Tyler Cowen is an economist, educator, intellectual, and fellow member of the commentariat (though plenty more eloquent and prolific than me). He's out with a new book... which he describes as "Not quite a book"

It has four chapters, is about 40,000 words, is fully written by me (not a word from the AIs), and it is attached to an AI with a dual page display, in this case Claude.  Think of it as a non-fiction novella of sorts, you can access it here.  You can read it on the screen, turn it into a pdf (and upload into your own AI), send it to your Kindle, or discuss it with Claude.

This is a 4-part series — corresponding to the four parts of Cowen's "book" — where I share some extracts and comments from the book as I'm making my way through it. You can read Chapter 3 here.

The full book, The Marginal Revolution: Rise and Decline, and the Pending AI Revolution, is available here:
https://tylercowen.com/wp-content/uploads/2026/03/TheMarginalRevolution-Tyler_Cowen.pdf

(giving it away openly, like Max Hillebrand did for Praxeology of Privacy is very cypherpunky... and also fits exactly with the times, where the value of generated written art or content drops to zero... so creators might as well give it away. I'm on this theme all the time, #796401, #1467035)

Chapter 4: Why Marginalism Will Dwindle, and What Will Replace It?Chapter 4: Why Marginalism Will Dwindle, and What Will Replace It?

This last chapter is, probably, why Cowen embarked on this journey in the first place.

He started with Jevons, probably the most important of the 3 marginal-revolutionaries, even though he didn’t use the word marginal (but "final degree of utility"), the concepts were there.

Fast-forward through the economist's century and to the LLM age, Cowen, like the rest of us, sees the AI writing(!) on the wall and what it's doing to production processes he's familiar with. It's not looking good for popular writing in general (but that's OK, Cowen is old and will live off his savings plus, parasitically, off the rest of us via boomer communism #1468968)... but it's not looking good for economics research either.

“The underappreciated news is that marginalism is on the way out. Furthermore, this is old news, though the trend is accelerating” (p. 85)“The underappreciated news is that marginalism is on the way out. Furthermore, this is old news, though the trend is accelerating” (p. 85)

“Marginalism will be like a sacred book that inspired later religions, but is no longer read or debated.” (p. 85)“Marginalism will be like a sacred book that inspired later religions, but is no longer read or debated.” (p. 85)

Marginalism and price theory, these foundational pillars of economic thought, are there to help us understand what's going on, the trade-offs involved... in the Freakonomics-fame type of conversation, it's to cut the social problem through an economic lens and gain greater understanding of the mechanisms involved.

When LLMs run the show, and AI analyzes data, there's nothing for us to "understand," no price-theoretical structure to grasp or intuit. Marginalism, while very essential to the step-by-step feedback on the inside of an LLM, does nothing for grasping the decision-making and results of what's going on under the hood.

LLMs don’t carry “theory”; they don’t price theory or construct predictions to test. They let their algorithms “build the ‘theory’ for us”,

Price theory, as an approach, is not identical to marginalism. But so many of the basic economic concepts from intermediate micro use marginalist ideas, so price theory has been a comfortable home in which marginalism has flourished, including marginalism as an active research program.

What role is intuitive microeconomics supposed to play in such a system? Big data, flexibility of estimation, and out of sample prediction are prioritized, not concordance with what an economist, even a brilliant one, is geared to expect or even able to understand. (p. 101)

“Sadly, price theory is fading in relevance, and it is taking marginalism down with it.” (p. 92)“Sadly, price theory is fading in relevance, and it is taking marginalism down with it.” (p. 92)

The training of Large Language Models (and other forms of AI) will enshrine marginalism into their basic operating concepts, as those models are trained on writings that understand marginalist concepts. Indeed, if you ask current top-tier LLMs questions about marginalism, they will do very well. But in those frameworks, marginalism will be a sliver of a much broader pie, baked into the inputs we fed our tools of artificial intelligence, but eventually invisible. It will not be at the center of the major analytical constructs and debates, or weigh very much on the minds of economists. (p. 85)

AI reminds Cowen of the early days of marginalism… a revolutionary technique that undermines the previous ways and that’s yet to be appreciated or even remotely understood by the old guard. “Progress in economic methods will be arriving at flood-like speeds, just as it did once marginalist insights were digested and applied” (p. 103)

The academic economists in the crowd can weigh in on this (I don’t really have a view):

One can go back and read major papers from the 1980s, or even the 1990s, and marvel that anyone ever might have believed such results (did they?). Today, you know that for better or worse all stops have been pulled out, at least for papers that are accepted at the top journals for their empirical work. (p. 85)

Has economics, the practice of the discipline, become more rigorous and robust?

if you return to the history of economic thought and ask which trend or school of thought has won out, it is not Austrian, marginalist, monetarist, Keynesian, Marxist, or anything along those lines. The dominant strand in economics today is empirical. And transplanted into history of economic thought language, you could start with William Stanley Jevons as a major driver of that trend. (88)

Open a top econ journal and you’ll see work that isn’t immediately economic in nature. Instead, it’s public health, it’s sociology, it’s social psychology, or statistical political science. It seems the profession and the journal editors went, “well, if it’s good empirical work, it belongs in our journal.” It’s why Emily Oster, who writes about children and child-rearing, is considered an economist; it’s why Erik Angner, who’s basically an improve-the-world policy guy, is an economist; it why John Lott (guns) and Jens Ludwig (gun violence) are economists; why Acemoglu/Johnson/Robinson are concerned economists; why Kahneman (psychologist) and his disciple Richard Thaler (psychologist masquerading as economics) are “economists.”

Later on, Cowen discusses machine-learning techniques for figuring out which sorts of defendant faces have the ability to influence sentencing by a judge

(“Are you not sure this kind of work should be done by economists? **You are way behind the times – those battles are over,”**p. 103).

We’ve basically given up or abandoned the core of what an economist meant. And with it went marginalism and price theory and whatever else might be considered the essence of economic thinking.

The dirty little secret is that what distinguishes economics as a field, right now, is a mix of higher standards, harder work, better math, and higher IQs. That is the real (dare I say marginal?) contribution of “empirical economics today,” not marginalism per se, though of course contemporary models typically are consistent with marginalist reasoning. (p. 86)
The confrontational or “social discomfort” side of marginalism, discussed in chapter one, is now hurting marginalism somewhat. It is not the main reason why marginalism as a series of intuitions is dwindling. Yet (at the margin!), as the economics profession has moved to the left, a diminished role for at least some marginalist intuitions is perhaps not entirely unwelcome. (p. 87)

The trade-off view is gone; social engineering and paternalism have replaced it.The trade-off view is gone; social engineering and paternalism have replaced it.

These days, the economics of economics doesn’t even matter that much… plug-and-play, do some computer sciency things…

The prevailing attitude is that the economics you can figure out along the way, or for some topics you may not need to know much of it at all. How complicated are all those economic principles anyway?

In academia, Cowen suggests price theory, i.e. economic principles, is done for:

After remarking on the poor intellectual state of the interviewees for economics slots at GMU (I presume?), Cowen remarks that while they are smart, highly educated, successful and able to tell you in detail “which estimation methods they have used in their job market paper” (p. 93), they “just don’t think naturally in terms of economic intuition, and the neglected intuitions of course include marginalism” (p. 94).

And the reaper comes for financial economists, too:

For a long time I have thought of finance as the most advanced and most successful branch of economics. It works with the highest quality data, has many of the most rigorous models, and the economic assumption of “people really do want money only” seems relatively justified in that sphere of endeavor. […] when you look at who the top hedge funds and trading firms hire, economists are playing a smaller and smaller role in the sector. Those finance PhDs from MIT are still out there, but they play an increasingly small role. They are being outcompeted by individuals with degrees in math, computer science, and physics. The requirements for quantitative skills at the extreme frontier are so great that very few economists – MIT degrees or not – remain competitive. (96-97)

“In the most prestigious field of economics, marginalism has been in full retreat for over 30 years, and it shows no signs of making a comeback.” (p. 98)“In the most prestigious field of economics, marginalism has been in full retreat for over 30 years, and it shows no signs of making a comeback.” (p. 98)

THIS bit is tragic, echoing in many ways the crashing-to-zero value (#1467035, #1425743, #796401) I’m chirping around here (p. 106):

The bottom line is this: through machine learning, the social sciences have as of late come up with new means of generating, testing, and evaluating novel hypotheses. We used to view that as an especially “human” activity, but the machines are encroaching on that too.

Maybe marginalism, and with it economic reasoning proper, is dead and what we’re seeing are cadaveric spasms.Maybe marginalism, and with it economic reasoning proper, is dead and what we’re seeing are cadaveric spasms.

All hail Jevons and the marginalists and us who came after them; may we all rest in peace.