Every generation believes it has discovered the technology that will finally break capitalism:

• The loom was going to destroy labor.
• Electricity was going to create mass unemployment.
• The assembly line was going to eliminate human relevance.
• Computers were going to wipe out the middle class.
• The internet was going to hollow out the economy.

Now AI is supposed to trigger a severe economic shock, displacing white-collar workers so quickly that demand falters and markets convulse.

The concern isn’t absurd. But history suggests it is incomplete.

Productivity and Prosperity Move Together

For over 200 years, every major productivity shock has increased living standards, not destroyed them:

• U.S. real GDP per capita has increased roughly 8–10x since 1820.
• Real hourly compensation has broadly tracked productivity over long horizons.
• Average annual hours worked have fallen dramatically since 1900.

Agriculture once employed ~40% of the U.S. workforce. Today it employs under 2%.
Manufacturing employment peaked mid-20th century and declined as automation improved. Clerical work has been systematically automated over the past four decades.

And yet:

• GDP per capita rose.
• Real consumption rose.
• Life expectancy rose.
• Leisure time increased.

The pattern is not subtle:

Productivity ↑ → Costs ↓ → Purchasing Power ↑ → Demand ↑ → New Sectors Emerge

To argue that AI will permanently collapse demand is to argue that this time productivity gains will not lower prices, will not expand purchasing power, and will not lead to new industry formation.

That is not a small claim. It is a radical one.

Industrial revolutions do not flatten the curve. They steepen it:

• Steam power.
• Electricity.
• Mass production.
• Computing.
• The internet.

Each wave accelerated output per person.

AI is far more likely to be another inflection point than a reversal.

Displacement Is Real. Collapse Is Not.

Technological revolutions eliminate jobs:

• They eliminate tasks.
• They compress categories.
• They hurt specific regions.
• They create inequality spikes.

The Industrial Revolution displaced artisans. Globalization displaced manufacturing hubs.
Software displaced clerical workers. Short-term dislocation is real.

But displacement is not destruction.

Human labor reallocates. Capital reallocates. Entirely new demand appears in sectors that did not previously exist:

• In 1900, no one worked in software.
• In 1950, no one worked in digital advertising.
• In 1990, no one worked in cloud infrastructure.

If in 2000 someone had told you that by 2026:

• Bank tellers would largely disappear.
• Travel agents would collapse.
• A trillion dollars of offline retail would migrate online.
• Car manufacturing would become highly automated.

You would likely have predicted mass unemployment. Instead, GDP per capita roughly doubled. Employment rose. Entire new sectors emerged.

Sectoral collapse does not automatically translate into macro collapse.

The Transition May Be Violent, But It Is Rarely Instant

The strongest objection to the productivity thesis is not permanent collapse.

It is speed:

• Technology moves fast.
• Markets move faster.
• Institutions move slower.
• Labor moves slowest.

That gap can produce real turbulence.

Financial markets price the future instantly and often overshoot in both directions. Expectations compound. Narratives cascade. Capital reallocates before the real economy has time to adjust. Governments respond reactively. Workers cannot retrain overnight.

That mismatch can absolutely produce ugly quarters, even ugly years. However, technological capability is not the same as economic replacement. We have seen this before.

Fifteen years ago, self-driving trucks were supposed to eliminate one of the largest job categories in America. Truck driving was the most common job in many states. The technology progressed rapidly. Investors extrapolated. Commentators predicted structural unemployment.

Today, autonomous systems exist, but long-haul trucking remains largely intact. Regulation, liability, insurance, infrastructure, edge cases, and economics slow real-world diffusion dramatically.

The same pattern holds more broadly. Even transformative technologies follow diffusion curves. Electricity, refrigeration, telephones, computers, and the internet did not saturate households overnight. Adoption followed S-curves stretching over years if not decades, not quarters.

Every technological revolution feels instantaneous while it is happening. The data shows it is not.

AI is no exception.

AI’s capabilities are real and improving rapidly. Models are increasingly competent at coding, reasoning, multimodal tasks, research assistance, and workflow automation. The technology is not a toy. It is already meaningfully increasing productivity in specific domains, and it will likely become far more powerful over the next decade.

But capability growth and economic saturation are different phenomena. A tool can be extraordinary and still take years to fully diffuse through institutions, regulation, labor markets, and global infrastructure.

Despite the intensity of the narrative:

• Billions of people globally have never used an AI system.
• A minority use free chatbots.
• Only a small fraction pay for AI tools.
• An even smaller fraction rely on AI as a core coding scaffold.

AI feels saturated inside tech and financial circles. At global scale, it is still early.

Markets extrapolate instantly. Diffusion unfolds gradually.

That gap creates volatility. It does not automatically create collapse.

None of this means displacement will be painless. Certain roles may compress quickly. Certain asset prices may re-rate violently. Certain regions may suffer. The dispersion between technological change and policy response may widen before it narrows.

But history suggests two moderating forces:

• First, adoption takes longer than headlines imply.
• Second, labor reallocates rather than disappearing.

The danger is not that AI eliminates work overnight.
The danger is that markets price elimination faster than economies can adapt.

Those are very different risks.

What Jobs Will People Do?

common objection is practical: displaced white-collar workers are not going to become plumbers, carpenters, or massage therapists. That’s true. Historically, displaced workers do not simply shift into existing blue-collar roles.

They move into categories that did not exist before.

• In 1995, “social media manager” was not a job.
• In 2005, “app developer” barely existed.
• In 2010, “cloud architect” was niche.

Technological revolutions expand the adjacent possible. They create new layers of coordination, services, tools, and industries that are invisible beforehand.

The discomfort comes from not knowing what those jobs will be yet.

But that uncertainty has accompanied every major shift in history.

The Real Risk: Transition Friction

None of this minimizes the turbulence.

Every productivity shock creates:

• Temporary inequality spikes
• Geographic concentration of gains
• Skill mismatches
• Political backlash
• Social instability

The winners and losers are rarely the same people.

The dispersion between technological change and policy response may indeed be widening. Financial markets may indeed amplify both optimism and panic.

These are legitimate concerns. However, they are concerns about transition dynamics, not permanent economic collapse.

Historically, institutions adapt:

• Education systems expand.
• Labor protections evolve.
• Competitive markets transmit productivity gains into lower prices.
• Capital reallocates into new sectors.

Adjustment is uneven, but it happens.

To assume permanent collapse is to assume permanent institutional paralysis.

That is possible. It is not the historical base case.

AI Reduces Cognitive Friction

AI is not merely automation.

It reduces the cognitive cost of doing almost anything:

• Starting a company.
• Writing code.
• Conducting research.
• Launching globally.
• Serving customers.
• Translating across languages.
• Making complex decisions.

Lower friction expands markets:

• When entrepreneurship becomes easier, more firms form.
• When coordination costs fall, markets expand.
• When information asymmetry shrinks, capital allocates more efficiently.

That is expansion logic, not collapse logic.

The Thesis

For a true economic collapse to occur, we must believe:

• Productivity gains will not lower prices.
• Purchasing power will not expand.
• New sectors will not emerge.
• Labor will not adapt.
• Institutions will not evolve.
• Competitive markets will fail to transmit gains.

History suggests the opposite. The more plausible future is not systemic collapse.

It is a volatile but powerful productivity acceleration:

• There will be dislocation.
• There will be inequality spikes.
• There will be political noise.
• There may be brutal market cycles.

However, over time, productivity increases tend to expand output, raise living standards, and increase human optionality.

AI is not the end of economic progress. It is the next chapter.

I’ll be exploring the specific implications for marketplaces in the next episode of Playing with Unicorns. The macro conclusion is the same: the opportunity lies in understanding how AI expands the economic pie, not in assuming it destroys it.

We have seen this movie before. The ending has never been collapse.

It has been transformation. It has been expansion. And most often, it has been acceleration.