I really enjoyed the format of ’s book review lists, so I’m going to make it a yearly tradition to summarize my reading for the year in a similar manner. I’ll list the completed books in chronological order below. For books that were notable (whether notably bad, notably good, or containing some notable idea), I’ll include a short blurb with my thoughts and my major takeaways.

#1. Sean McMeekin, Stalin’s War: A New History of World War II

A history of World War II that starts by asking the following questions:

1. Which leaders (if any) were in power pre-, during, and post-war in order to really influence the course of events (the build-up, the actual fighting, and the treaty negotiations afterwards)?

2. Which leaders (if any) benefited the most from the war?

3. Which leaders (if any) had a vested interest in the European powers going to war?

4. Which leaders (if any) successfully bent the leaders of the other powers to their will?

McMeekin asserts that, when we stop to consider these questions, there is only one leader who satisfies all four: Joseph Stalin.

Starting from this point, the book outlines a contrarian history of the war, centering on Stalin's role in instigating the war between Germany and the Western Powers, attempting to contain Germany through resource constraints, and, finally, manipulating the Western Powers to achieve his war aims as cheaply as possible.

I’m not enough of an expert on WWII history to refute or support McMeekin’s claims, but it certainly made for one of the most interesting history reads I’ve ever had.

#2. Nicolas Cole, The Art and Business of Online Writing: How to Beat the Game of Capturing and Keeping Attention

#3. Amaury De Riencourt, The Coming Caesars

Published in 1957, this book outlines the progression of American government from its original, firmly aristocratic character (we didn’t even vote for president in the beginning of the Republic!) to its then-new mass democratic flavor. De Riencourt draws parallels between this progression and the progression of the Greco-Roman world leading up to the end of the Roman Republic and the beginning of the Roman Empire. He uses this parallel to analyze the trajectory of the US and predict the arrival of an eventual American Caesar.

Unfortunately, this book has become extremely relevant in recent years.

#4. Niall Ferguson, The Pity of War: Explaining World War I

Another contrarian historical perspective, this time blaming the United Kingdom for the outbreak of World War I. It argues that had the UK decisively demonstrated its willingness to enter the war before its outbreak, or had instead committed to staying out of it, the total devastation for everyone would have been significantly reduced.

Given the death and destruction wrought by World War I, as well as the knock-on consequences of the war (the Spanish Flu pandemic, the Great Depression, & finally World War II), it’s hard to argue that the UK’s decision to support France was optimal in the long run. All of the European powers would likely have been better off if the UK had left France to fend for itself and it had lost to Germany more swiftly. It hardly seems like the counterfactual could be any worse than what Europe endured from 1914 to 1945. However, I’m not sure that I’m totally convinced that the war was England’s fault in the sense that Ferguson asserts. As with McMeekin’s book, I’ll need to do more reading on the WWI period to really evaluate these claims.

#5. Samo Burja, Great Founder Theory

Some books aim to inform, diving into great detail on a specific concept or event, analyzing all of its minutiae from every angle. Other books aim to enlighten, expounding a big idea that attempts to fundamentally shift your worldview.

This book is one that aims to enlighten.

Burja’s big idea is a sort of re-skinning of the Great Man of History ideal — except, this time, it incorporates institutional design as its centerpiece.

His theory basically goes as follows: we can measure an individual's “greatness” by the effect they have on shaping the world. Any shaping that is limited to the individual’s lifetime is inherently highly constrained, and thus, they cannot be considered a Great Founder. (Note that, in Burja’s view, this means that Napoleon is not a Great Founder for his military campaigns, but he is a Great Founder for things like standardizing units of measurement, creating conditions that led to growing nationalism throughout Europe, and reorganizing the French military structure to support the first foray into total war). In order to ensure that the individual continues shaping the world long after their death, they must create some mechanism by which to transfer their worldview down through the generations. These are our institutions, and their founders are those who have made the world we live in today.

The remainder of the book examines the implications of this theory from various angles.

#6. Amaury De Riencourt, The American Empire

#7. Peter Attia, Outlive: The Science and Art of Longevity

#8. Alonso Cueto, La Hora Azul

#9. Chris Miller, Chip War: The Fight for the World’s Most Critical Technology

I was late in reading this one, but it remains just as relevant (if not more so) today.

I wish I had read it earlier because it quite clearly lays out the case for investing in Intel as the fab of the future in the US.

#10. William L. Hamilton, Graph Representation Learning

#11. Patrick McGee, Apple in China: The Capture of the World’s Greatest Company

One of the best business books I’ve ever read. It treats the history of Apple as a history of its manufacturing division, centering on its ability to churn out massive volumes of some of the most advanced devices in the world. By centering the manufacturing story, China is inevitably centered as well — at first as an apprentice, later as a partner, and finally as an uneasy frenemy of sorts.

This book will teach you so much about Apple, about manufacturing, and about China. I cannot recommend it enough.

#12. Jeff Pepper, Mulan, Woman Warrior: An Easy-to-Read Story in Simplified Chinese and Pinyin, 240 Word Vocabulary

I started learning a bit of Mandarin in 2025, and so far it’s been great fun! I typically like to learn languages through reading (I’ve used this technique before in Spanish and, to a lesser extent, in French). So, to get started on my journey, I looked for some simple yet enjoyable books I could work my way through.

This led me to discover Jeff Pepper, who has simplified many culturally relevant Chinese texts to make them accessible to beginner and intermediate audiences. These include The Journey to the West, various Chinese folktales, a biography of Confucius, and more. Mulan, Woman Warrior is one of the easier of his published books (and tells the same story as my favorite childhood movie!), and so I found it extremely enjoyable to read. I’m very much looking forward to working through The Journey to the West this year.

#13. Dan Brown, The Secret of Secrets

Gemini defines “narrative thrust” as follows:

Narrative thrust is the compelling force that drives a story forward, creating momentum that pulls the reader from one scene to the next, keeping them engaged and eager to discover what happens next

Narrative thrust covers a multitude of sins. Dan Brown, for all his faults, knows how to tell a story with thrust. The Secret of Secrets is no different.

#14. Fyodor Dostoevsky (translated by Michael R. Katz), Crime and Punishment

This is a book that I very much did not enjoy reading, but cannot stop thinking about now that I have read it.

For those who have read Dostoevsky, I’m sure the reasons why I didn’t enjoy the experience of reading the book are obvious — the stilted prose, the confined narrative arc, and the oppressive, suffocating atmosphere that permeates every scene. But for those same readers, the reasons I can’t stop thinking about it must be equally as obvious — the psychological drama, the probing questions into the nature of morality, and the battle between logical rationality and subconscious thought.

Two themes really struck me throughout this book.

The first is Raskolnikov’s ability to reason himself into positions that his subconscious refuses to comply with. To him and other characters in the book, his reasoning is flawless when he expounds on his life philosophy (which, to summarize, is that he is a Napoleonic figure, an example of a higher man, and, as such, is unbound by the constraints of standard morality). In particular, when he is alone and able to ruminate, he frequently finds himself reinforcing this idea in his head. However, whenever confronted with real human connection throughout the story, his Ubermensch idealization breaks down and gives way to a more Christian morality. In addition, there are moments, even when Raskolnikov is alone and avoiding human connection, where thoughts bubble up from his subconscious, questioning his crime. This dichotomy between what he wants to be (the Ubermensch) and what he subconsciously knows he is (the Christian) tears him apart internally throughout the novel.

The second theme concerns the book's ending. After admitting to his crimes, Raskolnikov is imprisoned in Siberia. At the beginning of his imprisonment, he continues his oscillation between feeling himself an Ubermensch and hating himself for it. Only after a year of isolation, during which Sonya Marmeladova, the daughter of a man Raskolnikov helped earlier in the novel, fell ill and could no longer visit him, did Raskolnikov repent and release himself from his delusions. In a sense, the imprisonment in Siberia acted as a death & resurrection for him, after which he had shed the sins of his past life and resurrected as a moral, human agent rather than an aspiring Ubermensch.

Both themes capture the centrality of human connection & emotion in our cognition — Raskolnikov can’t maintain the behavior implied by his reasoning when he comes into contact with others, and he only fully gives up his delusions when he experiences the pain of briefly losing his last human connection, Sonya, during her illness. This tension between self-overcoming and human connection is worth keeping in mind, particularly for those of us who do aspire to more and, in doing so, shed a bit of our natural humanity.

There is another theme I’d like to explore, but I’ll refrain from going on too much of a tangent here. I’ll just say that there are also really interesting parallels between the second theme I mentioned, the life of Edmond Dantes from The Count of Monte Cristo, and the life of Napoleon. I hope to explore these parallels in-depth in a future article.

#15. Tae Kim, The Nvidia Way: Jensen Huang and the Making of a Tech Giant

A really fun history of today’s most important company. Lots of great insights into how Jensen works.

#16. Harry M. Schey, Div, Grad, Curl, and All That: An Informal Text on Vector Calculus

This book builds up the machinery of multivariable calculus through the lens of electromagnetism and physics applications. I’ve been trying to go back and review some math (e.g., calculus, linear algebra, etc.) from a more physics- and geometry-forward perspective so that I can improve my intuitions for the machinery of those topics. Div, Grad, Curl, and All That was great for that purpose.

#17. Michio Kaku, Quantum Supremacy

Yes, I’m aware that many of Kaku’s books devolve into quackery. Yes, I’m aware that Scott Aaronson called this “the worst book about quantum computing… that I’ve ever encountered”. And yes, it does live up to that illustrious billing.

And yet, I still find myself feeling invigorated to learn more and explore new areas every time I read one of Kaku’s books. I’m not sure what this says about me, but here we are.

#18. Richard P. Feynman, Surely You’re Joking, Mr. Feynman! Adventures of a Curious Character

Surprisingly, Feynman’s autobiography focuses quite little on his actual work in physics. Instead, we get to peer into the creative and whimsical mind at play of one of the titans of 20th century physics. Expect lots of stories of mischief and side quests, like lockpicking at Los Alamos or learning to play the drums to join a band during Carnival in Brazil.

#19. Dwarkesh Patel, The Scaling Era: An Oral History of AI, 2019–2025

This book cuts up many of Dwarkesh’s interviews and groups them together in chapters organized by theme (e.g., “Chapter 1: Scaling”, “Chapter 2: Evals”, etc).

Even if you’ve listened to all of these podcasts previously (as I have), the book still provides a lot of value through its theme-based organization and the extensive footnotes/marginalia that Dwarkesh has included to clarify topics and provide additional detail.

Overall, a nice book to quickly assess where we are and where we’re going in AI.

#20. Tyler Cowen and Daniel Gross, Talent: How to Identify Energizers, Creatives, and Winners Around the World

Finding above-average (or even great) employees is not too difficult — we broadly know how to conduct interviews to find people who are competent in their role and conscientious in their execution.

But how do you find people who are truly exceptional?

In Talent, Cowen and Gross aim to answer this question by covering interesting, non-standard interview techniques to help quickly assess whether someone is an outlier among outliers. Very useful if you are a venture capitalist, startup founder, or anyone else aiming to find people who are truly at the tails of the distribution.

Overall, this was a bit of a lighter year of reading than I usually have. In 2026, I’m aiming to increase both the volume and the complexity of the reading I’m doing — hopefully, there will be many more math, physics, and deep learning books in next year’s list than were in this year’s.

In addition, I really like learning history (both political and business) through biographies, and I got away from that a bit in 2025. I’m going to work on incorporating many more of those in 2026.

Lastly, my reading over the last few years has slanted heavily towards non-fiction, but, like Raskolnikov, I’ve been missing the human connection and insights that great fiction can bring. My goal for 2026 is to explore many more of the great books, particularly the seminal novels of the 19th century.

There is a common meme called “the last person who knew all of X”. Some variants include:

• “The last person who knew everything” — often referring to Thomas Young (1773-1829)

• “The last person who knew all of math” — often referring to David Hilbert (1862-1943)

• “The last person to know all of physics” — often referring to Enrico Fermi (1901-1954)

The implication of these statements is that it is now impossible (or at least unheard of) to know everything in these domains. And the dates above show that this phenomenon of the generalist ended somewhere between 100-200 years ago depending on how broadly we want to define the field of study. Thus, we have had around a century of specialization, with the idea that “becoming a true generalist is impossible” largely accepted as fact, whether explicitly or implicitly, by academics everywhere.

The march from high school diploma to undergraduate degree to PhD results in ever-greater degrees of specialization and an ever-narrowing aperture through which to view knowledge. The benefit commonly attributed to the narrowing of our education is that it helps us to reach the frontier of knowledge and thus make novel contributions more quickly. Its proponents say that no one can know everything anymore, and thus it is a waste of time to branch out to fields other than your chosen field of study.

However, my view is that we are artificially constraining our best and brightest by continuing to espouse this narrative that it’s impossible to be an effective generalist. Useful knowledge is not broadly distributed — it is concentrated in a handful of topics contained in only a handful of subjects. You can sense this intuitively if you walk through the aisles of your nearby Barnes & Noble: there are tens of thousands of books spread across the shelves, and the books on the shelves change every few weeks or months. But how many of those titles will be remembered 1 year from now? 10 years from now? What about 100 years from now? It would be a safe bet that no knowledge or insight contained in any new release will be foundational moving forward. By contrast, the knowledge contained in the core of useful books can be re-applied across different sets of facts in disparate fields to solve problems more effectively and creatively than many specialists.

Specifically, I claim that this useful core consists of roughly 300 books of the 170 million unique titles in existence (roughly 0.00017% of all books), and that this core will afford you ~99% of the value of all knowledge ever produced. That is, if you are ~98-99th percentile 18 year old high school graduate (roughly corresponding to ~1500+ on the SAT), then you should be able to get to the forefront of knowledge in the core academic subjects (or have the tools necessary to rapidly reach the forefront of non-core subjects) in about 300 books.

In other words, knowledge is highly power law distributed.

The core subjects that need to be learned to realize this outsized payoff consist of:

• Physics

• Mathematics

• Computer Science

• History

• Philosophy

• Literature (with the canon changing depending on your cultural context)

It is quite possible, within the constraints of our 300 book total, to reach the level of knowledge of a 2nd or 3rd year PhD student at an elite institution in each of these subjects. This is possible for two reasons:

1. Each of these fields has a small core of topics that are explored from many different angles to form the many subtopics of that field. For example, symmetries & action principles in physics, linearity & topology in mathematics, or computability & complexity in computer science.

2. Each of the core topics in these fields provides insight into multiple other fields in the list. Learning mathematics aids both physics and computer science (and, to an extent, philosophy). Learning philosophy aids both history and literature. And so on.

Moreover, once you have reached this level of knowledge across each of these core subjects, you will be able to dive into any subfield of these fields or any field not listed in this core and rapidly reach the frontier of research. This is only possible because other fields use the core insights of these fields as their fundamental scaffolding. Chemistry, for instance, is ultimately the study of valence electrons governed by the laws of quantum mechanics; biology is the complex chemical machinery of life; economics is the application of multivariable calculus, game theory, and psychology; and political science is the ongoing, real-world stress test of history and moral philosophy. When you own the "root nodes" of the knowledge graph, the specialized nodes are often just specific parameters applied to general frameworks you have already mastered. Consequently, the transition from this generalist foundation to a specialist frontier is not a climb up a new mountain, but a lateral step onto a bridge you have already built, allowing you to perceive structural similarities between disciplines that the siloed expert remains blind to.

Once you accept this framework, you see that learning becomes much more about curation rather than exhaustively studying all that has been published. It is the via negativa — choosing not to read a vast majority of writing so that you can instead learn a vast majority of knowledge. Thus, the barrier to becoming a modern Thomas Young is no longer cognitive capacity, nor is it the "impossibility" of the expanding universe of knowledge. The barrier is the discipline to ignore the noise of the 99.9999% in favor of the signal of the 0.0001%, and the patience to master those root nodes before attempting to climb the branches. The library of the generalist is small, but it is heavy.

End note: As an example of what this curation looks like, here is a condensed roadmap I am using to go from no knowledge of physics to a graduate school level [link]

While listening to ’s excellent podcast with Stephen Kotkin (go listen to it if you haven’t yet. I’ll wait.), Kotkin made a very interesting point on the dynamics that led to the Communist revolutions in Russia and China. I’ll provide an excerpt below (emphasis mine):

[In Russia and China], you have this peasant land hunger. The peasants are often without their own holdings. They work on someone else's property, or their holdings are so small that if there's a little bit of bad weather let alone a massive drought, they're on the verge of starvation. Subsistence level agriculture is not politically stable…

You need to deal with the peasant land hunger so that it becomes a stabilizing political force. You have the peasants get the land and then they have a piece of the status quo and want to retain the system, versus the peasants not having the land and they want to overthrow the system to get the land…

So the peasants had their own revolution in 1917 and 1918, which was not about the socialist parties. It's not about the Bolsheviks, it's not about Lenin, it's about the peasants seizing the land. But that creates an intense radicalism that becomes the platform for the socialists in the cities to gain and hold power in the system. You don't have that in the German case.

Kotkin’s claim here is that a large, landless peasant class provides a powder keg, and the Communist movements in Russia and China were smart enough and opportunistic enough to capitalize on this latent instability in the system. By contrast, in Germany, the peasants were predominantly landowners, which provided the stability to weather the Communist movements. Meanwhile, in England, peasants as a class were starting to give way to the urban factory worker, who was able to progressively win voting rights and the right to organize into labor unions.

This result with a Red Russia & China, a fascist Germany, and a capitalist UK seems obvious in retrospect, but for an observer in pre-WWI Europe, this would actually be an extremely surprising result. It was often assumed by Communist intellectuals (including the leaders of the Communist movement) that the global Communist Revolution would begin in England and Germany, which were perceived as the ripest environments for labor to overthrow the capitalist class.

So why did these dynamics play out so counterintuitively? To understand this, let’s analyze the sociological model implicit in Kotkin’s explanation for the causes of the Russian and Chinese revolutions. Broadly, I think Kotkin’s framing of the issue rests on separating the populace into two groups

1. The Dampeners represent the social classes who have a stake in the prevailing order—they have this stake because they stand to gain from political and economic stability and to lose from political and economic instability. In the tsarist Russian context, this would include the tsar and his bureaucrats, as well as the large landowners.

2. The Powder Kegs represent the social classes who feel they do not have a stake in the prevailing order—that is, they feel they do not stand to gain from political and economic stability and, similarly, do not stand to lose from political and economic instability. In the tsarist Russian context, as we’ve discussed, this group is most easily seen in the landless peasant class.

The Dampeners, then, can be thought of as any groups that are currently benefiting and, importantly, feel as though they are benefiting from the current economic, political, and social structure. They represent the vanguard of the current society, defending its interests since the society’s interests are their interests as well. That is, their natural inclination is to dampen any dramatic oscillations in their society in order to defend their current interests and their perceived upside. The Powder Kegs, on the other hand, can be thought of as disenfranchised groups that do not have a stake in the nation’s future. Due to their lack of downside in the face of societal collapse and lack of upside if society remains stable, this class is incentivized to flip the game board over and try to start over. However, in many cases, they will remain latent without an outside spark to light the powder keg.

In this framing, we can think of the Powder Kegs as generators of civilizational tail risks.

An Explanation of Tail Risks & Non-Ergodicity

In the finance world, tail risks are low probability events that cause immense losses (i.e., the left side of the normal distribution in the above image). Tail risks in finance and many other domains also tend to have the following characteristics that make them especially pernicious:

1. They tend to more closely follow a fat-tailed distribution—that is, low probability events occur more often than would be expected by a standard normal distribution modeling the event (e.g., 1-day stock returns) have the following characteristics that make them especially pernicious:

2. The impact of the tail risk is typically much higher than the impact of a more typical data point. For example, a single day when the stock market crashed can wipe out years’ worth of gains.

Given the above, in fat-tailed domains with high-impact left-tail events, we need to be especially careful about defending against (ahem, dampening the effects of) tail risks. And society is the fat-tailed domain. As difficult as forecasting in finance is, forecasting future political and social developments using history might as well be astrology. It’s highly qualitative, subject to opinion, and, as a result, suffers from human status quo biases (or “nothing ever happens,” as the Twitterverse says). Hence, we almost definitely underestimate low probability events in human society. Thus, the distribution is fat-tailed. In addition, we know that the risk of tail events in human society is immense—just look at the Bronze Age Collapse or Europe after the fall of the Roman Empire. Thus, the distribution of impactful societal events is fat-tailed with extremely high-impact tail events.

You may now be thinking, “Okay, these tail events are definitely a concern. But they are, by definition, low probability, so do we really need to worry about them?

Enter non-ergodicity.

To discuss non-ergodicity, we’ll start by intuitively defining ergodicity. Let’s imagine a volume such as a box and some time-dependent process like the motion of a particle in the box. Now, suppose we take two distinct measures of the box:

1. We look take a particular point in time and compute the average position of all particles in the box.

2. We fix our attention on one particle and track its trajectory over time. We then compute the average of this one particle’s position based on the data we record as we follow it around the box

In the first measure, we are computing a spatial average—we look spatially across the box, identify the positions of all particles in it, and then find the average position of these particles.

In the second measure, we are computing a temporal average—we focus on one particle across time, identify all positions that it visits, and then compute an average of these positions.

Mathematically, we can identify the spatial average as the standard expected value of a continuous random variable that you should be used to if you’ve taken an intro course in statistics or probability:

Meanwhile, the temporal average is given by the following:

If the above two notions of an average are equal for a given random process, then we say that it is ergodic. Continuing our physical intuition, this equality is saying that, if we follow the motion of one particle in our box, it will visit all of the locations in the box in the exact proportions that we would expect to find all of the particles at a fixed time. That is, we can use spatial and temporal averages interchangeably because they are the same for ergodic processes. The particle motion example was chosen deliberately, as Brownian motion (i.e., the random motion of particles in a volume) is one of the canonical physical processes exhibiting ergodic behavior.

Now, non-ergodic processes are random processes in which the spatial average and the temporal average are not equal. The behavior of one data point over time does not equal the average behavior across all data points at one instant in time.

Why do we care to make this distinction? Because whenever people talk about averages, they almost always default to the spatial average and then go on to use these averages to make predictions about particular data points. For example, people will often say, “The average yearly return for the S&P 500 is 10%, and therefore I can model my portfolio as gaining 10% per year.” I hope now the issue with this thinking is clear: our hypothetical stock allocator has computed a spatial average and then applied it to one data point over time without checking if the process is ergodic! And, spoiler alert, stock returns are a canonical example of non-ergodic processes, so this reasoning would lead to dangerously faulty conclusions!

Let’s now explore exactly why the combination of underestimating tail risks and confusing ergodic and non-ergodic processes is so dangerous with a concrete example.

Tail Risks, Non-Ergodicity, & Society

We’ll start by building a simple model of the real GDP growth of a standard developed economy that has a latent probability of collapse. We can model this as a discrete-time process, where each year is treated as a single period. In any given year, we will estimate a 0.1% chance of ruin (i.e., the hypothetical country fails and its real GDP goes to 0). This can represent governmental collapse, nuclear war, incurable diseases that wipe out the population, etc. In the vast majority of years (with a probability of 99.9%), we’ll forecast that the country’s real GDP growth is normally distributed with a mean gain of 2.5% and a standard deviation of 0.5 percentage points.

This setup combines the main concepts of tail risks and non-ergodicity that we mentioned previously:

1. Tail risks — we have a 0.1% chance of ruin in any given year. So our risk is extremely rare but also very impactful, just as we’ve defined tail risks.

2. Non-ergodicity — when a country fails (i.e., its GDP goes to 0), it doesn’t have the chance to start over. Hence, if we follow the failed country over time, after it fails, it doesn’t have the chance to explore the full state space (and hence the temporal average will differ from the spatial average).

These two effects, when combined, produce very shocking effects, even though at first glance it might look like we only have a 0.1% chance of catastrophe.

We’ll now run this simulation for 500 years and observe the effects. In this simulation, every country will start with a GDP of $1.

I want to start by directing your attention to the left side of the graph. You’ll see that a full 39.7% of countries in this simulation collapsed, with their GDPs going to 0. Our initial intuition that we had a 0.1% chance of catastrophe was incredibly wrong—we have a 0.1% chance of catastrophe in any given year, but the chance that society fails in any year increases as time advances! Rather than saying “we have a 0.1% chance of catastrophe”, which sounds relatively benign and pushes people towards complacency, it would be far more accurate to say “we have a 40% chance of catastrophe in the next 500 years”, which sounds much more dire.

Now, take a look at the legend in the top-right corner of the graph. Here, you can consider the mean to be the spatial average we’ve been discussing—hence, in the vast majority of conversations where this data is getting discussed, someone would likely say “the average country over this 500-year sample saw its GDP grow from $1 to $12,032.61, a gain of 1,203,160%”. Hence, the reported “average”, the spatial average, actually doesn’t capture the experience of living in these conditions at all! You would think, based on seeing an average gain of over 1 million percent, that all or most countries developing under these conditions would be on the march to utopia. However, these countries only have barely better than a coin flip’s chance of not collapsing into complete chaos in this 500-year sample!

To better understand the dynamics of this process, we need to instead look at the temporal average. We can do this by computing the geometric mean of returns for each individual path and then averaging this set of geometric returns. This yields an average yearly loss of 39.5%, with an expected ending value of $0, aligning much more closely with the collapse dynamics we see in the histogram. This brings us to an even broader insight that can be made here, which we can see in the plot below.

As time moves forward in our sample, more and more of the countries are collapsing. Taking this to the limit, eventually, all societies in this simulation will collapse. Even more generally, when the probability of catastrophic ruin is greater than 0, any society will collapse given enough time.

To tie things back into the beginning of our discussion, what Kotkin has identified in his analysis of the latent instability in the 20th-century Russian and Chinese societies is that the presence of disenfranchised peasant groups increased the yearly chance of ruin for these societies, and thus led to their demises. Combining with our analyses above, we see that when one or more Powder Keg groups exist in a society (making the probability of ruin non-zero), the long-run survival probability of that civilization is exactly 0. The more such Powder Kegs exist, and the more groups that appeal to these groups, the greater the probability of collapse in a given year.

On the flip side, if we re-run the simulation with an order of magnitude lower probability of ruin (i.e., a 0.01% risk of collapse per year), we see a huge reduction in risk in the 500-year window. Only 4.7% of societies in this scenario collapse, reducing our risk of ruin by 35 percentage points.

Given these analyses and the conclusions we’ve drawn, it would seem that identifying potential Powder Kegs and implementing policies to let off some steam are some of the most important actions a government can undertake. If we can identify these groups and pacify them, we can substantially improve the survival probability of our society over large timescales and avoid outright collapse. The late Russian tsars and Chinese emperors failed to identify and pacify their Powder Keg group. The Communist parties in those countries, however, did not overlook the peasant Powder Keg group and were thus able to leverage them to bring the country’s risk of ruin to reality.

With this lesson in mind, let’s look at a Powder Keg group that is roughly analogous to the landless peasants in tsarist Russia and examine some ways we can defuse the impending explosion.

A 21st Century Powder Keg & Its Potential Solutions

To identify our Powder Keg, let’s return to our definition of the group: we need to find a social class that doesn’t have a stake in the prevailing political order. That is, we need to find a group that doesn’t stand to gain from the country’s success since it doesn’t own a stake in the country, and also has the subjective feeling that it will never achieve such a stake in the current system. Hmm… where can we find such a group?

If you clicked on any of the myriad links in the previous paragraph, I think it’s clear that Gen Z fits the bill—they are currently dealing with extremely high housing costs that have put home ownership out of reach for many members of that generation (at least, that is, in the cities with the jobs they’d want to work and places they’d want to live). Most importantly, this generation feels like it will never get better. There is a very strong, pervasive sense that society is fundamentally set up so that they won’t succeed. We can see this dynamic starting to play out politically, with Gen Z shifting strongly to Trump during the presidential election, who openly espoused the need to tear down the current economic and social order to “Make America Great Again”. We are also seeing this in New York City, where Gen Z voters are turning out in strong numbers to vote for Zohran Mamdani, the once‑underdog‑turned‑favorite mayoral candidate who has some fairly openly held anti-capitalist views.

Whether or not the issues Gen Z faces require a radical overthrow of the prevailing social order, much of Gen Z certainly feels that this is the case. And that is enough for a radical group (like the Communists did in 20th-century Russia and China) to leverage this group as the bedrock for a revolution. Which means, to protect our civilization from collapse, we have to resolve the issues they’re facing and give them (and future generations) a stake in the current social order.

The way I see it, having a stake in American society as it’s currently constructed essentially amounts to two things:

1. Owning a home

2. Owning a piece of the stock market

Owning a home is the traditional view of success in the United States, and so the inability to buy causes immense stress to younger generations. It makes them feel shut out from the typical path to prosperity that their parents and grandparents have followed. And, as we’ve discussed, simply feeling that there is no hope is enough to provide a latent layer of unrest, regardless of its truth. By providing them with a pathway to home ownership, we could reduce the anxiety and anger directed at the US social order. In addition, home ownership typically connects people to a community more than renting does, as they can form more permanent roots, which could also help to assuage the social isolation felt by many in the younger generations.

Similarly, the more modern view of success amounts to having significant capital in the US stock market. Gen Z generally feels resentment towards US companies and CEOs, mostly because they feel that they are being “robbed” or taken advantage of by these companies. This feeling is likely augmented due to their lack of ownership in the stock market—if they owned a slice of all of these corporations, it would be much harder to feel robbed since Gen Z would, in turn, be getting richer from their success.

So we now have a two-pronged approach—give young people a vested interest in US corporations and a vested interest in the housing market. Both of these assets will reduce the desire for societal disruption and instead incentivize stability and growth, since younger people would now be directly harmed by disruption and directly benefit from growth.

What would policies targeting these two areas look like?

Housing Policy

To defuse the Gen Z Powder Keg through housing, we must confront the sacred cows of American urban planning and economic policy head-on, adopting measures that prioritize radical supply-side interventions over the palliative demand-side subsidies that have only inflated bubbles and entrenched resentment. The goal here is not mere affordability tweaks but a structural overhaul that floods the market with housing units, crashing prices to levels where ownership becomes a default expectation rather than a distant dream, thereby converting potential revolutionaries into vested stakeholders who crave stability to protect their newfound equity.

First, deregulation stands as the linchpin: easing zoning laws, building codes, and permitting processes to unleash denser, faster development. In cities like San Francisco or New York, where Byzantine zoning ordinances have artificially constricted supply, we could emulate the Houston model on steroids. Imagine upzoning entire neighborhoods to allow mid-rise apartments and mixed-use developments without the endless environmental impact studies or community vetoes that currently serve as de facto barriers erected by incumbent homeowners (our modern Dampeners) to preserve their scarcity-driven windfalls. By slashing approval timelines from years to months, we'd not only multiply housing stock exponentially but also incentivize innovation in modular construction and prefabrication, driving down costs through economies of scale. This isn't about anarchic sprawl; it's about recognizing that the current regulatory thicket is a form of crony capitalism, protecting legacy interests at the expense of intergenerational equity. The contrarian insight? Such deregulation would disproportionately benefit high-density urban cores, where Gen Z wants to live, fostering vibrant, walkable communities that counteract social atomization, turning isolated renters into networked homeowners with skin in the civic game.

Complementing this, implement a land value tax (LVT) while eliminating traditional property taxes. Drawing from Georgist economics an LVT taxes the unimproved value of land itself, not the structures or improvements upon it. This shifts the fiscal burden onto speculative land hoarders and absentee owners, who currently profit from scarcity without contributing to productivity, and rewards those who build densely and efficiently. In practice, this would discourage vacant lots in prime areas (a plague in places like Los Angeles) and encourage vertical development, as the tax incentivizes maximizing output per square foot. Eliminating property taxes, which penalize improvements, removes the disincentive to renovate or expand, further accelerating supply. The net effect? Housing prices plummet as land speculation collapses, making entry-level ownership feasible even for baristas or entry-level coders. Critically, this policy is revenue-neutral or even positive for governments, funding infrastructure without the distortionary effects of income or sales taxes.

Next, a targeted reduction in union power within the construction sector to slash labor input costs. Unions, while historically vital for worker protections, have in many blue states metastasized into rent-seeking guilds that inflate wages far above market rates through prevailing wage laws and project labor agreements, adding 20-30% premiums to public projects alone. By reforming or outright repealing these mandates (perhaps via federal preemption in interstate commerce-affected builds), we could introduce competitive bidding from non-union labor, including skilled immigrants under expanded visa programs. This isn't anti-labor animus; it's a recognition that artificially high costs perpetuate the very inequality unions purport to fight, locking out younger workers from affordable homes while padding the pensions of Boomer incumbents. The result? Faster builds at lower prices, with spillover effects into related industries. Tie this to broader labor market fluidity (e.g., right-to-work expansions) and we create a virtuous cycle where construction booms absorb underemployed Gen Z, giving them immediate income upside alongside future ownership stakes.

Finally, ramp up domestic energy production to crater energy input costs, which permeate every stage of housing development from raw materials extraction to HVAC installation. Fracking, nuclear deregulation (streamlining NRC approvals to Yucca Mountain levels of efficiency), and offshore drilling expansions could flood the market with cheap BTUs, reversing the green energy mandates that have jacked up costs via intermittent renewables and grid unreliability. By prioritizing baseload abundance we not only halve construction energy bills but also enable all-electric homes that are cheaper to operate, further lowering the ownership threshold. In our tail risk framework, this policy acts as a multiplier: cheaper energy stabilizes the broader economy, reducing the probability of ruin from exogenous shocks like oil crises, while directly empowering the housing supply surge that integrates Gen Z as Dampeners.

Implemented holistically, these policies could halve median home prices in high-demand areas within a decade, per supply elasticity models, transforming Gen Z's latent rage into conservative impulses. After all, nothing quells revolutionary fervor like a mortgage and rising property values.

Stock Market Policy

Shifting to the equity side, the objective is to forge a direct umbilical cord between young Americans and the productive engine of capitalism: the stock market. Rather than the anemic 401(k) matches or tax-advantaged IRAs that presuppose stable employment and disposable income, we need audacious, universal mechanisms that inject ownership stakes from cradle to career, preempting the feeling of disenfranchisement that fuels Powder Kegs.

The recently enacted MAGA (Money Accounts for Growth and Advancement) accounts (rebranded as Trump Accounts in the final text of the One Big Beautiful Bill) represent a strong starting point in this. These accounts automatically seed every eligible newborn with a $1,000 federal contribution (under a pilot for births from 2024 to 2028), invested in diversified funds tracking U.S. equity indices like the S&P 500, with tax-advantaged growth and additional contributions capped at $5,000 annually from taxable entities or unlimited from tax-exempt organizations. Access is tiered—no withdrawals until age 18, then partial for qualified purposes like education, entrepreneurship, or homebuying, expanding at 25 and fully unrestricted at 30—with favorable long-term capital gains taxation on qualified distributions. These accounts instill from infancy a visceral sense of ownership in corporate America, equating personal gain to national prosperity. To secure broader political alignment and ensure the program's longevity beyond electoral cycles, however, a further rebranding is essential, shedding the partisan connotations of "MAGA" or "Trump Accounts" in favor of a neutral, inclusive name like "American Future Funds" or "National Prosperity Accounts." This move transcends left-right divides, framing the initiative as a non-ideological investment in collective stability; by inviting buy-in from all societal factions, we mitigate the risk of future repeals or sabotage, transforming it into an enduring institution that unites rather than polarizes, much like Social Security evolved from its contentious origins into a bipartisan bedrock.

However, these accounts don’t go far enough: the $1,000 seed is paltry in the face of compounding's long horizons (barely scratching $10,000 by age 18 at historical 7% real returns), eligibility skews toward newborns and young children (leaving Gen Z and millennials largely sidelined without retroactive catch-ups), contribution limits throttle broader participation, and the delayed access risks alienating a generation already primed for immediate disruption over deferred gratification. To truly dampen tail risks, we must expand them aggressively: scale the initial deposit to $5,000, extend automatic enrollment with catch-up lumps ($10,000+) for 18-35-year-olds funded by reallocating corporate tax windfalls from buybacks to "citizen shares," eliminate age-based restrictions in favor of penalty-enforced long-term holds, and broaden investment mandates to include total-market ETFs for true diversification. This supercharged version subverts the Marxist narrative of "exploitation by capitalists" by making everyone a micro-capitalist, diluting class warfare into shared prosperity. In tail risk terms, widespread equity ownership dampens systemic shocks; if a critical mass holds stakes, political pressures shift from "eat the rich" to "protect the market," reducing the annual ruin odds by fostering a Dampener supermajority.

To amplify this foundation even further, redirect any universal basic income (UBI) pilots or proposals straight into mandatory stock purchases for recipients under an expanded "Universal Basic Equity" (UBEI) framework, rather than cash disbursements that dissipate into consumption traps. Imagine channeling equivalent funds (say, $1,000 monthly equivalents) into the augmented MAGA/Trump Accounts or similar diversified portfolios, perhaps via Vanguard-esque vehicles with algorithmic rebalancing. This forces skin-in-the-game; recipients become vested in corporate efficiency and innovation, as their "income" derives from dividends and appreciation rather than zero-sum transfers.

Collectively, these stock policies embed a profound incentive alignment: young people gain from US economic ascendance, viewing disruptions—like tariffs or regulations—as direct threats to their wealth. This vested interest in continued success transmutes Powder Kegs into Dampeners, slashing civilizational ruin probabilities and extending our society's survival horizon indefinitely.

Conclusion

In revisiting Kotkin's insights on the peasant-driven upheavals that toppled empires in Russia and China, we've unearthed a timeless sociological truth: societies teeter on the edge of ruin when disenfranchised masses, our Powder Kegs, accumulate without recourse, their latent volatility amplified by opportunistic sparks. Through the lens of tail risks and non-ergodicity, we've modeled this as an inexorable march toward collapse in any system where the annual probability of catastrophe exceeds zero, where spatial averages seduce us into complacency while temporal realities reveal the fragility of civilizational paths. Gen Z, as the 21st-century analog to those landless peasants, embodies this threat not through malice but through a profound sense of exclusion from the American dream's twin pillars: homeownership and equity participation.

The policies outlined—radical housing deregulation paired with Georgist taxation, union reforms, and energy abundance to crash barriers to entry; alongside expansions of rebranded, universal stock accounts and equity-directed UBI to democratize capital—are strategic dampeners designed to integrate the alienated into the status quo's vanguard. By granting tangible stakes in stability, we slash that ruin probability, extending our society's viable horizon from inevitable doom to indefinite prosperity. This represents a first attempt at pragmatic ergodic engineering: in a fat-tailed world, the true revolutionaries are those who preempt the keg's ignition, ensuring that history's surprises favor continuity over cataclysm. Ignore this at our peril; embrace it, and we might just forge history’s first immortal civilization.

Returning to the language learning grind has me reflecting on different modes of learning, their interplay, and how people often get them wrong. To provide us a jumping off point, I’m going to define two broad modes of learning:

1. Explicit Learning, which we can consider any learning that requires teaching rules or procedures. This type of learning is often accompanied by a textual explanation of the procedure and some examples to practice applying that procedure. Most subjects in school are taught this way: e.g., learning a new approach in math or working through a grammar workbook in Spanish class.

2. Implicit Learning which we can consider any learning that occurs as a mostly unconscious process. A helpful rule of thumb is that any learned material that requires a response in seconds falls under the purview of implicit learning - at that time scale, there isn’t enough time to work through a procedure (and hence to use the tools of explicit learning).

Now, anyone who has learned a language in school (or read my not-so-subtle leading example in the Explicit Learning section) knows that explicit learning is the modus operandi of foreign language instruction in the United States. Virtually the whole process occurs in a grammar textbook where the student is shown some concept, such as the conjugation of verbs ending in -ar in the present tense in Spanish. The student then reads through the “logic” of how this grammar concept works and works through a set of examples in order to “apply” the procedure.

I hope the scare quotes I used above made it abundantly clear how ridiculous I think language learning logical or procedural is. In fact, there is no logic to language grammar since it did not grow out of a logical process! Language grew organically alongside humans, and so it has all the same messy, non-logical characteristics that we see in all things biological and anthropological. This is why every grammar “rule” that is taught has dozens, if not hundreds, of exceptions! We’re applying a method of learning suitable for math to a completely unrelated field in language learning that categorically does not follow a logical structure with teachable procedures!

(Trigger warning for anyone traumatized by this teaching method: I am about to show an example of one of these accursed grammar worksheets.)

Now you might be thinking: “Okay, so easy fix — we just move all of language learning and any other similar fields to implicit learning. Explicit learning remains reserved for the logical subjects of math and physics. QED.” In this paradigm, all of the non-rigorous subjects would eschew explicit learning, and all of the rigorous subjects would focus only on explicit learning.

However, despite this aligning more with the natural contours of each subject, we would be leaving potential learning speed on the table by sticking only to implicit or explicit learning. I will give two examples to illustrate the point:

• Language Learning (Implicit Learning Paradigm) — Suppose you are learning a foreign language and, like the good language learning student we’ve outlined thus far, you’re doing it fully through immersion (that is, implicit learning). You avoid all grammar explanations and worksheets, and instead spend all of your time listening to podcasts, reading books, and watching YouTube videos in your target language. If you come across a particular grammar construction that you don’t understand, you just try your best to understand it in context and keep immersing. With this setup, it might take dozens of hours of immersion to understand the grammar concept, particularly for rare constructions or variants of the same construction that look unrelated on the surface. Meanwhile, reading about the grammar topic and doing some exercises might take under an hour and prime you for your immersion.

• Mathematics (Explicit Learning Paradigm) — Now suppose that you’re a math student and, like the good math student we’ve outlined thus far, you’re doing it fully through working through derivations and proofs (that is, explicit learning). You avoid any form of memorization or intuition building, and instead spend all of your time focusing on logically deriving all of your course content. Now, as the course moves on, it will assume knowledge that you learned earlier in the semester. However, since you didn’t bother to do any memorizing or intuition-building, you don’t remember those earlier concepts with any sort of automaticity! You need to rederive all the knowledge from scratch, and, after the course material builds up beyond a certain point, the cognitive load of this derivation becomes too much, and you start to fall behind.

As the two above examples show, even in subjects that are skewed towards one of the learning paradigms, the other still has an important role to play.

In implicit-dominant subjects, we can use explicit learning to identify or study patterns that we would like to notice during the implicit skill development. This would be like a language learning studying a grammar concept (explicit learning) to improve their understanding during immersion (implicit learning), or a basketball player breaking down defensive coverages in film (explicit learning) so he can make better reads during scrimmages and games (implicit learning).

In explicit-domain subject, we can use implicit learning to memorize and internalize concepts and procedures without concerning ourselves with their logical derivation. This would be like a mathematics student memorizing common integrals (implicit learning) to make it easier to understand the logical derivations of concepts that use integration, such as probability distributions (explicit learning).

Thus, broken down this way, the problem of learning any subject can be reduced to a three-part problem:

1. Where can you use explicit learning techniques in this subject?

2. Where can you use implicit learning techniques in this subject?

3. What should the balance and interplay of these learning paradigms be in this subject?

If you nail these three questions for your particular subject, then you will have achieved optimal learning speed and efficiency.

Subscribe now

Renewal and Markets

For most of recorded human history, progress was nearly too small to measure along nearly all relevant axes: life expectancy, infant mortality, GDP per capita, etc. Progress along each of these vectors was measured on the order of centuries or millennia, and any hard-won progress was extremely fragile, with events such as the fall of the Western Roman Empire taking centuries to recover from in areas such as Britain. However, beginning in the late 1700s, a confluence of technological, sociological, and ideological innovations led to capital P “Progress” — the sort of rapid, hockey stick growth that we’ve come to expect as a natural part of our world. One of these key innovations, perhaps the most important of them all, was the introduction of capitalism and the free market economy.

Capitalism and free markets together unleashed several key forces that ignited global growth: competition, distributed information processing, and incentive alignment. I’ll outline briefly the importance of each of these below:

• Competition. Competition is the driving force behind innovation and efficiency in capitalist economies. When companies vie for market share, they are compelled to improve their products, lower prices, and increase efficiency to attract consumers. This constant pressure to outperform rivals fosters an environment where only the most efficient and innovative businesses thrive. Competition not only benefits consumers with better and more affordable products but also pushes companies to continually innovate, ensuring that progress never stagnates.

• Distributed Information Processing. Information about consumer preferences, resource availability, and technological possibilities is widely distributed among individuals and businesses. This decentralized information processing allows markets to quickly adapt to changes. Entrepreneurs and businesses can respond to signals from the market, such as shifts in consumer demand or new technological advancements, enabling a dynamic and responsive economic system.

• Incentive Alignment. Capitalism aligns incentives by tying the success of individuals to the success of the companies they own or work for. When management and workers have a stake in the ownership of their companies, their incentives align closely with the economic growth and dynamism of the broader economy. Managers and employees who hold equity in their firms are directly rewarded for the company’s performance, encouraging them to work harder, innovate, and make decisions that enhance the company’s value. This ownership structure creates a powerful alignment of interests that drives productivity and fosters a culture of continuous improvement.

These three factors have endowed capitalist free market economies with a remarkable capacity to enrich society through a built-in mechanism for rejuvenation. While individual companies may not endure indefinitely, the market system itself thrives and evolves, welcoming new entrants and jettisoning outdated competitors.

The cycle of rejuvenation in free markets follows these steps:

1. Innovation Emergence: A new technology, methodology, or market becomes available, often driven by advancements in science and technology or shifts in consumer behavior.

2. Entrepreneurial Entry: Entrepreneurs and new companies seize the opportunity to exploit this new angle, creating products or services that meet emerging demands.

3. Incumbent Inertia: Established companies, often characterized by bureaucratic inertia and resistance to change, struggle to adapt to the new paradigm.

4. Competitive Displacement: The new entrants, leveraging their innovative approaches and exploiting the incumbents' weaknesses, begin to outcompete and displace these established firms.

5. Decline of the Old Guard: Incumbent companies shrink or exit the market, unable to maintain their previous dominance.

6. Cycle Continuation: The high-growth new companies eventually become the new incumbents, setting the stage for the next wave of innovation and disruption.

This cycle ensures that, while no single company survives for eternity, the system itself survives and grows more efficient and prosperous. And, crucially, virtually everything about the companies within the system can change. They can have different governance structures, different incentive structures, leverage different technologies, target different market segments, and more. While the system itself remains in place, its constituent parts can change to the point of being unrecognizable. This allows our economy to respond dynamically to rapidly changing technological and environmental factors and to leverage these changes for continuous improvement — that is, in the parlance of Nassim Nicholas Taleb, to be Antifragile.

The Ossified Nation-State

In stark contrast to the above, there is no equivalent process for rejuvenation and change in nation-states. Governments are often enshrined in constitutions that either remain largely unchanged (as in the United States) or undergo only superficial modifications while retaining the same overarching structure (as seen in many Latin American countries). The only mechanism for large-scale governmental change appears to be a total collapse of the existing order, which can come through:

• Violent Revolution: Internal uprisings that overthrow the existing order.

• External Conquest: Invasions and conquests that impose new governmental structures.

• Post-War Reorganization: Major wars resulting in governments' dissolution and reformation, such as Germany after World War I.

Historically, the structure of democratic republics was able to partially fill the role of free market economies in the context of nation-states. Democratic Republics such as the United States allowed voters to periodically replace presidents, cabinets, and policy agendas through regularly scheduled elections, avoiding the violent regime changes mentioned above. The combination of these regularly scheduled elections with de facto term limits on the executive branch of government created a system where new leaders, new ideas, and new policy structures were continually brought in to renew the government. However, several contemporary factors are undermining the Democratic Republic as a rejuvenating system:

1. Aging Politicians: Politicians are getting older and staying in office longer. Increased lifespans, coupled with the absence of term limits for Congress, contribute to this trend. Additionally, mass media and a larger, less informed voting populace make name recognition a crucial determinant of electoral success, favoring long-term incumbents.

2. Influence of the Supreme Court: The Supreme Court plays a more significant role than in the past as partisan interest groups frequently come to a gridlock. By elevating many decisions to the Supreme Court, the nation’s decision-making power becomes more and more centralized in a group of justices that receive lifetime appointments, limiting the capacity for new views to be brought in.

3. Persistent Bureaucracy: The bureaucratic apparatus surrounding the government has increased in size and does not change with each administration, so the personnel and ideologies making up the government are more constant from administration to administration.

All of these factors together make the US government increasingly brittle — it is less able to change in response to rapidly changing environments, and so is likely to break. Think of the difference between a wood plank and water. Water changes in response to even tiny perturbations, but the benefit is that will not "break" from a large impact. By contrast, a wood plank will resist most changes, until one change that is large enough to overcome its resistance snaps the plank in half. Similarly, the increased rigidity of the US government has now made it harder to bend, making it more likely to break in response to extreme stress and environmental, economic, or sociological change.

How do we move forward?

To improve the situation, we need to ensure a much higher turnover in the government. This includes personnel, policies, legislation, and ideologies. We need to ensure that the government can respond rapidly to changing conditions, which can only occur by allowing the governance structure in place to shift without the strong status quo bias we see today. There are a few factors that I think need to change to allow this to happen:

1. Radically reducing the size and scope of government agencies with unelected officials. Bureaucratic agencies since WW2 have been able to run roughshod over the country, interpreting and enforcing laws as they see fit and being largely impervious to public scrutiny. The opacity of the organizations and the inability of voters to remove officials from them through elections has allowed them to ossify, making the government apparatus significantly more stagnant from administration to administration. The Supreme Court’s ruling in Loper Bright Enterprises v. Raimondo and Relentless, Inc. v. Department of Commerce is a step in the right direction in this case, limiting the ability of agencies to actively interpret the law without much oversight (which can lead to ambiguous and shape-shifting regulatory implementations, creating huge strains on businesses).

2. Setting expiry dates on all government legislation. Laws existing in perpetuity create strong inertia in the government — laws that already exist will likely continue to exist because keeping them requires no action, whereas repealing them requires some action. This results in ever-increasing regulations, tax codes, and legal codes which sap dynamism from the economy and provide significant market opportunities to purely extractive companies and institutions. Both individuals and corporations would benefit from simpler tax codes and reduced regulatory burdens.

3. Imposing term limits on Congress and the Supreme Court. Term limits would ensure a regular influx of new ideas and reduce the likelihood of long-term incumbency leading to stagnation. In particular, poor choices for appointments to the Supreme Court are disastrous, as lifetime appointments can last for several decades before the spot is turned over to a new justice. By making appointments temporary, we can limit the negative impact of bad appointments while also ensuring that new generations of justices regularly filter into the court.

4. Implementing cognitive testing for all elected officials. This one is self-explanatory. We should not have to repeat the embarrassments of this national election cycle, where each party has competed to prove that the other’s candidate is the more geriatric and clueless. We as a nation should have the confidence (and frankly, the dignity) to be able to assume that our leaders are cognitively capable of handling the job.

The goal of the above measures is to create a system that fosters innovation and dynamism in the government by allowing the government to change. Each measure aims to remove bloat from the current system, increase competition for positions among elected officials, and reduce the status quo bias that is so prevalent in politics today. There is certainly substantially more that needs to be done above and beyond these suggestions, but it does present a starting point for a government that isn’t stuck doing a cheap parody of FDR’s administration.

16 years. We enter the school system at 5 years old and typically exit college at 21. In those 16 intervening years, we’re told that perfect is not only attainable — it’s our only option. We must score 100% on every exam, we must get an A for the class at the end of every semester. And there are no do-overs; there is no stumbling and getting back up again. One mishap on an exam permanently relegates us to second-class citizens, unable to receive the A that we so desperately need. As a result, we learn that failure is something to be avoided at all costs and in all circumstances.

Existing in this environment for over a decade of our lives (and sometimes over two decades for those of us who go on to graduate school) instills this type of thought into our psyches. Failure becomes not just undesirable — it ceases to be an option. Any failure is seen as the end of the world because, for at least 16 years, it was the end of the world for us. As a result, even though we may be outside of the school environment, we fear failure to the same degree and avoid the situations which may lead to it. This fear becomes self-reinforcing — the more we avoid situations out of fear that may lead to failure, the more we feel that our fear is justified. Our actions tell us that our emotions are reasonable, and so we avoid failure with increasing conviction and fear each passing day.

This avoidance of situations that may lead to failure is disastrous for long-term growth. The world is governed by probability — few, if any, actions that we undertake entail certain success. Those that do have a near-certain probability of success are simple, small-scale, and have been done by countless others before us, giving us a clear roadmap for how to achieve what we are aiming for. Unfortunately, using these three features, we have specifically defined an action that is not valuable, as it is not difficult (due to its simplicity), is not impactful (due to its small-scale nature), and is not novel (due to the preponderance of people who have completed the action before us). In short, by limiting ourselves to actions with a near-certain probability of success, we are limiting ourselves to the least valuable actions available to us, reducing the growth we can experience as individuals, as well as the impact and scope that our lives can have on our communities around us.

In order to shift our focus from actions that are not valuable to those that are, we must open ourselves up to a very real (and frightening) risk of failure. Perfection and its associated connotation of guaranteed success are not possible for most valuable actions, such as starting a business or approaching the person who may end up being your spouse. These actions bring along large probabilities of failure but massive rewards in the case of success. To overcome that need for perfection and the fear of failure which stops us from undertaking these actions, we first attempted to understand where our fear comes from, as we described in the introduction. Now we must also understand why repeated failure is a normal and necessary part of the process when undertaking valuable actions.

There are two key aspects that all of the most valuable actions share:

1. The probability of success is low, but the rewards are massive. It might take multiple attempts to achieve success, but the effects are life-changing.

2. The probability of failure is large, but the outcome associated with failure is not ruinous. You can get back up and keep going in the event of a failure.

The high reward that accompanies success ensures that we have a reason to continue attempting the action, despite failure. The high probability of failure along with its relatively low associated costs ensures that we almost definitely will fail along the way, but that costs are not so high that we cannot refine our approach and try again in the event of failure.

Moreover, if we look at each attempt at a specific action as an independent trial, we see that we can dramatically reduce the probability of failure (i.e. never having one success) by repeatedly trying this action. For example, if our probability of failure when applying for a new job is 0.9, then if we apply to 100 jobs, our probability of failure falls to 0.00002, virtually guaranteeing at least one successful job application. If instead we were focused on perfection, we would try to minimize that 0.9 probability of failure before taking any action. Even if we reduced the probability of failure by 80% through planning and preparation, we would still have a 0.18 chance of failure. In summary, repeated attempts at an action can dramatically reduce the probability of failure when compared to simply attempting perfection from the outset.

Put plainly, the above arguments say that, for high-value actions, the rewards for success are well-worth the costs of failure and that multiple failed attempts at an action increase our odds of eventually succeeding. As a result, it is in our best interests to place our fears to the side and put our best foot forward for any action which we want to accomplish. Waiting for the perfect time or trying to design the perfect situation for action is an impossible task in most cases and, even when it is possible, it results in lower chances of success than simply trying, learning, and iterating.

Putting it all together, we see that fear of failure stems from years of a perfectionist mentality being instilled in us from the school system. This fear of failure limits us to undertake only those actions which are not valuable, but this need not be the case. We can shed this fear by understanding its origins and how it limits us. We have seen that truly valuable actions cannot be achieved with a perfectionist mindset due to their high probability of failure. We now know that repeated action, not perfect action, is the key to success in high-value situations. And we can take this knowledge forward to achieve what before we were afraid to even dream of.

One year ago today, on a whim, I opened Netflix and searched for a new show to watch. While scrolling through the endless options, I found a show that sounded interesting: Money Heist. I remembered seeing the name in advertisements and internet discussions, and I hadn’t watched a heist show in quite a while, so I figured I’d give it a shot.

About five minutes into the first episode, I noticed that the audio seemed to be out of sync with the actors’ lips. I opened the audio settings for the show to investigate and discovered, to my surprise, that I was actually watching the show dubbed in English! The original audio for the show was in Spanish.

I had studied Spanish pretty seriously for several years in the past, but at the time I found Money Heist, I was on a five-year hiatus from engaging with the language. I decided it would be a fun challenge to try watching the show in its original Spanish in order to see how much I could remember from my earlier studies, so I switched on the Spanish audio and Spanish subtitles and settled in to finish the first episode.

On that day, my love for Spanish was rekindled, and I’ve been immersing ever since. After now having passed the one-year milestone, I decided to write this post to detail what I did, what worked (and didn’t work) for me, and where I plan on going next in my journey.

The Method

Note: if you’re already familiar with immersion learning and Refold, feel free to skip this section.

After getting hooked by Money Heist, I began scouring the internet for resources on how to improve my Spanish and what I could do to speed up the language learning process. I had read some of Stephen Krashen’s work in the past, so most of my search revolved around input-based methods — which is how I eventually stumbled on Matt vs. Japan and Refold.

These resources have been absolutely indispensable for my language learning journey, so I definitely recommend checking them out in full if you’re interested in learning a foreign language, but I’ll provide a brief summary of the methods here so we’re on the same page.

The key premise behind Refold’s ideology is the idea of language immersion. Refold asserts that no amount of explicit grammar study, textbook exercises, or speaking drills will get you fluent in a language. The only way to reach true fluency is to consume content — and to consume a lot of it.

They then split language immersion into three different types: intensive active immersion, free-flow active immersion, and passive immersion. Intensive active immersion entails focusing all of your attention on the content you’re immersing in (whether it be a show, podcast, book, etc.) and actively looking up words or grammar structures that you don’t know. On the other hand, free-flow active immersion gets rid of the dictionary and only requires you to focus all of your attention on your immersion. You allow yourself to follow the story without getting bogged down by the details. Lastly, there’s passive immersion, which is immersion that is done while performing another task that doesn’t require much attention. Examples include listening to an audiobook while you fold clothes or listening to a podcast while you drive.

In order to reach fluency, ideally, you will want to employ a mix of all three types of immersion. Intensive active immersion will be most productive on a per-hour basis, but it is also by far the most mentally taxing and most difficult to fit into a daily schedule. As a result, the majority of my immersion has fallen into the free-flow active and passive categories.

Refold also recommends the usage of a flashcard app called Anki, which is a form of spaced repetition software (SRS). An SRS basically optimizes the time between reviews of your flashcards so that you review a card just before forgetting it. Saving a review for just before (or just after) you’ve forgotten the content of the card has a couple of important effects. Firstly, it prevents you from wasting time by reviewing a card that you already know very well. Secondly, by waiting until you are on the verge of forgetting, reviewing the card becomes much more difficult. The difficulty of recalling the contents of your card actually causes your brain to reinforce the neural circuitry which contains that memory to a much greater degree than if your review had been easy (see Make It Stick: The Science of Successful Learning for more info on this phenomenon).

The role of the SRS is to help memorize important vocabulary and grammar structures. Cards typically consist of Spanish vocabulary on the front and English on the back at first. However, you very quickly move on to creating sentence cards, which are created from sentences that you find during your immersion in which there is one unknown element (either a word or grammar structure). Sentence cards consist of a full Spanish sentence on the front, and then one of a few options on the back:

• A translation of the sentence into English

• An English definition for a word you don’t know in the sentence

• A Spanish definition for a word you don’t know in the sentence

The best option of these three will depend on your level of ability in the language and the specific word or grammar structure you’re trying to learn. Combining the SRS with immersion is an extremely powerful tool and allows you to rapidly expand your vocabulary, helping you quickly move onto more challenging immersion material.

Now that we have the basics of immersion-based learning and SRS usage ironed out, I’ll talk about what I did on my journey and how I found it worked out.

Reading

Stats:
Books: 20
Pages: 5132
Words (estimate): 1,539,600

Reading has been by far my most successful domain throughout my Spanish journey. I bought a couple of graded readers right after starting Money Heist (Classic Spanish Stories and Plays and Read & Think Spanish) and started working through them. Immediately, I found that I loved picking up vocabulary and grammar through extensive reading but struggled to do the same through listening. As a result, I shifted the majority of my immersion time to reading. Most of my reading immersion for the first few months consisted of free-flow active immersion because of the difficulty of looking up words — I felt that it broke the flow of my reading if I looked up too many words and made significantly affected my enjoyment of the story.

After the first few months, I decided to set up an old Kindle I had lying around and start using it to read Spanish ebooks. This was a massive game-changer for me because of the ease with which I could look up unknown words. I no longer needed to interrupt my reading to go to SpanishDict or WordReference in order to find a definition. Instead, all I needed to do was highlight the word and get the definition without breaking the flow of the book. As a result, I was able to do intensive reading at essentially the same speed that I did free-flow reading previously, which helped my vocabulary and grammar knowledge to explode.

Even after getting the Kindle, I was still very bad about sentence mining from my reading immersion. I found the process to be extremely tedious, and so I would have a week where I’d stay on task followed by months where I would create 0 new sentence cards. I eventually stumbled upon Migaku Kindle, which marked another turning point in my reading journey. This Anki extension allows me to automatically import Anki cards for all of the word lookups I perform on my Kindle, reducing the time for sentence card creation to essentially nothing. As a result, since incorporating this add-on, I’ve created thousands of sentence cards and significantly expanded my vocabulary.

Using this approach, I now feel fairly comfortable reading the vast majority of material that I find. For example, I recently read Crónica de una Muerta Anunciada by Gabriel García Márquez while making few, if any, lookups. While there were definitely words I didn’t know, they didn’t impact my enjoyment or understanding of the story.

Takeaways:
- Read a lot to acquire more vocabulary and grammar
- Buy a Kindle or some other e-reader for fast word lookups
- If using a Kindle, make frequent use of the Migaku Kindle add-on for Anki

Listening

Stats:
Active Listening
- Netflix: 180 hours
- YouTube (estimate): 300 hours
- Podcasts (estimate): 50 hours
Passive Listening
- Podcasts (estimate): 20 hours
Total hours (estimate): 550 hours

I began my listening journey focusing mostly on watching with Spanish subtitles. As a result, at the start of my journey, listening was more of a hybrid listening/reading activity for me. Later on, as my vocabulary expanded, I began to slowly wean myself off of subtitles. I began by removing them entirely from YouTube immersion while keeping them on for Netflix immersion. I then started to remove subtitles on easier Netflix shows, while leaving them on some of the slang-heavy shows I was watching such as Élite or Pablo Escobar: El Patron del Mal. Starting around April 2021, I removed subtitles on all content and now focus purely on listening without any reading support.

Due to my heavy emphasis on reading, my listening has suffered a bit comparatively. I’m now at the point where I understand essentially 100% of vocabulary and grammar structures that are used in TV shows, podcasts, or everyday conversation, but there are many times where I simply cannot parse the sounds and understand the words that are being said. It has been improving steadily since removing all subtitle supports in April, but it’s still far below my reading ability.

I also started doing much more passive listening in the past month (I’m actually passive listening while writing this article), and I’ve noticed that this has had a huge effect on my listening ability. While I’m usually distracted by other things while doing passive listening, the little bits of language that do seep in start to add up over time. I’ve been focused mainly on understanding the Dominican dialect lately by passive listening to a wide range of Dominican podcasts, and so far I have noticed a very rapid increase in my comprehension of that accent.

In summary, if I could do things over, I would have removed all subtitles from my listening practice significantly earlier — probably sometime around January 2021. In addition, I would have attempted to passive immerse as much as physically possible since Day 1. It’s quite easy to fit in at least a couple of hours of passive listening every day, and these hours begin to add up rapidly. Lastly, and perhaps most importantly, I would make sure to prioritize compelling content over comprehensible content as much as possible. I find that when a show is way above my level but extremely good, I learn significantly more than from a show that’s comprehensible but not compelling. Don’t let difficult content scare you away — if it looks enjoyable, dive in as soon as you can.

In my experience with Refold, listening is a skill that requires a ton of hours to develop and it cannot be improved much through other means. Yes, reading helps with actually knowing the words and grammar that are being said, but it can’t help you with hearing those words in rapid speech. So make sure to invest a lot of time and energy here.

Takeaways:
- Remove subtitle supports as early as possible
- Start passive listening from Day 1 and do it every day
- It is more important for your content to be compelling than for it to be comprehensible

Speaking

Stats:
iTalki: 43 lessons
In-person conversation (estimate): 6 hours

I deviated a bit from Refold methodology here, but I ended up starting iTalki almost immediately after beginning Spanish immersion in August 2020. I decided to take the plunge in order to overcome some output-related fears I knew I had from my previous experience with Spanish. The iTalki lessons definitely helped a lot with getting over this fear, which facilitated in-person conversations once I felt ready for those.

In addition, the lessons helped expose me to gaps in my vocabulary and grammar knowledge, which I then targeted through immersion and Anki. This allowed me to systematically learn phrases which made my output more fluid and natural over time. I think this exposure of what you don’t know is probably the biggest advantage of speaking practice, so make sure to note down words or phrases that you wanted to say but couldn’t think of during your iTalki lessons.

When you do decide to start your iTalki lessons, make sure to find a friendly tutor who you feel comfortable speaking with. You will likely be spending dozens of hours conversing with this person, so it will serve you well to prioritize the quality and enjoyment of the conversation over any particular accent.

Takeaways:
- Start iTalki early if you have output-related anxiety
- Use immersion and Anki to address gaps in your knowledge that iTalki exposes
- Find a friendly tutor that you feel comfortable speaking with for hours

SRS

Stats:
Learned Cards: 671
Unseen Cards: 2760

Anki has been the biggest struggle on my Refold journey. I don’t particularly enjoy making or reviewing flashcards, so I’ve been incredibly inconsistent with using the app. I have two big recommendations here for making your life with Anki easier. Firstly, as I mentioned in the reading section, find extensions and add-ons that make your life easier. Migaku Kindle is huge for speeding up the card creation process for reading immersion. There’s also Migaku Browser Extension, which will help you make audio cards extremely quickly as well. Find as many extensions of this nature as you can so that you can reduce your card creation time as close to 0 as possible. The more effortless you make this process, the more likely you will be to complete it.

Secondly, for actually reviewing cards, try to schedule specific times within the day where you’ll use Anki. I find that if I leave it up to me to decide when I study my flashcards, I continually put it off until I’m eventually too tired to do my reviews. With a specific time, it removes the guesswork of when I should do my reviews. In that same vein, experiment with scheduling multiple review sessions per day. It can be intimidating to need to do 150 Anki reviews in a single session. However, if you split those reviews between three or four sessions throughout the day, it suddenly becomes a very manageable amount.

I highly recommend implementing these strategies to make Anki usage easier and more consistent. Whenever I’m using Anki on daily basis, I notice massive changes in my vocabulary and grammar knowledge. I can’t quite quantify how large the speed up is from using Anki, but it feels as if I’m progressing at least twice as fast, if not more, when I’m frequently making and reviewing flashcards.

Takeaways:
- Use extensions to make Anki card creation as easy as possible
- Schedule specific times through the day for Anki
- Experiment with splitting Anki sessions up throughout the day
- Make sure to actually do Anki on a daily basis. It makes a huge difference

Where Next?

My Spanish journey so far has been one of the most challenging and fulfilling projects I’ve ever undertaken. I’m unbelievably surprised with the progress I’ve been able to make, but, at the same time, I know I have much farther to go before I reach my goals with this language. I’d like to pass the DELE C2 exam, read the great works of Spanish language literature, live in a Spanish speaking country, and so much more over the coming years. I anticipate it will probably take me another 1000 hours of active listening and 30–40 additional books in order to reach that point, so I’ll have my work cut out for me. However, if the last year, has been any indication, it’ll be the most enjoyable work imaginable.

I hope my experiences on this journey have been helpful to any readers out there who are currently learning a language or thinking about starting. Feel free to leave a comment if you have any additional questions or want to know more about my experience learning Spanish. Thanks for reading and happy learning!

“Silence is a lesson learned from the many sufferings of life.”

— Seneca, Thyestes, 309

Imagine silence. Sit quietly for a moment and try to be absorbed by it yourself. Allow your mind 60 seconds of tranquility.

Were you able to do it?

During those 60 seconds, you may have been assaulted by the buzzing of your phone, the ding of emails hitting your inbox, or a deluge of anxious thoughts. Even finding a small sliver of silence within your day may have proved to be impossible.

As modernity has come upon us, a part of the endless wave of technological advances and social upheavals, we have become a world of noise. Technology connects us to the thoughts and concerns of other humans in a unprecedented ways. It is now possible to read or listen to the thoughts of millions of other minds at any moment. The triumphs, tragedies, and worries of this cacophony of voices fills every void throughout our day — it is nearly inescapable. With this stream of information turning into a roaring river of data over the last decade, we must ask ourselves: what are the dangers presented by an absence of silence?

In its absence, we risk losing contact with ourselves. The inundation of outside voices does not allow us room to engage in dialogue with our own thoughts. In The Art of War by Sun Tzu, he cautions to “know others and know thyself, and you will not be endangered by innumerable battles” with the implication that, if we are not to know our true selves, we will be endangered by innumerable battles. Our constant connectivity and distraction is thus directly threatening us, causing us to take on more battles than is necessary.

We see this scene play out amongst the many actors in our lives. Those who do not know themselves fall into the same challenges and battles day after day. The alcoholic who does not acknowledge his problem continues to lose himself in the drink. The oblivious narcissist forfeit relationships as she ignores the wants and needs of others. The clueless food addict marches unceasingly towards an early grave. But these dangers are not limited to others — you and I alike share this same lack of self-knowledge.

Examine yourself and your experiences and attempt to locate points in time where you endured a hardship due to a lack of knowledge of yourself. Think about how the understanding provided by self-reflection can prevent other such occurrences. How would your life improve if you could avoid these battles?

You can only truly know by attempting it yourself. Set aside time for silence. Turn off the computer, leave your phone at home, and free yourself from distractions for just a short while. Allow yourself time to engage with the silence and, as a result, with your inner self. You may be amazed at what you find.

We all go through tragedy or hardships at some point in our lives. It’s an unfortunate fact of existence. When we experience these hardships, the sense of loss can be profound. In the extreme cases, when we lose a loved one or watch a valued relationship collapse, it can feel as if part of us is dying. And that’s because it is. Our close relationships feature as prominent parts of our personal story. They make up an important part of who we tell the world (and ourselves) that we are. Without them, we have to retell our personal story and discover who we are without them in our lives.

In addition, when we imagine the future, we almost always include some of our closest relationships in those plans. Losing a significant other can destroy the world we constructed in our mind’s eye more effectively than a nuclear blast could. It can cause us to feel directionless and like a major part of our identity has gone up in flames.

Tragedy is like a forest fire. If allowed to rage unchecked, it can destroy the whole forest and everything living within it. But if controlled, it can clear the way for new, productive growth.

It is important that we take action and seek to fill in the void that the forest fire has left. The true dangers lie when tragedy strikes, but we do nothing to control the flame and create new growth. This passivity leaves a gaping void in both our personal story and our vision for the future, making us feel lost, afraid, and directionless. These feelings can then spiral out of control and drag us into the abyss of depression and lethargy.

However, the destruction of our previous identity also allows us room to craft a new narrative about ourselves. We are no longer tied to the expectations of others or the concrete visions we’ve had about the future. We’re free to throw the canvas away and start anew, painting the picture we truly want. This sentiment is captured well by Marcus Aurelius,

Think of the life you have lived until now as over and, as a dead man, see what’s left as a bonus and live it according to Nature. Love the hand that fate deals you and play it as your own, for what could be more fitting?

— Marcus Aurelius, Meditations, 7.56–57

So do not lose hope in the face of tragedy or hardship. Persevere and use the opportunity to rewrite your story. You just might create your masterpiece.

Motivation is a fickle creature. One moment it’s burning in your belly like a fire, pushing you on to complete your task, and the next it’s vanished entirely. As a result, relying on motivation in order to get things done is very likely to be doomed from the start. When motivation leaves us in the middle of a project — whether it be for school, work, or a hobby — we feel lost, as if we’re drifting at sea with no direction.

These feelings can be extremely detrimental to our mental state and productive effort. More importantly, the anticipation of these feelings can cause us to engage in a form of self-sabotage. If we know that our motivation will leave us during a project, why even start it? We begin to withdraw into ourselves, avoiding new tasks and experiences because we know we won’t follow through on them. We start to actively give up, even before the fire has left us, in order to avoid the embarrassment that occurs when motivation leaves us without our permission or foresight. We start to wonder if we’re broken because of our consistent loss of motivation — if we’re doomed to always start new projects, never finishing them.

The key is to stop relying on motivation. We will never be able to control when motivation arises and when it abates. What we can control are the systems we put in place to support us in the pursuit of our goals when our motivation leaves us — or when it never appears at all.

These systems, or habits, allow us to automate our productivity by prioritizing the key components of our projects. We do this by first identifying the small steps that can be taken each day in order to reach towards our goals and then decreasing them to the simplest possible form. For example, if your goal is to be a reader, you know that you need to read every day to cultivate that identity. We have a few options we can use to decrease “read every day” into a simpler form — we can change it to “read one page of a book every day”, “read one sentence from a book every day”, or even just “open a book every day”. We now have a daily objective which is much simpler to implement and accomplish, decreasing the effort required to continue making progress on our project.

Even more importantly than the decreased perceived effort, we’ve now removed all ambiguity in how to make progress on a day-to-day basis. A large problem with motivation, and the subsequent loss of it, is that it provides no direction on where to go. It simply tells you, “I need to go somewhere.” By identifying our overarching goal and then breaking it down into easily-achievable daily microgoals, we provide ourselves with a detailed roadmap showing where we’re going and how to get there. Try implementing these microgoals in your daily life in order to work towards some of your larger ambitions, and allow the roadmap to lead you to your destination.