seeing the same pattern everywhere I looked. Layoffs in San Francisco. Agencies folding in Hong Kong. Friends who used to bill $200 an hour now competing with $20 AI subscriptions.
And I kept thinking: we've seen this before. Not the technology—the economics. The moment when a whole class of people wakes up and realizes the skills they traded their youth for are suddenly worth less than the electricity running the server.
We just dress it up differently now. We call it "AI transformation" instead of "structural collapse." We tell each other it's a golden age of creativity, not a violent reorganization of who gets to eat.
But history doesn't care about our press releases. And if you want to survive the next three years, you need to stop listening to the utopian keynote speeches and start looking at what actually happens when empires suddenly don't need their soldiers anymore.
The Soldiers With No War
Everyone knows Qin Shi Huang for the Terracotta Army and the Great Wall. The textbooks paint him as a madman—burning books, burying scholars, working people to death for ego.
But look at the economics instead of the morality play, and you see something else entirely. Qin inherited a machine built by his predecessor Shang Yang: a ruthlessly meritocratic military state where peasants could climb to nobility by killing enemies. For decades, war was the only path out of poverty. Millions of people optimized their entire lives around combat skills.
Then in 221 BC, Qin conquered everyone. China was unified. The war was over.
And suddenly the emperor had millions of highly trained, heavily armed professional soldiers with exactly zero peacetime skills. Send them home to farm? They'd become private militias for local warlords and the empire would fracture within a year. Keep them employed? Doing what?
So he built the Great Wall. The Epang Palace. A national highway system stretching thousands of miles. Massive, non-revenue-generating public works projects that consumed the energy of an obsolete military class.
The wall wasn't just defense against nomads. It was a jobs program for people whose only marketable skill was now illegal in peacetime.
The Modern Parallel Nobody Wants to Name
For forty years, our Shang Yang reform was the university system. Study hard. Learn Excel. Master a programming language. Write standard marketing copy. Pass the CPA exam. Do these standardized things, and you'll be rewarded with a salary, a mortgage, and a place in the middle class.
AI just ended the war.
The skills that took millions of people years to acquire—junior legal review, mid-level coding, data analysis, standard copywriting, financial modeling—are now instant, cheap, and infinitely scalable. The soldiers are standing around in a unified empire, holding weapons that nobody needs anymore.
And just like Qin, our modern emperors are proposing modern Great Wall projects to absorb the excess. Universal Basic Income. Mandatory "human-in-the-loop" regulations. Government retraining programs for jobs that won't exist by the time the certificates print.
I don't think these are evil ideas. I think they're Qin economics—ways to manage obsolescence rather than solve it. And when the system that funds them runs out of money, the wall stops getting built.
The Liberation That Wasn't
The American Civil War is supposed to be a simple story: North fights South, slavery ends, freedom wins.
But look at the economics underneath. The North was industrializing. They wanted protectionist tariffs to force Americans to buy expensive local goods instead of cheap British imports. The South was export-driven, built on cotton. The tariffs gutted them—essentially using Southern wealth to build Northern railroads.
And here's the part they don't teach you: slavery was already dying a slow mathematical death. The mechanical reaper and industrial tools were making human field labor economically obsolete. The war just accelerated the collapse.
When the slaves were "freed," what actually happened? They had no capital. No land. No seeds. No tools. So they went back to the same plantations, borrowed money for equipment from the same white landowners, and split the harvest. They were technically free. Economically, they were trapped in perpetual debt.
They called it sharecropping. It was slavery with a different contract.
The Digital Plantation
I hear the same language now from AI companies. "Liberation from drudgery!" "Be a creative solo-entrepreneur!" "Let AI handle the mundane so you can focus on what matters!"
But if you don't own the foundation, you're not free. You're just on a different plantation.
The new landlords are Google, Microsoft, OpenAI, Anthropic. The new "land and tools" are the foundational models, the compute clusters, the cloud infrastructure, the API endpoints. If your entire business is renting their APIs to execute freelance tasks—if you're a "ChatGPT prompt engineer" or an "AI-enhanced agency" running on someone else's infrastructure—you're not an entrepreneur. You're a digital sharecropper.
They'll raise API prices. They'll change terms of service. They'll introduce platform taxes and token limits and subscription tiers that slowly compress your margins to exactly zero. You're free from the 9-to-5, but you're economically chained to the infrastructure you don't control.
I've watched it happen to friends already. Agencies that pivoted to "AI services" built entirely on OpenAI's API. Six months later, their unit economics evaporated when token costs shifted. They weren't running businesses. They were sharecropping on digital land that wasn't theirs.
How to Actually Survive
I'm not writing this to be depressing. I'm writing it because I think there's a narrow window to make the right moves before the window closes. Here's what I'm actually doing at Mercury, and what I'd tell anyone I care about:
Stop Selling Time. Start Selling Judgment.
If you're billing hours for execution—writing copy, managing ads, coding features, crunching data—you're in the soldier class. The work you're selling is being automated in real time.
We stopped being an agency because agencies sell execution. We now sell what we call the Judgment Layer: deciding which problems are worth solving, vetting AI outputs against business reality, owning the relationship and accountability that algorithms can't hold.
AI is your infinite army of interns. Your job is to be the general who points them at the right hill and takes responsibility when they charge the wrong one.
Own Something That Can't Be Rented
You can't own GPT-5. You can't own Azure. So what capital can you accumulate?
Proprietary data. Verified authority. Un-gameable truth.
The internet is currently drowning in synthetic garbage—AI-generated articles, fake reviews, hallucinated citations. In that flood, the scarce resource is trustworthy signal. Real case studies. Actual telemetry. Genuine relationships with journalists and analysts. First-party data that models must cite to remain accurate.
Don't spend your time trying to trick AI crawlers with SEO hacks. Build the thing that the AI needs to reference. Become the source that the machine can't afford to ignore. That's your land. That's your capital.
Live Where the Rules Are Fuzzy
AI dominates "low entropy" environments—clear rules, structured data, predictable outputs. Coding. Accounting. Standard analysis. Anything with a right answer.
You need to move into "high entropy" spaces: strategic ambiguity, complex human relationships, B2B trust negotiations, ethical judgment calls, creative direction that requires taste.
An AI will always give you the safest, most mathematically average answer. It will suggest the strategy that worked for the median case in its training data. But breakthrough business success comes from seeing the anomaly that the data missed—the edge case, the cultural shift, the emotional undercurrent that statistics can't capture.
That's where humans still win. Not because we're smarter, but because we're weirder. Because we have biases and scars and irrational convictions that occasionally align with reality in ways the average can't predict.
The Honest Truth
I don't think 2026 is going to be a golden age for everyone. I think it's going to be ruthless. I think a lot of people who did everything right—who studied hard, who got the degree, who played by the rules—are going to find themselves holding skills that the market no longer values.
History doesn't repeat, but it rhymes. The Qin soldiers didn't fail because they were lazy. The sharecroppers didn't fail because they were stupid. They failed because the economic architecture shifted underneath them, and the skills they'd optimized for were suddenly written out of the equation.
The question isn't whether AI is good or bad. The question is whether you're building on land you own, or sharecropping on someone else's digital plantation.
Choose now. The walls are going up either way.
— James, Mercury Technology Solutions, Tokyo, May 2026
