trades, not experienced enough for the knowledge work their credentials supposedly prepared them for. The doors they were promised would open are increasingly painted on the wall.

The cruelest part? Their parents often still don't understand what happened. They're operating on 1990s assumptions in a 2026 economy, offering advice that's not just unhelpful but actively counterproductive. "Network more." "Polish your resume." "The degree will pay off eventually." They don't see that the rules changed, because the rules worked fine for them.

And the rules did work, once. That's what makes this a tragedy rather than a scam. The credential system wasn't a conspiracy. It emerged from real forces solving real problems, and it delivered real value for decades. But institutions have inertia. The playbook that worked for the parents was already obsolete by the time the kids followed it. And in a world that's changing faster than any previous generation has experienced, assuming the past predicts the future isn't just wrong. It's reckless.

The Bifurcation

Right now, we're living in the messy middle.

Startups have mostly stopped caring about credentials. When you're in survival mode, hiring someone who can't do the work is an existential threat. You don't have the luxury of legal defensibility or HR process. You need someone who can ship, and you'll figure out if they can ship by having them ship something.

The Databricks CEO put it bluntly: "I don't believe in interviews because I think some people interview well, and they might not be good at all." His approach: give candidates actual problems the company is working on right now. If you're hiring someone for marketing, work with them to fix something that's broken in marketing. See if they can do the work by having them do the work.

Palantir is recruiting high school graduates directly into software roles, skipping college entirely. "Skip the debt. Skip the indoctrination. Get the Palantir Degree." Alex Karp: "Once you come to Palantir, you're a Palantirian. No one cares about the other stuff."

Meanwhile, Fortune 500 HR departments are still posting job requirements that read like they were written in 2008. Bachelor's degree required. 3–5 years experience. They've announced skills-based hiring initiatives. Forty-five percent of employers claim to have dropped degree requirements. But Burning Glass found the truth beneath the press releases: 47% of companies that publicly removed degree requirements aren't actually hiring more non-degree workers. They changed the posting. They didn't change the hiring.

This is the bifurcation. The companies that will win are already operating on the new rules. The companies clinging to the old rules are doing theater, saying the right things while their actual hiring practices remain stuck in the credential paradigm.

The question isn't whether the transition will happen. It's how long the laggards can sustain the pretense.

What Comes Next

The pendulum swings back to where it started, but it's not a return. It's a spiral.

Before industrialization, before mass credentialing, the question was simple: "Show me what you can do." The masterpiece, the demonstrated skill, competence proven through work. We moved to "show me your papers." The credential, the proxy, the signal that was supposed to predict competence. Now we're moving back to demonstration, but the mechanism is different. Not local guild observation but global digital portfolios. Not a masterpiece evaluated by masters in your town but work visible to anyone with an internet connection.

Same principle. Different technology. Different failure modes.

Work trials are spreading. Linear, the software company, invites candidates for paid 2–5 day trials where they work on real projects with actual internal tools. The result: 96% retention over four years. It's expensive, yes. It takes time. But it answers the only question that matters: can this person do the work?

Portfolio culture is replacing resume culture, at least in fields where work can be demonstrated. Open-source contributions. Public projects. Writing that exists on the internet for anyone to read. The new credential isn't a piece of paper saying you learned something. It's evidence that you built something.

Skills assessments are becoming standard. 85% of employers now use some form of skills-based hiring. Not because they've become enlightened about equity and access, but because they've been burned too many times by candidates who interviewed well and performed poorly.

For employers, the logic is straightforward: a degree tells you someone completed coursework. It doesn't tell you whether they can manage a project, troubleshoot a system failure, or build client relationships. Skills assessments answer a different question: can this person do the work?

This is the future of hiring: provisional, demonstrated, continuous. You get in by showing you can do the work. You stay by continuing to do it. The credential is the output, not the input.

The New Failure Modes

But pendulums swing too far. They always do.

The guild system's failure mode was gatekeeping and stagnation. Masters protecting their positions. Knowledge hoarded rather than shared. Innovation suppressed. The credential system's failure mode was inflation and disconnection. Signals detached from substance, everyone running faster to stay in place.

The demonstration-based system will have its own pathologies. We can already see them emerging.

Performative building. If the new credential is public work, expect a flood of shallow projects designed to look impressive rather than be useful. The GitHub contribution graph will become the new GPA. Gamed, optimized, emptied of meaning. We'll learn to distrust portfolios the same way we learned to distrust degrees, and we'll need new ways to distinguish signal from noise.

Continuous audition. When the credential is ongoing performance rather than a one-time achievement, the pressure never lets up. At least the old system let you stop proving yourself once you had the degree. The new system demands perpetual demonstration. For some, this is liberation. For many, it will be exhausting.

New gatekeepers. The credential system's gatekeepers were universities and professional associations. The demonstration system's gatekeepers will be platforms, algorithms, hiring managers with their own biases. The barriers to entry might be lower, but they won't be absent. Someone will always be deciding who gets seen and who gets overlooked.

The experience paradox. If you need to demonstrate you can do the work to get the job, how do you get the experience to demonstrate? The old system answered this with internships and entry-level positions. The new system might create its own Catch-22: you need a portfolio to get hired, you need to be hired to build a portfolio.

The Spiral

History doesn't repeat, but it rhymes.

We started with "show me what you can do." The masterpiece, the demonstrated skill, competence proven through work. We moved to "show me your papers." The credential, the proxy, the signal that was supposed to predict competence. Now we're moving back to demonstration, but the mechanism is different. Not local guild observation but global digital portfolios. Not a masterpiece evaluated by masters in your town but work visible to anyone with an internet connection.

Same principle. Different technology. Different failure modes.

The transition will be messy and uneven. It will strand a generation trained for the old rules. It will advantage people who can demonstrate over people who can only certify. It will create new inequities even as it dissolves old ones.

But the direction is set. The credential system spent a century building itself up and a decade hollowing itself out. What comes next will be built on the rubble.

Show me what you can do.

Not what you studied. Not what you paid for. Not what some institution certified.

Show me.