Episode #183 | Published January 23, 2026 | Duration: 37:05

A response to “Capital in the 22nd Century” by Philip Trammell and Dwarkesh Patel, a paper arguing that AI will create permanent techno-feudalism unless we implement aggressive global taxation. The paper is serious, internally consistent, and built on a foundation that is profoundly wrong.

Philip Trammell and Dwarkesh Patel published a paper arguing that artificial intelligence will create wealth inequality so extreme it makes the Gilded Age look like a socialist utopia, and that without aggressive global taxation, we are heading toward permanent techno-feudalism.

The paper is serious. The authors are serious. The logic, given their starting assumptions, is internally consistent. The starting assumptions are profoundly wrong.

Understanding why smart people believe wrong things is one of the most valuable exercises available. Nothing builds conviction like dismantling a sophisticated argument at its foundation.

Imagine a boat taking on water in the middle of the ocean. Someone writes an elegant paper analyzing the water distribution between compartments. They measure flow rates. They model which passengers will drown first and which will stay dry longer.

Their solution: redistribute the water more evenly among the passengers. That way everyone drowns at the same rate.

Boats don’t float because water is distributed fairly. Boats float because someone understood buoyancy and hull design. The leak is a knowledge problem, not a distribution problem. Moving water between compartments fixes nothing.

Wealth works the same way.

The paper builds on Piketty’s framework: r > g, meaning the rate of return on capital tends to exceed the growth rate of the economy. The rich get richer because that is what capital does.

Trammell and Patel extend this: what happens when AI substitutes for all human labor? Labor’s share of income goes to zero. All returns flow to capital owners. If r > g holds, those owners compound their advantage forever. Permanent dynasties. Game over.

Their solution: wealth taxes, inheritance taxes, global coordination to prevent capital flight.

The paper treats wealth as a stock of stuff that exists independently of knowledge. Capital is factories, machines, servers, land. It has a rate of return because that is what capital does.

But capital is not a thing. Capital is embodied knowledge. A factory is valuable because it embodies knowledge about manufacturing. Without that knowledge, it is a pile of metal. The return on capital is really the return on the knowledge embedded in it.

Knowledge does not compound automatically. It has to keep being true. It has to survive criticism. It has to remain relevant as the world changes.

The railroad barons of the 19th century owned enormous capital. By r > g logic, their descendants should own everything by now. So why don’t the Vanderbilt heirs own Apple? Because you cannot buy knowledge before it exists. Railroad knowledge had nothing to do with semiconductor physics. New knowledge created new winners.

They assume AI will end disruption. But disruption is new knowledge making old knowledge obsolete. If AI is smart enough to replace all labor, it is smart enough to create new knowledge. New knowledge disrupts its owners’ arrangements as readily as anyone else’s.

They assume the future is predictable. Knowledge creation is not predictable. If we could predict what we will know in 2100, we would already know it. That is a logical impossibility.

They assume redistribution solves the problem. Redistribution does not create knowledge. It requires institutions that work. If those institutions are degrading, redistribution becomes another arena for corruption.

They focus on relative inequality over absolute capability. The poorest person today has capabilities the richest person in 1800 could not buy at any price. If knowledge creation continues, the floor keeps rising. The pathological case is not inequality. It is absolute deprivation.

The real problem is whether we maintain the conditions for knowledge creation. Everything else, including inequality, is second order.

Knowledge creation requires systems where truth is easier to verify than to fake. Where errors get caught and corrected. Where new knowledge can emerge from anywhere and displace old knowledge that stopped working.

When constraint quality is high, knowledge accumulates and the ability to transform the physical world grows. When it is low, information floods in but nothing sticks.

Every time a paper like this gets taken seriously, the gap between people who understand and people who do not gets wider. The confusion is the arbitrage.

There exists a form of capital that anyone can own. It requires no permission. It cannot be debased. It cannot be easily confiscated. Verification is cheap and falsification is prohibitively expensive. High constraint quality by design.

The question is not who owns the stuff. The question is whether we are still solving problems. If we get that right, the distributional questions become irrelevant. The pie keeps growing. New knowledge keeps disrupting old arrangements so that dynasties cannot persist.

If we get it wrong, no amount of clever policy will save us. We will be redistributing the water on a sinking boat.

01:42 – Why read papers you disagree with

05:35 – Summary of the paper’s claims

05:46 – Piketty’s r > g framework

06:29 – AI substitution and permanent dynasties

07:43 – The proposed solution: global taxation

09:40 – The sinking boat analogy

12:12 – Wealth is knowledge, not stuff

13:48 – The replication crisis in science

16:31 – The fundamental error: wealth as stuff vs knowledge

17:40 – Railroad barons: why dynasties don’t persist

20:24 – Error 1: AI will end disruption

21:55 – Error 2: The future is predictable

23:01 – Error 3: Redistribution solves everything

24:37 – Error 4: Relative vs absolute inequality

26:06 – The real problem: constraint quality

30:04 – Where the alpha lives

30:49 – Bitcoin as high constraint quality capital

34:35 – Conclusion: an epistemological crisis