The Distillation of Intelligence
The Distillation of Intelligence
Meet the Experts Earning $150 an Hour to Train the AI That Will Replace Them
The Rise of the Mercor Economy—and Its Cruel Paradox
Headlines about AI layoffs have been loud and impossible to miss. But quietly, underneath all that noise, something stranger is taking shape: a polished, well-paid "shadow economy" of expert knowledge brokers. In April 2026, Goldman Sachs estimated that AI had already wiped out roughly 16,000 net U.S. jobs over the previous year—and 72% of those jobs belonged to white-collar professionals: lawyers, financial analysts, writers, and managers.
Yet in the middle of that same collapse, a startup called Mercor has emerged as one of the biggest bridges for turning human expertise into machine-readable code. By mobilizing roughly 30,000 elite professionals, Mercor is powering an unprecedented transfer of expert knowledge—from human minds into AI systems.
Who are these elite professionals? PhDs from top U.S. universities. Doctors. Lawyers. Financial experts. Senior software engineers. In other words: almost exactly the people whose jobs are disappearing.
The new gig looks simple enough. Work from home, on your own schedule, typing out your professional thought process at your computer. Pay ranges from $45 to over $250 an hour, averaging around $108—with some contracts, often for U.S.-based business owners or senior software engineers, reaching $500 an hour. On paper, it looks like one of the smartest, best-paid remote jobs around.
But this job is hiding a brutal paradox.
That generous paycheck? It's payment for training data—data being used to build a highly skilled AI agent that will eventually replace the very expert being paid to train it.
This article lays out the facts behind the fast-growing "Mercor economy" and then asks a harder question: what happens to society once knowledge and judgment have been fully transferred to machines?
1. The Rise of an Industry That Hunts for Experts
Founded in San Francisco in 2023, Mercor—along with rivals like Surge AI and Outlier—is now one of the fastest-growing forces in the AI industry.
In one sentence, what they do is this: recruit elite professionals, then build the infrastructure to transplant their intelligence into AI.
Frontier AI labs like OpenAI, Anthropic, and Google are running into a serious wall: a shortage of high-quality data. The internet's general text has already been scraped dry.
To make AI genuinely expert-level—not just competent—these labs need something rarer: the deep reasoning of real professionals. A lawyer's legal interpretation. A doctor's diagnostic process. A researcher's careful reading of a paper. This is the raw material of reinforcement learning from human feedback (RLHF) and chain-of-thought data—and it can't simply be scraped from the web. Someone has to produce it, live, in real time.
Mercor built its business on exactly this gap. Using its own AI-powered interview system, it rapidly screens and pools elite talent from around the world, then connects them to AI labs—OpenAI, Anthropic, Google DeepMind, Meta AI, xAI, and others.
The work itself: experts review AI-generated output in their field, evaluate it for accuracy and quality, and—crucially—explain why they'd reach a different conclusion. That reasoning, that "how I think through this," is exactly what gets fed back into the model.
For professionals, this looks like the smartest gold rush of the moment: short hours, serious pay.
And on an individual level, it's a completely rational choice. If a nurse or a senior financial analyst is staring down a brutal job market, taking $150 an hour to put decades of hard-won skill to use is simply good financial sense.
Mercor's CEO frames this as an entirely new category of work: a limited-time window—he's floated roughly ten years, though that timeline keeps shifting—during which millions of professionals get paid to teach machines the judgment they currently hold. Mercor, in other words, has created a labor market where humans train the AI that will one day do their job.
Other companies running similar operations:
Scale AI / Outlier
Surge AI
Invisible Technologies
Turing / DataAnnotation.tech
Handshake AI
A word of caution: Mercor's rapid rise and the buzz around "$150 an hour" have already attracted fake websites and phishing scams. Be careful.
2. "Distilling Intelligence"—and the Road to Fully Autonomous AI
What these experts are really providing isn't fact-checking or tagging. It's something deeper: the instincts built from years of experience and the logical steps behind every judgment call.
The industry has a name for this: distillation.
It's the process of pouring the complex know-how locked inside a human mind into an AI model—turning it into reproducible software. Right now, every major AI lab is hungry for exactly this kind of expert data.
Anthropic's own framework, Constitutional AI, points toward where this is heading: a model that no longer needs a human evaluator at all, judging its own performance instead against a built-in set of principles.
Here's the catch: the more data these $150- and $250-an-hour experts hand over, the closer AI gets to doing their job on its own—and the closer labs get to finishing that Constitutional AI vision.
Which means this shift, from "AI trained by human evaluators" to "AI that evaluates itself," acts as a kill switch for the entire labor market that's currently paying these experts so well. Once a model can judge and correct its own professional reasoning, the demand for human experts in that market disappears—not maybe, but necessarily. The direction is clear. Only the timeline is up for debate. That $250-an-hour rate is only available while the bridge is still standing.
Which brings us to the paradox at the heart of this whole system—a cold, macroeconomic irony: the better and faster an expert does this job, the faster their own profession's market value erodes.
In other words, these professionals are being paid handsomely, right now, to help commoditize—to help erase the value of—the very expertise they're selling. And that's simply what's happening, as a matter of fact: decades of hard-earned expertise and analytical skill are being extracted as data and rebuilt into infrastructure that anyone can use, cheaply, on demand.
3. A Reflection: What Will Humans Value Once Knowledge Itself Has Been Distilled?
We are living through a quiet but historic moment: the systematic codification of human expertise and its handover to machines.
The highly paid experts teaching AI their final layer of judgment are, in a sense, the last-mile delivery drivers of automation in complex, specialized fields. This isn't a sudden layoff. It's something stranger—a deliberate, well-funded, quiet dismantling, built specifically to teach machines the subtle art of human judgment.
As that last mile toward Constitutional AI closes, a fundamental question remains for the global workforce: once human judgment has been fully systematized and commoditized, what unique value is left for the expert?
When logical thinking and specialized knowledge—the very things society once placed at the top of the ladder—become available from AI at close to zero cost, where does human worth go from here?
What follows are a few ways of thinking about that question—not answers, exactly, but starting points.
① From the Economy of Reason to the Economy of Emotion
If AI comes to handle correct answers and logical reasoning flawlessly, then the things that resist efficiency—the irrational, the un-optimizable—may become far more valuable than they are today:
Picking up on the faint nuance or mood behind someone's words
Simply sitting beside someone else's pain, even when it can't be put into language
Holding onto a feeling or a conflict that doesn't resolve neatly
As the "knowledge economy"—where knowledge itself carried economic value—comes to an end, society's center of gravity may shift toward something rawer and more human: an Emotional Economy.
② After the Breakdown Comes the Self
Alchemy and psychology share a concept called nigredo—the "blackening," a stage where the old self breaks down into confusion and chaos—followed by rubedo, the "reddening," where a truer self finally emerges.
Watching your own expertise and titles get replaced by AI can feel exactly like this: a collective nigredo of intelligence itself.
Think of everything that's gone into it. Studying through the night for certification exams so you wouldn't fall behind your colleagues. Grinding through entrance exams years earlier to get into a prestigious school. Years spent as a ronin, chasing admission to a famous university. Ten or twenty million yen spent on an MBA from some globally competitive program. Five or six years preparing for the bar exam. Graphic design, software engineering, medical diagnosis, finance, and accounting—decades of accumulated human effort, in field after field, now on the verge of being handed over to AI.
Soon, humans may no longer need to prove any of it. And right now, in this brief, rainbow-lit window, people are getting paid well to hand over their hard-won intelligence—to help build their own replacement.
The identity you built on top of your intelligence is about to come apart.
So what's our survival strategy?
Think about the person who was once Japan's best at the abacus. When the calculator arrived, they probably set the abacus down and found something new to master. The tool became obsolete—but the skill of the person using it never disappeared.
AI has now absorbed nearly every pattern from the average of the past. Even our conversations with AI right now are being distilled into more data. What will our relationship to that distilled intelligence look like going forward?
But there's one thing AI can never distill: the raw, lived texture of being human. What does it mean to keep living a genuinely personal life alongside AI—to keep pursuing individuation, that one-of-a-kind self built from both light and shadow?
For those who've spent a lifetime building intellectual skill, maybe the answer isn't "AI is taking this from me." Maybe it's: let AI take over the purely intellectual part—and let that free you to think more widely, more deeply, and more strangely, and go looking for a talent in yourself you never even imagined you had.
③ Empathy as the Antenna of a New Era
In a world where AI handles data processing and logical construction, the kind of high-sensitivity empathy that used to get dismissed as "too sensitive" or "inefficient" may turn out to be something else entirely: a crucial antenna, tuned to catch the earliest tremors of human unconscious life—and of the times themselves—before anyone else notices.
4. Looking at the Economics: How Will the Structure Actually Change?
Once autonomous AI—constitutional models, advanced agents—has fully absorbed this distilled intelligence, a real question remains: how will people actually pay rent, keep a roof over their heads, and raise children?
Across the internet, a common thread keeps showing up: the 20th-century model of employment—sell your intelligence and skills to a company in exchange for a paycheck—is breaking down. Both the social system and the shape of individual work are being redefined from the ground up.
At first glance, that sounds terrifying. But it's worth remembering: before Japan's era of rapid economic growth, most people ran small businesses, farmed, or worked as craftspeople. It was only during that boom—roughly the mid-1950s through the early 1970s—that Japan shifted, almost overnight by historical standards, from a nation of self-employed workers and farmers to a nation of salarymen. That, too, was a massive structural transformation. And society survived it, then thrived.
So a shift in economic structure isn't, on its own, something to fear.
Still—this time is different. This time it's intellectual labor itself being replaced by AI, which really is unprecedented for our species. It's uncharted territory. Then again, the future has always been uncharted.
Here are some of the predictions circulating—not certainties, but a way to prepare together for what's coming.
What the Future Social and Economic System Might Look Like
As professional labor loses value, individual effort alone won't be enough to support a household or raise children. Three kinds of social infrastructure are being seriously discussed:
Universal Basic Income (UBI) and data dividends. Funded by taxes on AI companies
and large-scale automated production—a "robot tax" or "AI tax"—this would give every citizen unconditional basic income. There's also growing discussion of a further dividend: compensation for the accumulated human knowledge and expert training data that made these AI systems possible in the first place.
Falling costs of living (a near-zero marginal cost society). As AI and robotics fully automate construction, agriculture, logistics, and energy, the baseline cost of housing, food, healthcare, and education could drop sharply—meaning people would need far less cash to sustain a household than they do today.
What Work Itself Might Look Like—and the Rise of the Solopreneur
Once basic survival is largely underwritten by social infrastructure, the work people do for extra income—or for meaning—is likely to shift toward hyper-personal micro-businesses.
The one-person company becomes normal. Work that once required dozens or hundreds of employees—app development, content production, small-scale marketing, specialized consulting—becomes something a single person can run, commanding a team of AI agents as their "staff." A one-person business earning enough to support a family stops being unusual.
Physical, local, hands-on work rises in value. Precisely because screen-based knowledge work gets automated, in-person service and craftsmanship become more valuable, not less. Home maintenance. In-person childcare, education, and elder care. Local farming. Direct community-building. Anything requiring genuine physical presence or emotional care becomes a serious income source.
A community economy built on trust, empathy, and story. Rather than paying for AI's flawless, uniform output, people start paying for something else: knowing who made a thing and why—what story and what feeling went into it. Small, devoted communities built around a specific maker's craft or experience become a real economic engine.
A Realistic Picture: Keeping a Home and Raising Kids
Rather than the old model—tied to one company, one fixed paycheck, nine-to-five—a hybrid model is emerging:
Basic living costs: covered by social income (UBI or dividends) plus dramatically lower infrastructure costs
Education, housing, and everything beyond survival: covered by a one-person AI-powered business, plus income from local, in-person, empathy-driven work
As labor shifts from forced survival to self-expression and participation, people may find they have more time again for parenting and daily life—and the sharp line between "work" and "life" may simply start to dissolve.
In Closing
The rise of the Mercor economy is quietly overturning something we've long taken for granted: the very definition of what knowledge and intelligence are worth.
In a world where your own expertise gets distilled into AI, and every form of logic becomes automated—where, exactly, does the value that only you can create still live?
As the era of competing on knowledge draws to a close, we may be standing at the threshold of a deeper question: what does it actually mean to be human?
My hope is this: that we can meet this new economy not with fear, but with something closer to curiosity—that each of us finds a talent we didn't know we had and discovers just how much one person, alone, can still do.
Like a surfer who's learned to love the wave rather than fear it, I hope we can ride this turbulent chapter together.
Comments