We Are the Horses Now - Part 1 of 2

But We Do Not Have to Be

Two systems are collapsing at the same time.

The first is education. School is where you ostensibly learn how to think: how to hold contradictory evidence in your head, how to synthesize conflicting accounts, how to build an argument from the ground up. But in 2026, students are outsourcing that work to AI and still getting As. The struggle that builds judgment is now optional.

The second is the early-career job market. Work is where you used to sharpen thinking through apprenticeship: learning by doing, watching experienced people make decisions under pressure, accumulating repetitions until judgment becomes instinct. But those entry-level roles—the analyst jobs, the coordinators, the junior positions where you learn how organizations actually function—are disappearing.

The result is a generation that never learned to think in school entering a job market that no longer teaches thinking on the job.

The question is not whether AI will produce economic value in aggregate. It will. Productivity will rise. The economy will grow. The question is what happens at kitchen tables when that growth does not translate into paychecks. When the overall pie gets bigger but individual slices get smaller. When your job becomes “human in the loop” managing AI outputs, but you never did the actual work, so you have no idea when the AI is subtly, dangerously wrong.

⸻

In 2026, a high school junior sits down with a U.S. history assignment on the Constitutional Convention. She opens ChatGPT, pastes in the prompt, and three minutes later has 1,200 fluent words on James Madison, the Virginia Plan, and the Three-Fifths Compromise. She tweaks a few phrases, hits submit, gets an A.

What she does not do is wrestle with how the men who wrote “all men are created equal” also encoded fractions of humanity into the founding document: enslaved people counted as three-fifths of a person for representation, women could not vote at all. She does not practice holding that contradiction. She gets the answer without the struggle.

Two weeks earlier, Amazon announced another round of corporate job cuts tied explicitly to automation and AI.¹ Not warehouse workers. Program managers, business analysts, operations coordinators. Even software engineers are being cut as AI tools handle more of the routine coding work.

The student is optimizing her way out of learning. The company is optimizing its way out of teaching. Neither is building the judgment capacity the economy is about to need.

The Arithmetic of Who Gets What

After World War II, when American productivity rose by $100, workers took home about $62 in wage gains. Productivity and compensation moved together.²

That stopped in the late 1970s. American companies began moving manufacturing jobs overseas. The work still got done. Productivity kept rising. But workers in Ohio and Michigan and Pennsylvania were no longer doing it. Workers in China and Mexico and Vietnam earned a fraction of what American workers had.

Since then, when productivity rises by $100, workers here capture maybe $44. The rest flows to profits, capital income, and top earners.³

Put that in real terms. A household earning $60,000 today would be earning closer to $90,000 if productivity growth had translated into wage growth the way it did in the postwar era. That $30,000 gap went somewhere.

AI is many things, some of them incredible. But in a lot of cases, it is also just the latest low-cost labor play. Instead of moving the work overseas, we are moving it to software. Same logic. Same result.

Economic modeling suggests AI could add one to two percentage points to annual productivity growth by the mid-2030s.⁴ If current patterns hold, workers might capture only $30 of every $100 in new value created, with $70 flowing to capital. The kitchen table gets more precarious even as GDP numbers climb.

The Thing About Horses

In 1910, roughly 25 million horses powered American commerce. By 1960, fewer than three million remained.⁵ The horses did not retrain. They became obsolete. The population collapsed by 90 percent.

Are knowledge workers the horses in this story, or are we the farmhands?

When horses disappeared, farmhands moved into factories over 50 years. Workers adapted because new jobs emerged and the timeline allowed for transition.

When manufacturing jobs disappeared starting in the 1980s, factory workers did not smoothly transition into knowledge work. The Midwest and the South hollowed out. Entire towns built around factories that closed. Wages that never recovered. Communities that never recovered. Deaths of despair rose. Political extremism followed.⁶

If you strip away the ideology and culture war rhetoric, much of what we call MAGA is a howl of rage from regions that were economically abandoned. Bitterness toward those who participated in an economy where all the upside flowed elsewhere.

Knowledge workers right now are somewhere between the farmhands and the horses. Which one they turn out to be depends on whether the economy creates accessible pathways into new work, and whether the transition happens slowly enough for people to adapt.

Right now, neither condition is being met.

Two Different Views

Marc Andreessen, a well-known venture capitalist, wrote “The Techno-Optimist Manifesto.” Progress for progress’s sake. Rejecting tradition, expertise, institutions. He writes, “We are not primitives, cowering in fear of the lightning bolt. We are the apex predator; the lightning works for us.”

It is a document about power. About who gets to decide what progress means.

I run Lanai, which offers a AI @ Work platform that helps employers understand what AI is doing in their workforce. We built a system of record for AI workflows so companies can see impact and steer decisions about what to stop, start, or continue with respect to the growing shadow workforce of AI assistants and agents. This is not abstract for me. I work with these companies every day.

The operators I work with look nothing like the Silicon Valley crowd Andreessen represents. These are empathetic, courageous leaders who see the contradictions ahead. Many have kids using ChatGPT for homework. They see what is happening in education and at work.

It is not hard for them to fast forward 10 years and imagine knowledge workers looking back the way manufacturing towns looked back. They know that without deliberate cross-sector design, knowledge workers look a lot more like horses who were made obsolete than farmhands who adapted.

They understand that humans will be inventive and imagine new opportunities. But they want to participate in building a compelling path forward. Not just wait and see what emerges from the wreckage.

These are the people deciding every day whether to cut jobs or invest in retraining. And they are not interested in eating their young.

The Scale and Speed

McKinsey estimates that generative AI could automate roughly 30 percent of hours currently worked in the United States by 2030.⁷ Millions of jobs in coordination, administration, customer service, and junior analytical roles.

From 2024 through early 2026, more than 145,000 technology sector workers were laid off, with a growing share of cuts explicitly tied to automation.⁸ Amazon reduced corporate headcount while investing more than $100 billion in AI infrastructure.¹ Salesforce eliminated thousands of customer service roles after deploying AI systems.⁹ Accenture cut more than 10,000 positions.¹⁰

And, yes, even software engineers are being automated. Displacement is happening at manufacturing scale but on a compressed timeline. Five to eight years instead of fifty.

The Choice We Face

The horses had no agency. They could not negotiate the terms of their obsolescence. They could not demand that productivity gains be shared.

We can.

Right now, a CEO somewhere is looking at a spreadsheet showing that AI could eliminate 30 percent of her coordination roles and increase quarterly earnings. The math is clear. The board will approve it. Wall Street will reward it. And she knows that in 10 years, when she needs to promote someone to run the division, there will be nobody left who did the repetitions necessary to understand how the business actually works.

She has kids using ChatGPT for homework. She can fast forward this movie. She does not want to be the person who does to knowledge workers what was done to manufacturing towns.

But she does not have a clear alternative. Nobody is offering her a roadmap for how to capture AI’s productivity gains while building the next generation of leaders. And she cannot do it alone.

This is the choice: we can let that CEO make the decision that is right for her shareholders and wrong for everyone else, one company at a time, until we replicate what happened in the Midwest and the South in every suburban office park in America. Or we can organize ourselves to build something different.

That something different has a name. It is called the continuous learning economy.

Next week, I will show you exactly what it looks like as a moonshot. Not vibes. Not aspiration. A moonshot the way moonshots actually work: clear vision, explanation of why now, and the specific coalition of actors who can deliver it. Who does what. How it gets funded. How the economics pencil out.

I am, I think, a pragmatic optimist with a very active imagination and an urgency to act now while we still can. I believe we can build systems that work for hundreds of thousands of people, not just the people who own the infrastructure.

I appreciate the perspective of smart investors. They see opportunities others miss. They fund innovation that changes the world. But their perspective comes from a specific corner of the universe. One where the goal is deploying capital and maximizing returns. One very different from the universe where operators make daily decisions about people’s livelihoods.

Those operators have to look employees in the eye. They have to navigate trade-offs between short-term economics and long-term responsibility. They have to build systems that actually function for thousands of people.

Building the continuous learning economy requires their expertise. People who have struggled with those trade-offs. People who understand that you cannot run a company, or a country, without people who know how to think. People who can design systems for whole humans.

The horses could not organize themselves. We are not horses. Next week, I will show you what it looks like when we act like it.

Footnotes and Sources with Annotations

¹ Amazon job cuts tied to automation (2026): Source: CNBC – Amazon cutting hundreds of corporate roles as it automates functions with AI tools (January 2026). URL: https://www.cnbc.com/2026/01/17/amazon-ai-job-cuts.html

² Productivity–compensation link post‑WWII Source: Economic Policy Institute (EPI), The Productivity–Pay Gap dataset, updated 2025. URL: https://www.epi.org/productivity-pay-gap/

³ Worker share of productivity gains since 1979: Source: Bureau of Labor Statistics (BLS) productivity data combined with EPI wage series; summarized in Washington Center for Equitable Growth (2024). URL: https://equitablegrowth.org/research-library/productivity-and-pay-divergence/

⁴ AI‑driven productivity projections: Source: McKinsey Global Institute, Generative AI and the Future of Work (June 2024). URL: https://www.mckinsey.com/mgi/research/technology-and-innovation/generative-ai-future-of-work

⁵ Horse population decline 1910–1960: Source: U.S. Department of Agriculture, Historical Statistics of the United States: Colonial Times to 1970, Series K 476‑481. URL: https://fraser.stlouisfed.org/title/historical-statistics-united-states-colonial-times-1970-45 Annotation: Provides census counts of horses and mules showing collapse from ~25 million (1910) to ~3 million (1960). Anchors the “horses vanished” metaphor in data.

⁶ Manufacturing decline and deaths of despair: Source: Anne Case & Angus Deaton, Deaths of Despair and the Future of Capitalism (Princeton UP, 2020). URL: https://press.princeton.edu/books/hardcover/9780691190785/deaths-of-despair-and-the-future-of-capitalism. Annotation: Ties regional economic collapse to rising mortality and polarization. Supplies factual base for the Rust Belt–MAGA argument.

⁷ 30 % of hours automatable by 2030: Source: McKinsey Global Institute, Generative AI and the Productivity Frontier (December 2023). URL: https://www.mckinsey.com/mgi/research/technology-and-innovation/generative-ai-productivity-frontier

⁸ Tech‑sector layoffs 2024–2026: Source: Layoffs.fyi aggregated dataset (accessed  February 2026). URL: https://layoffs.fyi/ Annotation: Tallies 145,000 U.S. tech‑sector layoffs since 2024; used for time‑bounded scale reference.

⁹ Salesforce AI‑related cuts: Source: Wall Street Journal – Salesforce to Lay Off Thousands as AI Automates Support Roles (December 2025). URL: https://www.wsj.com/articles/salesforce-ai-job-cuts-2025

¹⁰ Accenture workforce reduction: Source: Reuters – Accenture to Cut 10,000 Jobs as It Doubles Down on AI Services (October 2025). URL: https://www.reuters.com/technology/accenture-job-cuts-2025-ai-2025-10-11/