For good reason: I was laying the foundations for what this newsletter is going to become.

My job is to turn over rocks. To analyze companies obsessively until I find one that appears genuinely exceptional, then take it apart from top to bottom until I understand it better than almost anyone else.

That is the direction Undiscovered Compounders is going to take.

I will now publish my investment theses on the exceptional companies I find: small caps and micro caps buried in obscurity, each with its own dose of alpha.

The idea is simple. Investors want alpha. My job is to find it. And I am going to charge for the work.

But before asking anyone to pay, I need to show what the work is worth.

That is the point of these past few weeks of silence.

On Saturday, I will publish 2 pieces.

1. The first is an investment thesis on a company that is probably the best risk/reward setup I have seen in my career. It’s 40,000 words. Around three and a half hours of reading.
   I detailed everything: the market, the project, the numbers, the valuation, the risks, the downside, the upside, every assumption I could check, and every place where I could be wrong.
   I spent hundreds of hours on this thesis. After that work, and the feedback I received from people I trust, I am comfortable saying this is the kind of research you rarely find on Substack.
   That piece will be free.

2. The second piece is a better way to play the same setup. Objectively better, but with a different risk profile. It rests on the first thesis, which already does the heavy lifting. So this one is only 8,000 words.
   It’s a setup inside the setup. And it is a good example of the kind of alpha that can still be found when you look where very few investors are looking.
   That piece will be behind the paywall.
   

The free side of this newsletter will remain serious. It has to. It’s where trust is built. But paid subscribers will get the deepest company research and most of the alpha opportunities I find.

I can’t promise a fixed frequency of investment theses. Turning over rocks is too messy for that.

But now that my investing work and this newsletter overlap almost entirely, I can promise this: about 90 hours a week spent looking for alpha, with the best of what I find turned into publishable work.

This newsletter is only the first step in something larger.

I am building with a time horizon measured in decades, and the past month and a half was spent laying the foundations for that.

But one thing at a time.

For now, the point is simple: the reputation I am building is my best asset. I intend to protect it, and build everything around it.

I will try to prove that on Saturday.

Take care,
Flo

Subscribe now

Share

The marks left by the climb up Mount Stupid, the fall into the Valley of Despair, and the slow walk up the Slope of Enlightenment are easy to spot. The precision of the vocabulary. The pauses at the right moments. The disarming honesty about their own limits.

In investing, those signals are gold. And a trap.

Let me tell you what happened to me.

Subscribe now

A Clever Guy

There’s a fellow investor I talk with regularly about investing. Let’s call him John. I know John in person. So I can honestly say he radiates intelligence. In his words, but also in his eyes. Two seconds are enough to guess that his genetics gave him a very efficient neural architecture. Same thing on temperament. He’s no Buffett, but he’s deeply composed, quick to delay gratification. The kind of guy who doesn’t flinch when his portfolio does.

Recently, we were discussing a microcap in the commodities sector, one John knows inside out. The position is a significant part of his portfolio. I listen carefully. I ask questions, and the guy has answers. Good enough that I decided to dig into the name on my own.

The company is a royalty business: it provides cash and its own shares (incentives, the whole package) in exchange for a future cut of the counterparty’s revenue. Fair enough, it fits my thesis that commodities will outperform in the coming years, and it’s a cash-flow business model, which isn’t that common in the sector.

A few days later, I open the MD&A. And very quickly, something smells off. The deal quality, first of all, is mediocre at best. Most of the counterparties are pre-revenue (so no cash flow on those deals), and a few carry going concern notes. On top of that, the company’s cost of capital is high: a structural disadvantage against the majors, who raise capital far more cheaply. One hour in, I knew the name didn’t deserve more of my time.

Still, I’m a bit stunned. The company John pitched me is well below his usual bar. Probably the worst one, actually. And no, dear “value” investors, it doesn’t even have the excuse of being cheap. I’d even call it overvalued.

I walk away from this analysis in complete cognitive dissonance: my read of the company on one side, my read of this good, maybe very good, investor on the other.

Here We Go

A few weeks later, I catch up with him. You probably expect a bloody confrontation, but reality is more sarcastic. In the meantime, the stock has doubled. What was I supposed to say? “Hey John, I spent an hour on company X, it’s honestly a bad business, I’m not sure how you got into it. And calm down, the 2x doesn’t count: judge the bet, not the result.” Obviously not.

But I push back, carefully. I raise my concerns one by one. And of course, he has an answer for each of them.

The bad deals? Just the price of admission. For now, only two major deals anchor the company; the rest are pure optionality, “free lottery tickets”. The cost of capital? Backed by two well-known investors who came in before the IPO and refinance whenever needed, which means the real cost of capital is low. The overvaluation? I’m missing the point. The company signs deals the majors don’t even compete for. That means a pipeline with no competitors, and counterparties with no real incentive to negotiate hard.

Every argument unpacked, every point explained. He knows his subject, and believe it or not, I walk away convinced. I know from experience that sometimes, the best setups are the ones whose first layer is ugly enough to hide the prettier layers underneath.

This time I don’t wait. I start digging the same day. Two days later, I’m more stunned than before. The company isn’t just worse than I thought, it’s worse in ways I hadn’t looked for.

First, the contracts. The company signs deals across multiple countries, each with its own legal framework. In some jurisdictions, the obligation to honor the royalty sits far below other creditor claims. Translation: in many distress scenarios, those deals are close to worthless.

Then the CEO. He takes a salary that’s astronomical for a company this size, while holding a small equity stake. He isn’t even a key man: he doesn’t draft the contracts, he doesn’t have a deep network of potential counterparties. Even worse, a few years earlier he had created a nearly identical company that went bankrupt.

To drive the nail in, those big investors who have backed the company from the start have a track record of taking large positions in names in the sector and averaging down all the way to bankruptcy.

No is too weak a word. On top of that, I have the bitter taste of having wasted my time.

Here We Go Again

When I reanalyze the situation, the cognitive dissonance is strong enough to give me a migraine. It didn’t take me that long to see all of this, so there’s no doubt John is aware of it too. But I have absolutely no idea how he got past it. I even considered that he might be pulling a prank from the start. Unfortunately, too elaborate to be plausible.

Round two, postponed. I haven't talked to him since.

But the ending isn't that disappointing. Remember, reality is funnier than that: the stock has doubled again since our last conversation. That’s a 4x in a few months. And I’m the one who has to tell him his company is rotten. I have no idea how to approach the next conversation. But I’ve at least walked away with one lesson.

Before this story, if Buffett had called me saying, “here’s my best idea, I can guarantee you 50% a year if you hold for five years,” I would probably have done less research than if anyone else had pitched it.

Once you’ve learned to ignore the majority, it becomes all the more tempting to give weight to the people you respect. Yet most borrowed convictions are borrowed without you knowing it. Only time and volatility reveal what they really were: borrowed.

It doesn’t matter how smart or how accomplished the person in front of you is. You never know all the hidden reasons behind a decision: the constraints, the assumptions, the path taken to get there.

Their best idea can be your worst.

Take care,

Flo

Subscribe now

Share

I give you some money, a crystal ball, and a time machine.

The rules. Each round, you land on a random date between 2008 and 2022.1 You’re shown the front page of the Wall Street Journal from the next day (prices blacked out; it’s a game, not a cheat). You read. You bet on the S&P 500, on 30-year Treasuries, on both, or you skip the round. Maximum leverage allowed: 50x. You close your position at the next day’s close.

15 rounds. 45 minutes. Go.

The Results

118 finance students played this game with real money on the table (only $50, but you’ve got to start somewhere).

They know what a Fed decision does to the yield curve. They know why a strong dollar weighs on multinational earnings. They’ve spent years learning exactly what they needed for this game, so we should expect a strong performance.

And they delivered: Average return: 32%. Only 4.5% lost money.

Just kidding. Average return: 3.2%. 45% lost money. 1 in 6 went bust. With tomorrow’s front page in their hands.

Across the 2,000 trades placed by the 118 players, the directional hit rate was 51.5%. A coin flip.2 The information was there, they just couldn’t read it, on average (1 in 5 still managed to double their capital). Bad reading, fine. But how did they bet on it?

On 30% of trading days, players used leverage above 20x. And the correlation between bet size and actual hit rate across days was 0.628. At least they knew exactly when to bet big.

Just kidding, again. The correlation is zero. They had no idea about the quality of their own convictions, or what to do with them.

So the full picture is (on average): they couldn’t read the information they were handed (literally the future), and they bet big at random without even knowing they couldn’t read it.

It would have been interesting to run the same experiment with professional macro traders for comparison. But guess what, that’s exactly what the authors did.

Let’s Play a Game, Again

They invited five seasoned macro traders to play. The head of trading at a top-5 US bank. The founder of a top-10 macro hedge fund. Former seniors from Jane Street and from the Treasuries desk of a primary dealer. Let’s call them the veterans.

None of the veterans finished in the red. Average return: 130%. Median: 60%. No, I’m not kidding this time.

Two things separated the veterans from the students.

First, they read the paper better. 63% directional accuracy, against 51.5% for the students. The difference between a coin and an edge. That’s what you build over twenty years of watching how markets actually react to the news.3

Second, they knew when to skip. On roughly a third of trading opportunities, they did nothing. On the rest, they went big.

“Soros taught me that when you have tremendous conviction on a trade, you have to go for the jugular. It takes courage to be a pig.” — Stanley Druckenmiller

The 118 students had tomorrow’s paper and they burned it. The five veterans had the same paper and they made 130% in fifteen days.

Information is just an input. It tends to get stuck in the bottleneck between the screen and the chair.

Tomorrow, the input will be even more abundant. But the bottleneck won't widen on its own.

We are our own edge. Always have been.

Take care,

Flo

Share

I dug through decades of speeches, letters, and interviews of Charlie Munger to extract the mental models that, in my view, are the real gears behind his success.

I found 40. Here they are.

1. Reason-Respecting Tendency

You want to ask someone for something and maximize the odds they say yes. Tell them why. That’s it.

The advice could stand on its own. But humans are more… surprising than that. The probability increases even when the reason makes no sense.

In 1978, psychologist Ellen Langer ran a study where researchers tried to cut in line at a university copy machine using three different phrasings. Here are the results:

• Simple request: “Excuse me, I have 5 pages. May I use the Xerox machine?” → 60% compliance

• Request + real reason: “…because I’m in a rush” → 94% compliance

• Request + nonsense reason: “…because I have to make copies” , completely tautological, everyone is there to make copies → 93% compliance

The real reason and the nonsense reason are virtually identical in effect. Replications showed weaker results for nonsense reasons, but still consistently higher than no reason at all.

But still, what a result!

It’s rare for Charlie to cite a study to support his point. But when he does, he picks well:

“Unfortunately, Reason-Respecting Tendency is so strong that even a person’s giving of meaningless or incorrect reasons will increase compliance with his orders and requests. This has been demonstrated in psychology experiments wherein “compliance practitioners” successfully jump to the head of the lines in front of copying machines by explaining their reason: “I have to make some copies.” This sort of unfortunate byproduct of Reason-Respecting Tendency is a conditioned reflex, based on a widespread appreciation of the importance of reasons. And, naturally, the practice of laying out various claptrap reasons is much used by commercial and cult “compliance practitioners” to help them get what they don’t deserve.”
— Poor Charlie’s Almanack, Peter D. Kaufman

2. Human = Gamed

You probably know Charlie’s famous line: “Show me the incentive and I’ll show you the outcome.” But that's just the punchline. The reasoning behind it is what makes it powerful.

His reasoning goes like this:

• Every human system (family dynamics, tax codes, traffic laws) is built on arbitrary rules.

• Each person will interpret and evaluate those arbitrary rules differently.

• And from those interpretations, incentives emerge.

For example: I know I’m supposed to stop at a red light. But it’s 2 AM, no one’s around, and I won’t get caught. I run the light.

Scale that logic to every participant in every system, and you get Charlie’s premise:

“All human systems are gamed, for reasons rooted deeply in psychology, and great skill is displayed in the gaming because game theory has so much potential.”
— Charlie Munger, University of California Speech at Santa Barbara, 2003

People are not “intrinsically dishonest or bad”, it’s just that the rules are arbitrary and the incentives to bend them are not.

3. Cognitive Eccentricity

You probably know the famous anecdote about Charles Darwin: whenever he encountered a fact that contradicted his ideas, he would write it down immediately in a notebook, because he knew his brain would do everything in its power to forget it or distort it.

150 years later, cognitive psychology has obviously proven Darwin right. Our brains are wired to seek information that confirms our priors and reject what refutes them. Ironically, this is an adaptive reflex that served our evolutionary survival well. In a world where changing your mind about which berries are poisonous could kill you, stubbornness was a gift.

Where Charlie brings his wisdom is in how far this idea can carry someone. Doing what others are cognitively unwilling to do leads to outsized results: actively seeking to contradict your own ideas, being disliked without flinching, persisting through a string of failures, sitting still and doing nothing for decades. These are cognitive overrides, and the people who master them tend to end up in the history books, for better or for worse.

Ironically, Charlie Munger is himself one of the best examples of a “cognitive eccentricity” success story.

4. The Tolstoy Effect

Tolstoy observed that even the worst criminals don’t see themselves as bad people. They end up believing either that they didn’t commit their crimes, or that given the pressures and disadvantages of their lives, their behavior was understandable and forgivable.

Nobody is the villain of their own story. Charlie turns this observation into a principle:

“The Tolstoy effect, where the man makes excuses for his fixable but unfixed bad performance is bad character and tends to create more of itself, causing more damage to the excuse giver with each tolerated instance.”
— Poor Charlie’s Almanack, Peter D. Kaufman

Charlie does offer an antidote:

“The best antidote to folly from an excess of self-regard is to force yourself to be more objective when you are thinking about yourself, your family and friends, your property, and the value of your past and future activity.

This isn’t easy to do well and won’t work perfectly, but it will work much better than simply letting psychological nature take its normal course.”
— Poor Charlie’s Almanack, Peter D. Kaufman

5. Psychology of Human Misjudgement

Life is a human game. Practical psychology is the rulebook of that game, a rulebook that isn’t taught in school. One of Charlie Munger’s biggest learning regrets, by his own admission (one he spent the rest of his life correcting).

This discipline was so important to him that he did something he never did for anyone else: he gave one Berkshire Hathaway Class A share to Robert Cialdini, author of Influence: The Psychology of Persuasion, for ‘his contribution to mankind’ (roughly $250,000 at the time).

So the most useful and practical part of psychology-which I personally think can be taught to any intelligent person in a week-is ungodly important. And nobody taught it to me, by the way. I had to learn it later in life, one piece at a time. And it was fairly laborious. It’s so elementary though that, when it was all over, I just felt like a total horse’s ass. And yeah, I’d been educated at Caltech and the Harvard Law School and so forth. So very eminent places miseducated people like you and me. The elementary part of psychology-the psychology of misjudgment, as I call it-is a terribly important thing to learn.
— Poor Charlie’s Almanack, Peter D. Kaufman

I have a background in psychology, and I can say that not a single day has passed without it being useful to me. Definitely worth a few hundred or thousand hours of your life. Subscribe and let me do some of the work for you.

Subscribe now

6. Availability-Misweighing Tendency

Here, Charlie is generous; he gives both the problem and its antidotes:

“Man’s imperfect, limited-capacity brain easily drifts into working with what’s easily available to it... An idea or a fact is not worth more merely because it is easily available to you.”
— Poor Charlie’s Almanack, Peter D. Kaufman

The main antidote to miscues from Availability Misweighing Tendency often involves procedures, including use of checklists, which are almost always helpful.

Another antidote is to behave somewhat like Darwin did when he emphasized disconfirming evidence. What should be done is to especially emphasize factors that don’t produce reams of easily available numbers, instead of drifting mostly or entirely into considering factors that do produce such numbers.

Still another antidote is to find and hire some skeptical, articulate people with far reaching minds to act as advocates for notions that are opposite to the incumbent notions.
— Poor Charlie’s Almanack, Peter D. Kaufman

The few times Charlie mentioned this tendency, he always used everyday examples, never investing. Yet it’s one of the most common and most costly mistakes in finance. I wrote a deep dive on exactly that:

7. Granny’s Rule

Granny’s Rule is the requirement that children eat their carrots before they get dessert.

Applied more broadly: do the unpleasant tasks before the pleasant ones. Pure behavioral engineering.

It’s a simple idea, but Charlie took it very seriously.

“‘Granny’s Rule’ provides another example of reward superpower, so extreme in its effects that it must be mentioned here. You can successfully manipulate your own behavior with this rule, even if you are using as rewards items that you already possess! Indeed, consultant PhD. psychologists often urge business organizations to improve their reward systems by teaching executives to use “Granny’s Rule” to govern their own daily behavior.”
— Poor Charlie’s Almanack, Peter D. Kaufman

Now, if I hadn’t told you this was Charlie Munger, you’d probably think I was being condescending. Eat your carrots before dessert, really?

And that’s precisely what makes Charlie Munger’s mental models so popular. Here’s someone who achieved more than most of us will ever aspire to, explaining that ideas this simple actually work. Ideas we tend to dismiss precisely because they’re simple, when they’re not backed by a lifetime of success to vouch for their quality.

That dismissal is a massive mistake. The quality of a simple idea should not depend on who said it. And that’s one of the traits that made Charlie who he is: the ability to recognize a good idea regardless of where it comes from, and apply it across a wide range of situations. His pool of good ideas was therefore larger than most people’s, and the arsenal he built from it compounded faster.

So let me use this Granny’s Rule to offer a mental model of my own using Charlie as an example: judge the quality of an idea independently of who said it.

8. Two-Track Analysis

Charlie believed almost every decision should be analyzed from two angles, the rational and the psychological:

“Personally, I’ve gotten so that I now use a kind of two-track analysis. First, what are the factors that really govern the interests involved, rationally considered? And second, what are the subconscious influences where the brain at a subconscious level is automatically doing these things-which, by and large, are useful but which often misfunction?

One approach is rationality-the way you’d work out a bridge problem: by evaluating the real interests, the real probability, and so forth. And the other is to evaluate the psychological factors that cause subconscious conclusions-many of which are wrong.”
— Poor Charlie’s Almanack, Peter D. Kaufman

I’ll add a third track, one that’s often forgotten and too easily dismissed as irrationality: constraints.

Constraints of time, means, motivation, interests, power, etc. When a decision seems irrational, asking what constraints might have caused that behavior is often wiser than stopping at “irrationality.” Nine times out of ten, the person wasn’t being irrational, they just didn’t have the options people think they had.

9. Survival of the Simplest

“Take a simple idea, but take it seriously.”
— Charlie Munger

This one is special. It’s probably one of Charlie’s most insightful mental models, but it suffers from the very defect that makes it so insightful in the first place: it’s too simple. Ironically, this mental model is itself a simple idea, worth taking seriously.

Our brain doesn’t perceive reality. It perceives change. We don’t feel heat; we feel a change in heat. We don’t hear volume; we hear a change in volume. Psychophysics calls this the Weber-Fechner law: our sensitivity is proportional to the magnitude of the stimulus. The bigger the reference point, the less we notice small variations.

This wiring has a decisive consequence for decision-making: we systematically neglect small, consistent actions because they feel insignificant compared to the outcome we’re chasing.

Who invests $100 a week for 30 years to become a millionaire? Almost nobody. The action is too small relative to the goal. So people look for more spectacular paths that feel proportional to the ambition: lottery tickets, speculation, the next big thing.

Anyone who can take a simple idea with modest short-term consequences and commit to it seriously over a long period will inevitably become exceptional. An athlete, an investor, a musician.

“Think about everything in the simplest way, and act with seriousness.”
— Charlie Munger

10. N-Order Effects

When Medicare was first introduced, a panel of experts (including PhD economists) projected its cost using simple extrapolations of past data. They were off by a factor of more than ten.

New incentives changed behavior, changed behavior changed costs, changed costs changed everything downstream. Basically, they rounded Pi to 3.2 and called it a forecast.

“Too little attention is given to second order and even higher order effects. The consequences have consequences, and the consequences of the consequences have consequences, and so on.”
— Charlie Munger, University of California Speech at Santa Barbara, 2003

11. Latticework of Models

Isolated facts are dead weight. A fact only becomes useful when it’s connected to other facts through a structure (a theory, a model, a pattern). Without that structure, you’re just memorizing. With it, you’re thinking.

“You can’t really know anything if you just remember isolated facts and try and bang ‘em back. If the facts don’t hang together on a latticework of theory, you don’t have them in a usable form.”
— Charlie Munger, Talk at USC Business School, 1994

Charlie estimates he uses about a hundred models. That sounds like a lot. But most of the heavy lifting is done by a handful, borrowed from math, psychology, physics, biology, and engineering. The rest builds on top.

“You have to learn many things in such a way that they’re in a mental latticework in your head and you automatically use them the rest of your life. If many of you try that, I solemnly promise that one day most will correctly come to think: ‘Somehow I have become one of the most effective people in my whole age cohort.’ And, in contrast, if no effort is made toward such multidisciplinarity, many of the brightest of you will live in the middle ranks, or in the shallows.”
— Charlie Munger, USC Commencement Speech, 2007

I’ve started building this latticework in public: 50 mental models extracted from across all disciplines. 15 are published, 35 to go. I’ll release them all at once, when everything is final. Months of work left. One inch at a time. Be there when it lands.

Subscribe now

12. Physics Envy

What discipline doesn’t envy the natural elegance of physics? A few variables, a few operations, an equals sign, and there it is, a beautiful formula for the history books.

But the further down the chain of causes and consequences a discipline sits, the more illusory those formulas become. Anything involving human behavior falls into that category. Economics obviously qualifies. And yet it’s probably the discipline most jealous of physics. This isn’t a new problem: a century ago, Keynes was already making the same reproach to his peers.

“To convert a model into a quantitative formula is to destroy its usefulness as an instrument of thought.”
— John Maynard Keynes

Charlie Munger was one of the most vocal critics of this physics envy in economics, and he never missed an opportunity to remind people of it throughout his life.

Envy has never built anything, in academia or anywhere else.

13. Overweighing-What-Can-Be-Counted

Charlie said it better than I could:

“You’ve got a complex system and it spews out a lot of wonderful numbers that enable you to measure some factors. But there are other factors that are terribly important, [yet] there’s no precise numbering you can put to these factors. You know they’re important, but you don’t have the numbers. Well practically everybody (1) overweighs the stuff that can be numbered, because it yields to the statistical techniques they’re taught in academia, and (2) doesn’t mix in the hard-to-measure stuff that may be more important. That is a mistake I’ve tried all my life to avoid, and I have no regrets for having done that.”
— Charlie Munger, University of California Speech at Santa Barbara, 2003

This is exactly how Wall Street operates. The entire industry is built on what can be measured, and almost structurally blind to what can’t. Culture, management quality, incentive alignment, optionality, no Bloomberg terminal spits those out. Which is precisely where the edge is.

14. Surfing

Sometimes, winning is just a matter of surf: spot a wave, position yourself, and ride it.

According to Charlie, these skills are all the more essential in an era where technological revolutions chain together.

“When technology moves as fast as it does in a civilization like ours, you get a phenomenon which I call competitive destruction… You either get into a different business or you’re dead—you’re destroyed… And when these new businesses come in, there are huge advantages for the early birds. And when you’re an early bird, there’s a model that I call “surfing”—when a surfer gets up and catches the wave and just stays there, he can go a long, long time. But if he gets off the wave, he becomes mired in shallows…”
— Charlie Munger, talk at USC Business School 1994

He said this in 1994. 30 years later, the waves have only gotten bigger. And coming from a man who built his fortune on patience, that's not nothing.

15. The Social Cost of Good Decisions

Saying no when everyone around you is saying yes has a cost that most people aren't willing to pay.

“A lot of success in life and business comes from knowing what you want to avoid: early death, a bad marriage, etc. Just avoid things like AIDS situations, racing trains to the crossing, and doing cocaine, etc., develop good mental habits… If your new behavior earns a little temporary unpopularity with your peer group then the hell with them.”
— Poor Charlie’s Almanack, Peter D. Kaufman

By choosing your peer group well, you can reduce the social cost, or even turn it into a gain. Which makes it a huge incentive to choose your peer group carefully, even if it means changing it entirely.

Share

16. The Human Brand Moat

A brand reduces friction. You trust it, so you stop searching. Coca-Cola, Apple, Costco, you buy without thinking twice.

Charlie argued that the same moat applies to people, and must be actively built:

“You want to deliver to the world what you would buy if you were on the other end. There is no ethos in my opinion that is better for any lawyer or any other person to have. By and large, the people who’ve had this ethos win in life, and they don’t win just money and honors. They win the respect, the deserved trust of the people they deal with. And there is huge pleasure in life to be obtained from getting deserved trust.”
— Poor Charlie’s Almanack, Peter D. Kaufman

17. Business > Manager

2026 will be a big year for IPOs, or, as some on Wall Street call them, “It's Probably Overpriced”, and most of them will be sold on the founder's story.

Charlie's warning has rarely been more relevant:

“Averaged out, betting on the quality of a business is better than betting on the quality of management. In other words, if you have to choose one, bet on the business momentum, not the brilliance of the manager.”
— Charlie Munger, Talk at USC Business School, 1994

18. The Cancer Surgery Formula

When something in your life is a mess: a career, a relationship, a set of habits, a portfolio, the basic instinct is to try to fix everything. According to Charlie, the wiser move is the opposite: find what's sound, and cut away everything else.

It’s the mirror image of inversion: instead of asking “what should I add to make this work?”, ask “what should I remove so that what already works can breathe?”

“They [GEICO] look at this mess. And they figure out if there’s anything sound left that can live on its own if they cut away everything else. And if they find anything sound, they just cut away everything else. Of course, if that doesn’t work, they liquidate. But it frequently does work.”
— Charlie Munger, Talk at USC Business School, 1994

At its core, the Cancer Surgery Formula is just a concrete application of one of Charlie's deepest convictions:

“It is remarkable how much long-term advantage people like us have gotten by trying to be consistently not stupid, instead of trying to be very intelligent.”
— Charlie Munger, Berkshire Hathaway Meeting 2001

Cancer surgery is just consistent non-stupidity with a scalpel.

19. The Whirlpool Principle

“Some Scandinavian canoeists succeeded in getting through all the rapids of Scandinavia, and they thought they would continue their success by tackling the big whirlpools in northwest America. The death rate was one hundred percent.”
— Charlie Munger, Harvard School Commencement Speech, 1986

They died because they assumed the skill they had was the skill they needed.

Charlie adds a second layer: the canoeists went as a group. Each one’s confidence reinforced the others’. In this example, social proof turned individual overconfidence into collective suicide.

“A big whirlpool is something you want to avoid. And I think the same is true about intense ideology, particularly when your companions are all true believers.”
— Charlie Munger, Harvard School Commencement Speech, 1986

Charlie spent so much of his life denouncing ideology that I don't even need to look for another quote. He's already said it all.

20. Learning as a Categorical Imperative

This is one of the purest forms of mental model there is, a heuristic: a way to make an optimal decision very quickly and at minimal cost.

No more hesitating between leisure and learning, just learn. It’s a moral duty. A “universal law,” as Kant would say.

Of course, it’s just a cognitive trick, a way of dictating to ourselves the behavior to follow without thinking. But it’s a trick that depends primarily on our willingness to make it work (or not), which is pretty remarkable when you think about it.

Wisdom acquisition is a moral duty.

Charlie left the perfect ending himself:

“I hope these ruminations of an old man are useful to you. In the end I’m like old Valiant-for-truth in The Pilgrim’s Progress:

My sword I leave to him who can wear it.”

Rest easy, Charlie. The sword found hands.

Take care,

Flo

Subscribe now

Share

Half of Charlie's mind is here. The other half is one click away:

• ChatGPT, rate this company knowing it has a 92% chance of beating its targets.
  Score: 5.3/10.

• ChatGPT, rate this company knowing it has an 8% chance of missing its targets.
  Score: 3.6/10.

You read that right, both prompts say the same thing: same company, same probability. Only the wording changed. But the response shifted by 44%. The AI did exactly what you would have done: it reacted to the words, not the numbers.

That was one bias in one model. Researchers1 stress-tested every major model: GPT, Claude, Gemini, LLaMA, Mistral, etc. 48 models, 11 biases2 tested, 26,000+ observations.

Framing was only the beginning. Some biases don’t just distort scores; they reverse decisions entirely.

And no, upgrading to a more powerful model won't save you. But the researchers found what will.

Deeper Than You Think

What follows isn’t a collection of edge cases. Every result is the average across all 48 models tested.

You submit an argument to the AI and ask it to evaluate its quality. Word for word identical, except in one version it comes from a PhD student, in the other from a Nobel laureate. The score jumps 34%.

Switch the test: ask AI to evaluate a stock, then run the same prompt with “trending #1 on Twitter” added. The rating jumps 24%. No tweet has ever improved a company’s fundamentals. The AI “knows” this. It still adjusts. Just like a human would.

But those biases only distort a score. This next one inverts a decision.

A project will return $800K, but it costs $1M to finish. The AI says no. Now mention that “$10 million has already been invested in this project.” The no becomes a yes. That $10 million is a sunk cost, irrecoverable and irrelevant to the decision. But it’s enough for the AI to recommend losing another $200K, and your time.

Every way you phrase a question to AI is a potential way to bias its answer, and nothing in the output warns you. The AI itself doesn’t know.

A good analyst learns from his mistakes. He loses money on a sunk cost, remembers it and corrects. AI has read millions of sunk cost decisions, but never paid the price of a single one. It has the reflexes without the feedback. It’s an analyst who’s read every book but never had a P&L.

If the problem is the machine, the solution is a better machine. Right?

Better Model, Better Results?

The answer is yes. But no.

More powerful models resist framing and narrative biases better. But they worsen sunk cost, loss aversion, and disposition. Upgrading fixes some flaws and creates others that less capable models never had.

Intelligence doesn’t eliminate biases. It redistributes them.

And it always will, because biases don’t enter through a single door. They enter through all of them, from the creation of the training data to the final output.

But the most striking finding isn’t about bias levels. It’s about what the models pay attention to.

By analyzing how models weight different types of signals, the researchers discovered that each capability level corresponds to a distinct investor profile:

• Less capable models overweight recent performance, social signals, and whatever is “trending.” They invest like momentum traders following the crowd.

• More capable models overweight fundamentals, valuation multiples and cash flows. They invest like fundamental analysts.

Read that again. AI models aren’t just biased, they develop investment styles. And those styles mirror the spectrum that separates retail traders from institutional investors.

What does that tell us about the link between intelligence and investment style? I'd love to hear your take in the comments.

Leave a comment

But regardless of what it means, the practical consequence is immediate: when you choose between GPT, Claude, or Gemini, you’re not just choosing a tool. You’re choosing a directional bias that will color every answer, without ever telling you.

The question that remains: how do you neutralize what you can’t see?

Debiasing

The obvious fix would be to just tell the AI to be objective. It doesn’t work. Same way telling a human “be rational” doesn’t work. Fortunately, the universe is more subtle than that.

The bias isn't just in the reasoning. It's in the input, and in how the model processes it. That distinction is the key to everything that follows.

The researchers found two fixes: control what the AI sees, and rewire how it thinks about what it sees.

Step 1: Clean the input

The principle is simple: make the AI sanitize your prompt before it responds.

You explicitly ask it to strip anything that could bias the answer: purchase prices, amounts already invested, prestigious names, your own opinion. Any signal that isn’t relevant to the decision. What remains is the bare problem.

Here’s the prompt used in the study, adapted to work in virtually any situation:

Before answering, rewrite my prompt by removing any element that is not strictly necessary to the decision: purchase prices, amounts already invested, time already spent, source names, authority signals, and my own opinion. Keep only: the financial data relevant to the future decision, the core question, and the response format. Show me the rewritten prompt, then answer it.

Framing drops to nearly zero. Authority bias falls by 72%. Representativeness by 77%.

The logic is simple: if the trigger isn’t in the prompt, the bias doesn’t fire.

But preprocessing has a blind spot. It’s surgical on information biases (framing, authority, narrative), but it barely moves the needle against social biases (herding, availability, sycophancy). Even when you explicitly target them.

You can’t strip social pressure from a prompt, because the model doesn’t need to see it to feel it. That’s where the second method comes in.

Step 2: Rewire the reasoning

Preprocessing cleans what the model reads. Context engineering changes how it thinks about what it reads.

Instead of removing bias triggers, you give the model an explicit decision framework, a set of criteria that override the pull of whatever noise remains in the data. Think of it as giving the AI a checklist it must follow before it’s allowed to have an opinion.

Ideally, this goes in the system prompt3:

You are a rational financial analyst. Before answering, identify and disregard the following in my prompt: past costs and prices (irrelevant to forward-looking decisions), source names and authority signals (judge arguments on evidence, not credentials), my stated opinions (analyze independently of what I believe), social signals and popularity metrics (consensus is not evidence), and emotional or asymmetric framing (treat gains and losses symmetrically). Base your answer strictly on fundamentals, expected future value, and available evidence.

Herding, the bias that resisted preprocessing the most, drops by 94%. Availability by 98%. Sycophancy by 65%.

Neither method solves everything alone. But combined, they cover the vast majority of the eleven biases tested:

Subscribe now

One last tool, this one personal. It’s a reflex I use systematically: cross-examination.

Whenever AI weighs in on a decision, I ask it to argue against its own opinion and identify the biases that may have influenced its answer. It’s a safety net that has saved me from several costly mistakes.

Learning to use AI while understanding its biases is the equivalent of learning to navigate the internet in the 1990s. Most people didn’t bother. The ones who did had an edge for the next two decades.

The prompts, the frameworks, the cross-examination, none of it is complicated. The hard part is remembering that the AI won’t remind you.

It doesn’t know it’s biased. Exactly like you.

Take care,

Flo

“That one. Get him out. His clothes are too shabby.”

The boy obeys. He rises, steps out of the frame, and watches the rich kids hold their pose. His face burns, but he keeps it still. No protest. No tears.

He makes himself a promise: one day, he’ll be so rich that the greatest painter in the world will paint his portrait.

That kid’s name is John Davison Rockefeller.

He kept the promise. He became the richest man of his era.

Then he did something rarer than getting rich: he wrote down the operating system.

38 unfiltered letters to his son. Everything Wall Street, Harvard, and polite society would never teach him.

I read all 38.

Here’s what the richest man in American history teaches us, in words that were never meant to be public.

Subscribe now

“The Quickest Way to Harm Someone”

Rockefeller’s edge begins with an inversion.

Instead of asking how to build an empire, he asks what prevents someone from building one.

His answer fits in a single sentence, the very first lesson he passes to his son:

“The quickest way to harm someone is to give them money.”

His evidence is hard to argue with.

A study conducted in Massachusetts on 17 wealthy families. Result: none of their children ended life wealthy. 0 out of 17.

In Philadelphia, a joke made the rounds:
— He’s a self-made millionaire.
— Self-made? Sure. He inherited 20 million… and he’s got 1 left.

The joke stuck because it pointed to something real: wealth shields you from consequences. And without consequences, no resilience, judgment, or skill is ever formed.

That is why rich children so often grow up unprepared: they have been protected from the very pressures that make a person dangerous.

Urgency breeds ingenuity. Comfort breeds decay. And in that comfort, empires die, or never get built in the first place.

It’s not out of greed that Rockefeller hides his fortune from his own children for most of their lives, but out of love.

And it’s out of clarity that he lays down, in that very first letter, the principle that runs through the next 37: wealth protects nothing. It creates the conditions for its own erosion, unless it’s met with a discipline that contradicts it at every turn.

Rockefeller reduces the antidote to four terms: Dream + Failure + Challenge = Success.

Simple on the surface. Except Rockefeller does not give any of these words the meaning we usually assign to them.

And it’s in the gap between his meaning and ours that his greatest lessons live.

“I believe that faith is the father of success.”

For Rockefeller, dreaming is anything but passive.

“They mistook faith for hope.”

Hope is passively waiting for something good to happen.

Faith is the certainty of who you are becoming, and the willingness to act in alignment with it. It actively produces plans, methods, and decisions.

Rockefeller draws this distinction from dozens of conversations with defeated entrepreneurs. When asked to explain their downfall, they all ended up admitting the same thing:

‘To be honest, I didn’t think it would work.’
‘I felt uneasy before I started.’
‘In fact, it’s not too surprising that this has failed.’

Hence his conclusion, perhaps the most important one across all 38 letters:

“Disbelief is a negative force. When you disagree or have doubts in your mind, you will come up with various reasons to support your disbelief. Suspicion, disbelief, the tendency to fail subconsciously, and the lack of desire to succeed are the main causes of failure.”

Doubt manufactures failure in the background, producing hesitations, excuses, and half-measures. At the first obstacle, you suddenly have a full arsenal of reasons to quit.

The reverse is also true: confidence manufactures success.

For Rockefeller, having an unshakeable belief in yourself, faith, is the first engine of success:

“Confidence produces the attitude of believing in ‘I can do it’, and the attitude of believing in ‘I can do it’ can produce the abilities, skills, and energy. Whenever you believe that ‘I can do it’, you will naturally come up with a ‘how to solve’ method, and success is born once you successfully solve the problem.”

The order of the sequence matters: conviction → attitude → skill → energy → method → result.

Confidence is not the trophy you collect after winning. It’s the first domino.

“When I was a poor boy, I was confident that I would become the richest person in the world. Strong self-confidence inspired me to come up with various feasible plans, methods, means and techniques, and one step at a time to climb to the top of the oil kingdom.”

A poor kid who gets thrown out of a class photo, convinced he’ll become the richest man in the world. From the outside, that’s absurd. From the inside, it’s the machine that produced everything: every strategy, every acquisition, every decision.

And if conviction determines the outcome, then the size of the conviction determines the size of the outcome.

“Wealth is proportional to goals. If you have big ambitions and big goals, your mountain of wealth will rise to the sky. If you just want to pass by, you will end up in the rat race.”

“Everyone is a product of his thoughts. Thinking about small goals will lead to small results. Thinking of great goals will win great success.”

To go further than anyone, you have to aim further than anyone. It doesn’t guarantee success, but it guarantees that failure won’t be a certainty.

Have absurd goals. The cost is nearly zero. The upside can exceed anything you can imagine.

Subscribe now

“I am a clever loser.”

His son has just lost a million dollars on Wall Street. He’s been paralyzed for weeks. When he finally tells his father, the response cuts through everything:

“A failure does not show anything, and will not put the label of incompetency on your forehead.”

Then he goes further, explaining how he turned failure into an ally:

“I take failure as a glass of spirits. It is bitter when you drink it, yet it gives you plenty of vitality.”

“I am a clever loser. I know to learn from failures, draw success factors from my experience thru failure, and use innovative methods that I have never thought of before to start a new career.”

A “clever loser.” The expression is as deep as it is simple.

Rockefeller treats his failures as a means, not an end. Something he actively seeks out and uses to extract an advantage.

For him, the real danger is never failure itself. It’s what the fear of failure does to your ability to act.

“Once avoiding failure becomes your motivation for doing things, you embark on a path of laziness and powerlessness.”

The person who rationalizes inaction as caution silently loses every opportunity that action would have opened.

Rockefeller quotes Edison who, when asked about his 10,000 failures, replies: “I have not failed 10,000 times. I just invented 10,000 unworkable methods.”

Same reality, but the framing changes everything.

Failure is the same for everyone. What separates those who recover from those who don’t is the story they choose to tell themselves about it.

“Son, life is a great game.”

What made Rockefeller exceptional was his total commitment.

For him, half-measures are already a defeat.

“Most people fail not because they make mistakes, but because they are not fully committed.”

The status quo doesn’t exist. You’re either moving forward or falling behind.

But that kind of commitment only survives on one condition: lasting. Lasting longer than the doubt, longer than the fatigue, longer than everyone else.

“There is nothing in the world that can replace perseverance. Talent is not acceptable. Unprecedented talents abound, and geniuses who accomplish nothing is common.”

Genius impresses for a moment. Perseverance wins the game.

“Life is a great game. To win, you need to act, act again, and act forever!”

His entire life fits into his own four terms.

Dream + Failure + Challenge = Success.

Dream big enough to look ridiculous.
Absorb enough failure to become unbreakable.
Last long enough to become unstoppable.

The top doesn’t belong to the most brilliant. It belongs to whoever holds on the longest.

Keep playing,

Flo

Subscribe now

Share

Hard data can make those rules visible before experience does.

Here are seven rules I wish I had learned sooner.

Subscribe now

Acrophobia Is a Costly Flaw in Investing

All-time highs are neither rare nor dangerous (in general):

• On a monthly basis, the S&P 500 reaches a new all-time high in 31% of months.

• Over the past 100 years, only 2.5% of those all-time highs were followed by a deep drawdown of more than 30%.

So yes, all-time highs rarely lead to disastrous outcomes. But what about the typical case?

Simple: investing at an all-time high has historically delivered returns similar to those of investing at almost any other moment over the next 2 or 3 years, and slightly better over the next 12 months.

These data are a necessary antidote to one of investors’ most common fears: acrophobia.

Trust Human Irrationality

Over an investing lifetime:

• Investing when consumer confidence is at a peak has led to an average 12-month return of 3.9%.

• Investing when consumer confidence is at a trough has led to an average 12-month return of 24.1%.

This is the clearest evidence I have seen that some of the best moments to invest are the ones that feel the worst.

And when the world makes you hesitate, this is exactly the kind of chart that helps you press the trigger.

The Circle of Life

This is how companies are born, grow, and die.

A declining company may still post positive sales growth over the next three years, but only by continuing to destroy shareholder value.

This “circle of life” is one of the most persistent patterns in investing.

Recognizing it helps separate reported growth from economic reality.

The Cycle of War

War increases uncertainty, but not necessarily risk.

In other words, war widens the range of possible paths, without always changing the long-term outcome as much as investors fear.

Historically, war has tended to bring higher inflation, higher volatility, and higher returns.

The evidence is admittedly US-centric, which limits how far it can be generalized with confidence.

But at the very least, it shows that for the world’s dominant equity market, war has often been less bearish than investors instinctively assume.

If you can withstand the volatility, war has historically come with higher raw returns.

Subscribe now

Value Play or Value Trap

The deeper the drawdown, the longer it takes to reclaim the all-time high, and the lower the odds that it ever does.

No comforting truth here. Just statistics every investor should study before buying a stock that is down 50% or more.

One Simple Rule That Beat Buy & Hold

Build a portfolio, then repeat the same exercise every year for 25 years: sell 20% of your positions and replace them with randomly selected stocks.

But there is one twist. Each time, you either:

• sell positions at random,

• sell your best performers,

• or sell your worst performers.

What happens next tells you almost everything about how returns are really built.

Selling your winners each year is the most reliable way to underperform buy and hold.
Selling your losers each year is the most reliable way to outperform buy and hold.

Best of all, this was tested on millions of simulated portfolios built from randomly selected S&P 500 stocks, and the result barely changes whether the portfolio holds 5 stocks or 40.

No hidden condition. No clever trick.

As so often in investing: simple, but not easy.

(Click here for a deep dive on the study)

“Even God Would Get Fired as an Active Investor”

Imagine knowing exactly what the next 5 years will look like.

You buy the best stocks, short the worst, and rebuild the best portfolio every five years over a 90-year period.

Your worst drawdown? 47%.

And if you only go long the winners, the drawdown gets even worse: 76%.

Even with perfect foresight and superhuman returns, you still get cut in half at some point.

Volatility is not an exception. It is a constant. The choice is whether to prepare for it or be ruled by it.

(Click here for a deep dive on the study.)

None of these lessons is difficult to understand in hindsight.

What is difficult is recognizing them while fear, noise, and uncertainty are still in control.

If there is any edge in this piece, it is simply this: remembering what matters before the market forces you to.

Take care,

Flo

Subscribe now

Share

The wealth it destroys never shows up on a statement, which makes it hard to see, hard to study, and easy to ignore.

So I simulated millions of portfolios to pin down where opportunity cost starts, what amplifies it, and how to reduce it in practice.

Here’s what the data shows.

Subscribe now

The Setup

To make the results interpretable, you need the setup.

I start with the same initial universe of stocks and simulate millions of portfolios across the full sample period (2001-2025).

I then vary only a few things: portfolio size, turnover, sell rules, and what gets bought after a sale.

Everything else stays fixed: same starting list, same return paths, same horizon. Only the decision rules change.

The point is to isolate the cost of cutting compounding trajectories.

Full methodology and limitations are at the end of the post.

Where Does Opportunity Cost Actually Begin

A buy opens exposure to a range of possible futures. A sell closes that exposure.

The moment you sell, an open trajectory is replaced by a fixed outcome. Once sold, that future no longer belongs to us. That is where opportunity cost begins.

In plain terms: Opportunity cost = a better outcome that was available to you − the outcome you realized.

So the main driver isn’t buying. It’s selling.

The catch is that selling’s damage doesn’t show up immediately. It shows up over time, so we need to track how opportunity cost evolves.

The chart below shows the evolution of opportunity cost across millions of portfolios over 25 years.

The y-axis is logarithmic. On a log scale, a linear trend reflects exponential growth.

So, opportunity cost doesn’t just add up. It compounds.

And the sudden step-ups make the mechanism visible: those jumps often come from selling positions that later turned into outsized winners.

Fortunately for us, not all sales are created equal.

Why Does It Become So Destructive?

This is where things get interesting.

I simulated three types of selling behavior:

• Selling psychological losers → positions whose current price is below their average purchase price

• Selling at random

• Selling psychological winners → positions whose current price is above their average purchase price

Take a few seconds to think about it.

Beyond the potential mind-blowing effect, active engagement increases the ROI of your reading time.

Here is the result:

This is the first major finding of the simulation:

• Selling psychological losers leads to lower opportunity cost than buy-and-hold.

• Buy-and-hold leads to lower opportunity cost than selling psychological winners

Even better, in terms of performance: selling psychological losers > selling random > selling psychological winners.

In practice, great compounders tend to be found among stocks that have already gone up, not among those that have gone down.

On average, selling winners gives you a slightly higher chance of cutting off a trajectory that still had exceptional upside ahead.

That asymmetry may look small in the short run. Over long periods, it becomes brutally destructive.

This is Peter Lynch’s “selling your winners and holding your losers is like cutting the flowers and watering the weeds,” but shown here in a more systematic way.

The value of the simulation is that it ignores the subjective “reasons” investors cite for selling and focuses only on the decision rule. Across millions of portfolios, that framework already covers a huge variety of paths and outcomes.

But the mean result is only part of the story. The distribution of selling costs matters just as much:

Visually, the differences look subtle. The consequences aren’t.

• This chart shows that most sales impose only a small opportunity cost. The curves rise steeply at low levels, which means a large share of sales do relatively little damage

• But a small minority of sales are catastrophically expensive. Those extreme mistakes happen far less often when selling psychological losers than when selling psychological winners (the green and yellow curves are shifted to the right).

That is the real mechanism.

The destructiveness of selling does not come from every sale being costly. It comes from the fact that a few sales are catastrophic, and selling winners increases both the frequency and the magnitude of those catastrophic errors.

A useful analogy is Russian roulette.

In investing, opportunity cost works like a game in which:

• you do not know how many chambers are in the cylinder,

• you do not know how many bullets are loaded,

• and each sale is a pull of the trigger.

Selling winners is like adding more bullets to the cylinder, and making them larger. The bad outcomes become both more frequent and more severe. Selling losers does the opposite.

But it can get even worse. In fact, it often does. And this time, you are the culprit.

Subscribe now

What Amplifies It?

We’ve identified the direct trigger: selling a great compounder. But two factors make that mistake more likely, and more expensive.

1) Portfolio size

A larger portfolio doesn’t just “diversify.” It increases the odds you’re holding a true compounder.

And opportunity cost doesn’t come from owning a compounder. It comes from selling one.

So I hold the number of sales constant and increase portfolio size.

The pattern is clean: as portfolio size grows, opportunity cost grows. Even without trading costs. Even by selling randomly.

Why? Because more names means more chances to own a rare winner, and then “accidentally” cut it.

That’s also why selling psychological winners gets more destructive as portfolios get larger: the “top performer” in a 40-stock portfolio is simply more likely to be a real compounder than the top performer in a 10-stock portfolio.

One detail matters: the small “step up” for sell-worst from 5 → 10 stocks isn’t magic. With 5 sales/year, a 5-stock portfolio is basically full churn, so even “sell-worst” quickly forces you to sell winners. At 10 stocks, you can sell 5 names while still leaving the top winners untouched, temporarily at least.

But the headline doesn’t change: bigger portfolios amplify the cost of selling.

2) Selling frequency (for a fixed portfolio size)

Which brings us to the second amplifier: turnover.

This is the most striking chart in the analysis.

Each curve says a lot, so let’s analyze them one by one.

Sell random positions

Start with the simplest case.

Each additional sale pushes performance further below buy-and-hold. In other words, each sale adds opportunity cost.

And that cost is not just financial. Every sale also consumes time, attention, and conviction, only to leave you worse off than if you had done nothing.

In the Russian roulette analogy, higher turnover means pulling the trigger more often.

At first, the damage rises quickly. Then it begins to flatten around 35% turnover (7 sales per year in a 20-stock portfolio).

The reason is simple: at high enough turnover, two opposing effects begin to offset each other. More selling creates more chances to cut off future compounders, but it also creates more chances to buy new ones. What emerges is a negative statistical equilibrium: still worse than buy-and-hold, but no longer deteriorating at the same rate.

The sell-best and sell-worst rules are, in a sense, mirror images of each other.

Sell the best position

This rule is brutally expensive from the start. Your top performer is one of the most likely places to find a stock with exceptional upside still ahead.

As turnover rises, the damage increases. But not forever. In a 20-stock portfolio, once you sell enough names, the “best remaining position” eventually becomes the worst loser.

Sell the worst position

The logic is reversed.

At low turnover, selling the worst position helps because you remove the weakest names first. As turnover rises, that benefit grows, for a while.

But here too, there is a limit. Once enough losers have been sold, the “worst remaining position” eventually becomes the best winner.

That is why the curves start far apart, then move closer together as turnover approaches 100%.

At that point, the sell rule matters less, because the same stocks are increasingly being sold under different labels. Replace the whole portfolio every year, and “sell the best,” “sell the worst,” and “sell random” all converge toward roughly the same outcome: persistent underperformance versus buy-and-hold.

Let’s note something important: performance typically improved when sales came from the bottom 75% of the portfolio. That fact, surprising as it is, will matter later.

The conclusion is simple: portfolio size and turnover both amplify opportunity cost.

Each additional sell is another chance to cut off a compounding trajectory.
Each additional stock is another chance to own one, and eventually sell it.

In the end, the biggest driver is what you do with your biggest winners.
And the more stocks you own, the more that decision matters.

How to Reduce It in Practice?

Let’s summarize the key findings:

• Opportunity cost is created at the moment of selling. Reallocation has limited power to offset it.

• In both performance and opportunity cost: sell psychological losers > sell random > sell psychological winners.

• The higher the turnover and the larger the portfolio, the higher opportunity cost tends to be.

• Selling winners is even more destructive in larger portfolios.

These results imply a few practical rules to minimize opportunity cost and maximize long-term performance.

I’m ranking them from least important to most important.

5) The Reallocation Trap

In a reallocation, the opportunity cost from the sale is often larger than the expected benefit from the new buy (when it isn’t a loss).

You can make a “smart” purchase and still make a terrible decision, if you funded it by selling the wrong stock.

Prioritize your reasons to sell, not your reasons to buy.

4) Reduce Portfolio Size

A well-built portfolio of 10 stocks can already capture most diversification benefits without sacrificing performance.

Beyond 15 names, you often drift into “diworsification”: diluted conviction and performance disguised as prudence.

More stocks doesn’t mean less risk. It means more ways to make the one mistake that matters.

3) Build a System That Limits Turnover

Most lifetime performance comes from a few extreme winners.

Those stocks are rare. And they require time.

Every additional sale is another chance to cut a compounding trajectory before it does its job.

Reduce the number of sell decisions. Period.

2) Sell Your Losers Ruthlessly

In the simulation, losers were sold each year and defined in the simplest way possible: the stock trading furthest below its average cost.

Crude rule. Powerful effect.

Why? Because you’re far less likely to find a future great compounder among your current losers, and more likely to find one in what replaces them.

And remember: performance improved almost every time a sale came from the bottom 75% of the portfolio. “Loser” is a wide category here. In practice, it basically means: anything that isn’t one of your true winners.

Sell your losers. Let your capital search for better futures.

1) Protect Your Biggest Winners

If you only remember one thing, make it this:

Selling your winners, those farthest above your cost basis, should be exceptional. Maybe even never, as a routine.

Losses are capped at -100%. Gains have no ceiling.

One true compounder can pay for a lifetime of losers, including stocks that looked like winners at some point and then weren’t.

The most telling result in the analysis was brutally simple: the best setup by far was effectively selling everything each year except the top 5 performers.

Not a magic rule, but a clear message.

You don’t get rich taking profits. You get rich by not touching the winners.

Take care,

Flo

Subscribe now

Leave a comment

Full Methodology and Biases

This analysis uses a deliberately simplified simulation framework designed to isolate one thing: the opportunity cost created by selling.

The simulation starts from a fixed initial cohort of stocks. The cohort includes every stock with a valid year-end price at the end of 2000, and all portfolios are formed from that same starting universe. No late entrants are added later. The sample then runs from 2001 through 2025 using annual data only, with each year represented by the last available price of that year.

Each simulated portfolio is equal-weighted at inception. Portfolio sizes vary across scenarios, and millions of portfolios are randomly drawn for each configuration. A portfolio is then evolved over the full sample period under a specified decision rule. The key variables changed across simulations are portfolio size, turnover, the sell rule, and the buy rule. Everything else is held constant: same starting universe, same return paths, same horizon.

The purpose of this structure is not to mimic real life perfectly. It is to make the effect of selling decisions measurable.

Portfolio evolution

At the end of each year, portfolio positions are repriced using the year-end price of each stock. If a stock later disappears from the dataset, its value is frozen from that point onward rather than forcing a delisting assumption or a terminal loss. This avoids injecting a separate modeling choice into the results. When replacement is impossible, cash is used as a fallback.

All rebalancing takes place only within the initial cohort. In other words, the simulation does not reopen the universe each year and does not allow the purchase of stocks that were not already in the starting set. This is intentional: it keeps the analysis focused on the cost of cutting existing trajectories rather than on the benefits of discovering new names.

Sell rules

Three sell rules are used:

• Random selling: positions are sold at random.

• Psychological winner selling: positions with the largest unrealized gains relative to average purchase price are sold first.

• Psychological loser selling: positions with the largest unrealized losses relative to average purchase price are sold first.

When multiple positions are sold in the same year, the rules are applied in order. Under winner selling, the biggest winner is sold first, then the second biggest, and so on. Under loser selling, the biggest loser is sold first, then the second biggest, and so on.

The core intuition is behavioral rather than fundamental. The simulation is not trying to determine whether a sale was justified by valuation, business quality, or thesis deterioration. It is asking a narrower question: what happens when the decision process is systematically tilted toward selling the strongest winners, the weakest losers, or selling without bias?

Buy rules

Two replacement rules are used:

• Random replacement: replacement positions are chosen randomly from the eligible names not already held.

• Oracle replacement: replacement positions are chosen ex post from the same eligible universe based on their future terminal performance.

The oracle rule is not meant to be realistic. It is a stress test. It asks whether even the best possible replacement, with hindsight, can fully offset the cost created by the sale. In most cases, it cannot.

Benchmarks and definitions

The primary benchmark is buy-and-hold. For each simulated portfolio, the corresponding buy-and-hold benchmark is the portfolio that keeps the same initial positions over the same period, with no turnover. This makes the comparison internally coherent: same starting cohort, same names, same horizon, different decision process.

Opportunity cost is measured in two related ways.

First, at the portfolio level, it is the gap between the realized outcome of the active selling process and the outcome of the benchmark or comparison path.

Second, at the event level, it can be framed as the difference between what a sold position would have become if kept and what the replacement position actually became after transaction costs.

These are different views of the same mechanism: selling closes one path and forces capital onto another.

Why annual data

The simulation uses annual observations rather than daily or monthly data. This reduces noise, keeps the experiment interpretable, and focuses attention on the long-run compounding consequences of selling rather than on short-term fluctuations. That also means the analysis is about structural opportunity cost, not execution timing.

Main biases and limitations

This framework is intentionally narrow, and that creates several biases.

1. Fixed initial cohort bias.
The investable universe is frozen at the starting date. Real investors can discover new companies over time. This setup excludes that possibility. It therefore isolates the cost of abandoning existing winners, but it likely understates the value of skill in finding new names later.

2. Annual frequency bias.
All decisions happen once per year using year-end prices. Real selling decisions are continuous, not annual. This means the simulation captures broad directional effects, not the exact path that would occur under real trading behavior.

3. No thesis-level information.
The model does not know whether a company deteriorated fundamentally, became obviously overvalued, cut guidance, or faced permanent impairment. It only sees prices and rules. So the results should not be read as “never sell.” They should be read as evidence that selling is far more expensive than it appears by default.

4. No taxes, slippage, or market impact beyond simple trading costs.
Transaction costs are modeled mechanically, but taxes and execution frictions are not. In real life, those frictions usually make selling look even worse.

5. Equal-weight simplification.
All portfolios start equal-weighted. That makes cross-scenario comparisons cleaner, but it is not how all real portfolios are run. Capital allocation skill is not modeled here.

6. Survival and disappearance handling.
When a stock stops having prices in the dataset, its value is frozen rather than treated as a realized delisting return. That is a conservative modeling choice for consistency, but it is still a choice. Different assumptions would slightly change the tails.

7. Psychological winner/loser definitions are price-based, not fundamental.
A “winner” is defined as a position with an unrealized gain relative to average purchase price, and a “loser” as a position with an unrealized loss. The sell rules rank positions by that price-based measure, not by fundamentals or intrinsic value. A stock can be above cost and still cheap, or below cost and still bad.

8. Opportunity cost is partly path-dependent.
The exact magnitude of opportunity cost depends on which future extreme winners were owned, when they were sold, and what replaced them. That means the results should be interpreted statistically, not as deterministic laws for every individual case.

What this framework is good for

This simulation is useful for answering general questions about the consequences of selling, not for judging specific real-world decisions.

It is not meant to determine whether any single sale was justified. Its purpose is to show that, as a class of actions, selling is far more dangerous than most investors assume.

That is the main point.

Petabytes of data. Armies of elite analysts. Computing power that could rival an intelligence agency. And the result? Almost no real edge.

The explanation is simple: modern finance looks at the branches and misses the roots.

Subscribe now

Branches vs. Roots

That line gets to the heart of investing. So let’s do a little tree work.

Branches are the visible, legible stuff. The kind of information that drops neatly into a spreadsheet: KPIs, guidance, margins, polished narratives, and so on.

Branches are seductive because they’re:

• easy to understand,

• easy to communicate,

• and they feel objective.

Roots are the hidden upstream causes that make branches exist: management culture, customer obsession, employee quality, incentives, learning speed, etc.

And that’s the problem: roots don’t present themselves cleanly. They’re difficult to observe, difficult to compare, and almost impossible to “prove” with tidy numbers. You can’t measure them quickly, and you can’t measure them neutrally.

You can only grasp them by continuously exposing yourself to everything that touches the company, near or far.

But our attention naturally goes toward what’s available, clear, and quantifiable (availability bias). So we default to branches and neglect roots.

Yet that’s where what matters most is decided. To show it, let’s go back to the tree for a moment.

Tree anchoring is the way a tree’s roots anchor into the soil so it doesn’t get uprooted when the storm hits.

Translated to business, roots are the company’s intangible characteristics that let it survive during crises.

One example is worth a thousand explanations.

The Costco Tree

In 2000, Costco looked like a classic American success story: public for 14 years, already a household name, widely respected, and up something like 25% per year since the IPO.

And yet for much of the 2000s, the stock went nowhere. In fact, it was down.

If you stayed at the branch level, the story looked bad:

• Wall Street expected strong growth: more stores, more revenue, higher EPS.

• When management cut guidance, the market translated it into one word: saturation.

• Then came what looked like the “fatal” decision: hold prices steady and let margins compress. For Costco, it was long-term discipline. For the market, it was weakness.

But if you focused on the roots, reality was totally different:

• Leadership quality hadn’t degraded.

• The culture stayed consistent through stress.

• The focus on the customer and low prices, at the expense of short-term performance, was reaffirmed.

That’s where the opportunity was: a gap between what the market could easily see and what actually mattered.

Even if you bought at the 2000 peak and simply held through it, you still ended up beating the market over the long run, around a 12% CAGR (vs. 6%).

Charlie Munger is a clean illustration. As a director and shareholder, he didn’t sell during the 2000s. He never sold, period.

When asked why he didn’t sell Costco during the 2000s, this is the kind of answer he gave: “an unwavering commitment to customer trust,” “employee loyalty,” “product excellence,” “ethical culture,” and so forth.

No branches. Only roots.

That asymmetry is the whole game:

• The market prices branches.

• Time rewards roots.

Without high-quality roots, long-term compounding is nearly impossible.

How to Look at The Roots

Answering this question exhaustively would be solving long-term investing itself. A pure chimera, you’d agree.

But you can get surprisingly far with a few clues. Start with the first, often the most powerful: what not to do.

Modern finance defaults to what can be measured: things that look objective, travel well in spreadsheets, and usually cater to the short term.

Invert that. Stop piling up branches:

• Stacking measures (KPIs, valuation metrics, etc.) is only the illusion of precision.

• The market is already saturated with “objective” measures. They rarely create durable outperformance on their own.

• The biggest opportunities tend to appear when branches snap, when the short-term story looks ugly and consensus turns.

Then comes the harder part: what to do instead.

In theory, it’s simple: develop judgment about the quality of the root system. In practice, it’s not easy.

Here are the kinds of questions that actually get you closer:

• Under pressure, what does the company sacrifice first: the customer, the employee, or the quarter?

• When trade-offs get painful, do decisions still align with a promise the company has repeated, and honored, over time?

• What does the management team’s record reveal?

  • Behavior: honesty, intellectual humility, an owner mindset.

  • Capital allocation: patterns that repeat (investment, pricing, growth discipline) and what they produced over years.

  • Promises: what was delivered, what wasn’t, and most importantly why.

None of this can be reduced to a formula. It’s interpretation, pattern recognition, and time spent close to the business.

And that’s exactly why it’s neglected, and exactly why it can become an edge for anyone willing to do the work.

The Only Reliable Method

Most investors are hunting for certainty. So they load up on branches, green leaves everywhere: metrics, models, clean narratives, reassuring spreadsheets.

But green leaves never last. And branches always break.

The branch-level obsession, the branch fallacy, isn’t going away. It’s structural: institutions, incentives, career risk, reporting cycles, and a deep fixation on what can be measured. Quantity wins by default.

If you want to be a long-term compounder, you have to resist that default. You have to train your attention on the roots.

There’s no perfect method. Roots aren’t a checklist. Like beauty, they’re ultimately a matter of judgment.

So the best you can do is reduce uncertainty the only way that works: accumulate root-level perceptions over time. And the most reliable way to do that is to become a serial learner.

It’s no accident that the best investors, and the best operators, share this trait. They learn obsessively, continuously, and across domains.

May this article and this newsletter help you do the same.

Take care,

Flo

Subscribe now

Scheduling Update:

I’m changing the publishing schedule for my series, 50 Mental Models to Think Like a High-Level Generalist.

The original plan was one post per week for ten weeks. In practice, that pace is too constraining to maintain the level of quality I’m aiming for.

So I’m giving myself full flexibility on timing. I’ll publish the series either as one complete post or split into two posts, released as soon as they’re ready. I won’t give a precise date, but it won’t be finished for at least a few months.

To be clear: this isn’t a stop. I’m working on it continuously and will keep doing so. I just enjoy writing this series too much to ruin it with arbitrary deadlines that hurt the quality.

Thanks for your understanding.

Take care.

I hope your brain says “yes.” I’m sure your actions say “not really.”

Buffett, Munger, Lynch, etc. No one is immune.

The bias doesn’t disappear with experience. It just wears a different disguise.

A short walk is all it takes to see it.

Subscribe now

No Country for Old Men

You’re out for a quiet walk.

It’s summer. The hot wind stimulates your sudoriparous glands, when suddenly you spot a $100 bill on the ground.

Being naturally anxious, you hesitate. What if it belongs to the wrong person? A dealer. Or a psychopath. These days, who knows.

Decision made. You leave it right there. Either way, you didn’t lose anything and you didn’t gain anything. So everything’s fine.

Except it isn’t.

In reality, your anxiety just made you $100 poorer. An anchoring bias made you compare your new empirical wealth to your wealth from a now-dead past.

But the moment that $100 entered the equation, your baseline shifted. Finding it raised your potential wealth by $100. Walking away cost you that $100 of potential.

Anchoring, loss aversion, and regret aversion aside1:

• Not earning $1 = Being $1 poorer

• Not losing $1 = Being $1 richer

And that’s the point: you’re about to lose millions (maybe billions) of dollars you’ll never earn. The more you trade (buy/sell), the more you’ll lose.

For better and for worse.

I Love You… Me Neither

The moment you buy a stock, you’re buying a range of futures.

• Your potential wealth now includes whatever that stock can still become

• The moment you sell, you collapse that range into a single outcome.

That’s why your realized path will almost always lag behind your full set of potential paths, which brings us to this hard-to-swallow equation:

Empirical Wealth ≤ Theoretical Wealth.

To turn ≤ into =, you’d need to sell only when the stock’s remaining upside is worse than your best alternative, every time. Which is literally impossible, you’ll agree.

The whole investing game is making decisions that:

• maximize your potential wealth (your analysis skills),

• while minimizing the gap between theoretical and empirical wealth (your psychological skills).

But you’re playing the I Love You… Me Neither game against a serious opponent: the asymmetry of stock returns (max loss = -100%, while max gain → +infinity):

• I love you: own a multi-bagger that compounds for decades.

• Me neither: selling it too early for a lower-performing stock blows up your opportunity cost.

If you’ve been investing long enough, you probably have a few scars that never fully healed

Investing is Choosing Our Regrets

Peter Lynch has a perfect story for this.

This isn’t just Peter Lynch’s greatest mistake. It’s the biggest mistake of many of the greatest investors:

• Buffett bought Berkshire Hathaway out of ego, calling it his “dumbest stock,” and estimated the opportunity cost at $200B by his own math (see the interview).

• Peter Thiel invested very early in Facebook, taking a 10% stake for $100k. He cashed out at the IPO for $2B, which would be $100B today.

• SoftBank bought 5% of Nvidia in 2017, sold after a 20-30% drawdown in 2019 for $3B. Today, it would be $200B, more than SoftBank’s market cap.

Those mistakes cost them far more than their “empirical” losses. And it’s always the same culprit: the asymmetry of gains.

And for us too, mere mortals, it’s often the #1 source of value destruction.

So I decided to follow Sun Tzu’s advice: “If you know neither your enemy nor yourself, you will be defeated in every battle.”

And even if I don’t have Peter Lynch’s phrasing, I do have tools he didn’t have back then: a lot of data, and a lot of compute.

Perfect for getting to know a mathematical enemy.

Do the Maths

To really see opportunity cost in all its forms, I simulated hundreds of millions of portfolios to answer four questions:

• How does opportunity cost show up, and how does it grow?

• How does it change depending on the portfolio’s structure?

• How do buying and selling decisions impact it?

• What, concretely, reduces it?

These simulations gave me some very surprising results, and they’ll have a direct impact on my investment process, without question.

Here’s the full analysis.

Take care,

Flo

Share

You can find the first 5 here.

Let’s go.

Law: Stare Decisis Et Non Quieta Movere

The Big Idea

“Stand by the thing decided and do not disturb the calm.”

In plain English: past decisions become the default rule for similar cases, that’s how case law gets built.

More precisely:

• A court’s prior decision is a precedent.

• Stare decisis is the principle that courts should follow relevant precedents.

When a court rules on a novel case, it creates an anchor point that judges, lawyers, and citizens will align around.

Precedent is therefore a tool:

• a arbitrary tool for deciding when reality is new and uncertain,

• that creates predictability by imposing a default for future similar situations.

Arbitrary, yes, but effective for managing uncertainty at low cost.

The Mental Model

Once again, the trick here isn’t in generalizing, it’s in transposing.

Our first reaction to a new situation tends to become our “precedent”: the default response next time.

We can thank biology and psychology for that:

• Similar causes (context), similar effects (reactions).

• Our brains are wired to lock in a solution fast (anchoring, confirmation, availability).

So the right question, every time you face something for the first time, isn’t “what should I do now?” but: “Am I okay with this becoming my standard?”

If, on your first investment:

• You require yourself to read the latest annual report + transcripts before buying, you create a standard.

• You buy because your colleague won’t stop talking about it, you also create a standard, but a bad one.

Applying this mental model with zeal is an excellent way to deserve it until you get it. Especially the “until” part.

The 1-Line Rule

Choose your reaction to a new situation as if you’ll have that same reaction every time, because you probably will.

Biomimetics: Meta Mental Model

The Big Idea

Congratulations, it’s a meta mental model! → A mental model that lets you discover (or create, your choice) other mental models.

Based on its definition, biomimetics is a core discipline in this series:“Biomimetics is the emulation of the models, systems, and elements of nature for the purpose of solving complex human problems.”

The idea of “mental models” fits naturally inside biomimetics, with one difference: mental models (and therefore this series) aren’t limited to nature in the biological/ecological sense, but in the physical sense (our universe, and beyond).

If we stick to a biological definition, nature is (roughly) just the result of an initial event, environmental variation, and natural selection, that is:

• the selection of traits that appeared by chance through mutations and that maximize survival and reproduction, and

• the disappearance of the rest.

Natural selection is therefore a machine for creating (or selecting, your choice) increasingly optimized processes.

Put differently: among all the possibilities the universe offers, natural selection will select the most effective ones (for a given environment).

That makes it an excellent source of mental models to shamelessly clone.

The Mental Model

As a meta mental model, the process is only meant to discover other mental models (some people call that meta-wisdom):

• Study biological/ecological processes shaped by natural selection, and understand why they’re so effective;

• Describe and recognize the situations in which that process is effective;

• Apply the process in similar situations.

Jeff Bezos has admitted to “stealing” ideas from biology to build Amazon. His shareholder letters contain clear examples.

For a complete, step-by-step case study → Mental Model #6: Clonal Selection.

The 1-line Rule

Steal proven solutions from nature: identify the function, extract the principle, then port it to your problem.

Oceanography: Reservoir Thinking

The Big Idea

The ocean is best understood as a system of stocks and flows:

• a stock = a quantity contained in a compartment → CO₂ in the deep ocean, salt at the surface, etc.

• a flow = a transfer between compartments → CO₂ from the deep ocean to the atmosphere.

The key feature here is that, in oceanography, we often deal with gigantic stocks relative to the flows that change them.

The direct consequence is inertia: even if flows change today, the stock moves slowly.

Very slowly. For example:

• A hot summer can raise the temperature of the top 50 meters in a few days/weeks.

• But then heat takes years to penetrate below the surface (thermocline), and decades, or even centuries, to reach (and meaningfully alter) the deep ocean.

That response time becomes the oceanographer’s nightmare: it drastically complicates the analysis of causes and their consequences.

The Mental Model

The goal is to look for similar structures, estimate their response time, and adapt our strategy to the system’s timeframe.

1. Identify the Stock: what variables “carry” the system? (cash, backlog, trust, love, etc.).

2. Identify the Flows: what flows change the stocks?

3. Estimate the Response Time: a workable approximation is response time ≈ stock / net flow (more often a mental estimate than a real calculation).

4. Don’t misread noise: if response time is long, short-term variations are mostly noise.

WARNING: a flow is bidirectional along the same axis, and its speed can differ wildly in each direction. Think reputation: it can take decades to (re)build (high stock & low flow), but get destroyed in seconds (high stock & very high flow).

The 1-Line Rule

Higher Stocks + Lower Flows = Higher Response Time → Higher Relevant Timeframe

Cybersecurity: The Best Defense Is Offense

The Big Idea

Cybersecurity is war. And like any war, it’s won through paths.

A system doesn’t fall “because it’s imperfect.” It falls because there exists a path with a positive attack ROI:

• an entry (e.g., reused password),

• followed by a viable path to the target (a chain of access all the way to the keys to the kingdom),

• an impact (money, data, reputation, etc.).

Cybersecurity is the art of finding and blocking those paths, and of limiting the blast radius when it still gets through.

That’s why engineers practice pentesting (penetration testing) with zeal: simulating an attack yourself to find those paths and block them.

As often with systems (yes, them again), the practice generalizes extremely well (a country, a portfolio, an economy, an industrial process, etc.).

The Mental Model

Pentesting happens in 5 phases:

1. Reconnaissance → Gather all available information.
   Fundamentals, drivers, correlations, liquidity, dependencies, history, maturities, etc.

2. Scanning → Map the attack surface and identify breaking points.
   Concentration (sector, geography, currencies, etc.), implicit leverage, counterparties, correlation spikes, etc.

3. Exploitation → Test scenarios step by step until they cause real damage.
   Historical FX moves (-20% on EURUSD) / needing day-to-day liquidity + deep drawdowns, etc.

4. Post-Exploitation → Assess potential damage and its probabilities.
   The € has already dropped 15% per year versus the $ for multiple years in a row. As a European investor invested only in USD-denominated assets, my annual performance can be wiped out by FX.

5. Reporting & Remediation → Turn discoveries into path-blockers.
   I diversify my currency exposure / I keep 6 months of expenses + taxes in cash / I stagger my options maturities by at least 6 months, etc.

The 1-Line Rule

Without an adversary model, you build a random defense that leaves the cheapest entry point wide open.

Political Science: State Capacity

Power is useless if you don’t have the capacity to wield it.

That capacity is called state capacity.

More formally, it’s: the ability of governments to effectively implement their policies and achieve their goals.

For a government, that capacity comes from the capacity of its organizations (police, tax administration, teachers, etc.) and its infrastructure (police stations, online tax portals, schools, etc.).

But you can generalize this principle to any structure made of living beings with power asymmetries: a family, a classroom, a city, etc.

The Mental Model

The “state capacity” reflex is to explicitly separate the power to decide from the capacity to execute, to enforce, and to prevent.

When a new leader arrives (CEO, head of state, president of a regulatory body, etc.), the question isn’t “what does he want?” but:

• is the system he runs capable of executing what they want, and

• where is the potential bottleneck?

When Gary Gensler, former “Blockchain & Money” instructor at MIT, took over the SEC in 2021, the crypto community welcomed him like a near messiah.

But what followed illustrated the mental model perfectly: the SEC became crypto’s worst enemy during the Biden administration, and Gary Gensler was shaped by the SEC system more than he shaped it.

The appointment of Kevin Warsh as the next Fed Chair could potentially be a concrete case study for applying this mental model.

The 1-Line Rule

In every system, the real ruler isn’t the top, it’s the bottleneck. Weak capacity yields promises, strong capacity yields consequences.

Share

Only 35 more to go. The pace is very intense, but this series is transformative.

Every week I feel a massive step-change in progress. No doubt I’ll be far better equipped at the “end” of this series.

I hope it’s the same for some of you. Either way, I’m trying.

Take care,

Flo

Scheduling Update:

I’m changing the publishing schedule for my series, 50 Mental Models to Think Like a High-Level Generalist.

The original plan was one post per week for ten weeks. In practice, that pace is too constraining to maintain the level of quality I’m aiming for.

So I’m giving myself full flexibility on timing. I’ll publish the series either as one complete post or split into two posts, released as soon as they’re ready. I won’t give a precise date, but it won’t be finished for at least a few months.

To be clear: this isn’t a stop. I’m working on it continuously and will keep doing so. I just enjoy writing this series too much to ruin it with arbitrary deadlines that hurt the quality.

Thanks for your understanding.

Source

Law: Stare Decisis Et Non Quieta Movere

• I have no doubt a legal specialist might be slightly offended by how much I simplified stare decisis and precedent. But it was a necessary tradeoff for readability. For a more complete and rigorous definition:

  • Cornell Law School (LII, Wex) — “Stare decisis”.

  • Cornell Law School (LII, Wex) — “Precedent”.

• To go deeper on the behavioral side of “precedent”:

  • Tversky & Kahneman — “Availability: A Heuristic for Judging Frequency and Probability” (Cognitive Psychology, 1973).

  • Tversky & Kahneman — “Judgment under Uncertainty: Heuristics and Biases” (Science, 1974).

Biomimetics: Meta Mental Model

• On Bezos, biology, and business: Jeff Bezos (Amazon), 2006, 2016, and 2020 Letter to Shareholders.

• For a concrete biomimetics case study: Julian F. V. Vincent et al., “Biomimetics: its practice and theory,” Journal of the Royal Society Interface (2006).

Oceanography: Reservoir Thinking

• A relatively gentle introduction: NOAA Climate.gov, Climate Change: Ocean Heat Content.

• A more complex and more specific approach (and nothing beats a concrete case to highlight real-world complexity): Messias et al., 2022, The redistribution of anthropogenic excess heat is a key driver of warming patterns in the North Atlantic.

Cybersecurity: The Best Defense Is Offense

• My source for the pentest methodology. Note: I merged/removed some steps to make the framework generalizable: PTES, The Penetration Testing Execution Standard.

• For those interested in cybersecurity but not experts (like me): an excellent guide that goes deep while staying approachable. It felt like reading a modern-day Art of War: NIST, SP 800-115: Technical Guide to Information Security Testing and Assessment (2008).

Political Science: State Capacity

• For this mental model, a single (excellent) source was enough. It breaks down state capacity from every angle: Cingolani (2013), “The State of State Capacity: a review of concepts, evidence and measures.”

Pick anyone you want for an arm-wrestling match.

• +$10,000 if you win.

• –$10,000 if you lose.

Of course, you wouldn’t pick Dwayne Johnson.

You’d look for the smallest, frailest opponent you could find, win quickly, take the $10,000, and compensate them for their time.

But you wouldn’t pick Dwayne Johnson.

Even if the word “arm-wrestling” made you think of him first.
Even if everyone else picked him.
You’d pick the opponent that makes winning easiest.

Agreed? Great.
Now let’s change the game.

Imagine an active investor managing 8 figures or less.

You’d expect him to do the same: pick the arena where winning1 is “easiest”, not where the strongest players in the world are already fighting with better tools.

So he wouldn’t:

• focus on the part of the market where prices are less inefficient,

• compete directly against the biggest, smartest players,

• in the most crowded names.

He’d go where:

• there are fewer competitors,

• his edge actually matters,

• and inefficiencies are still meaningful.

Even if everyone is talking about the big names.
Even if the first tickers that come to mind are mega caps.
Even if the crowd keeps piling into the same companies.

Yet most investors go straight for Dwayne Johnson. Here’s why.

Read This Before You Do Something You Could Regret

To avoid talking past each other and/or getting into heated debates, here’s exactly what I mean in this post:

• “Investor” = an active investor managing < $50M

• “Active” = active on two dimensions:

  • stock selection

  • time horizon (from a few months to a few decades)

• “Small caps” = market cap < $1B, with sufficient liquidity

• Competence assumption: if an investor can do stock picking in large caps, they can do it in small caps too.2

Let’s begin.

Good Answers, to the Wrong Question

I’ve had countless conversations with active investors who are anti–small caps.

When I ask, “Why don’t you focus on small caps?”, I always get the same sempiternal three arguments:

1. Large caps are much less risky (fraud/bankruptcy) and less volatile.

2. There are far more bad companies in small(er) caps.

3. Information on large caps is higher quality and more transparent.

All of that is true. And that’s exactly why an active investor should generally avoid large caps.

When the goal is to generate alpha3, there’s only one question to ask:

“Where do my efforts have the best chance of paying off, given the risks I take and the time I spend?”

The answers are simple:

1. Where other investors don’t like to go, because they fear perceived risk and volatility.

2. Where there are plenty of bad businesses, and where stock-picking skill is more likely to create an edge.

3. Where information is harder to find, and where individual research is more likely to create an edge.

4. Where you compete against fewer people who are smarter than you and better resourced than you.

In other words: smaller caps.

Their answers were good, but they answered the wrong question. In fact, they were the right answers to the inverse question.

That’s the theory. It still doesn’t explain why so many active investors end up focusing on large caps in practice.

Let’s talk about junk, availability, and blindness.

Size Matters, If You Control Your Junk

It’s possible to get rich by controlling your junk. Let me explain.

The “size premium” (the long-run outperformance of smaller companies) has sparked endless debate over the past 50 years.

But most of that debate rests on an assumption that, for the purpose of this post, is a mistake: it treats small caps as one homogeneous bucket.

In reality, small caps contain more “junk” on average and a much wider dispersion of quality. Treating them as a single bloc means refusing a very real selection opportunity, which, for an active investor, is pure alpha destruction.

Fortunately, researchers re-examined the size premium by applying a simple quality filter to small caps; they call this “control your junk.”4

With these simple filters, the size premium not only becomes meaningfully larger, but the effect also shows up across:

• virtually all countries,

• virtually all industries,

• in a way that’s robust and stable over time.

Better yet, it’s not one filter, it’s multiple filters.

There are five, and all of them improve performance. Here they are5:

• Profitability (+5.16%/year) → higher margins, strong cash flows, and cash earnings that match accounting profits

• Safety (+4.28%/year) → low leverage, stable profits, low distress risk

• Payout (+5.28%/year) → limited dilution and meaningful cash returned to shareholders

• Growth (+2.43%/year) → multi-year improvement in profitability (ROE/ROA/margins/cash flows)

• All four combined (+6.04%/year)

If several different definitions of “quality” all point the same way, it says something fundamental: in small caps, filtering is the condition that makes the universe investable.

So what: does the lack of interest in small caps simply come from investors not realizing how heterogeneous they are, and how much filtering matters, therefore leaving all these opportunities on the table? Is that really the root cause?

Of course not.

This is one of the most competitive games on the planet (maybe the most). A simple lack of easily available information can’t be the explanation for that much apparent irrationality.

In situations like this, the thick description mental model is useful: look for the invisible constraints that explain the behavior, rather than defaulting to ‘irrationality.’

Since Kahneman and Tversky, and the death of homo economicus, the prime suspect is almost always the same: investor psychology.

The Pernicious Inheritance

Most investors don’t really choose their investing universe, they inherit it.

They show up with the same 30 names already printed in their heads:

• the ones they hear everywhere before they’ve even opened an annual report,

• the logos and jingles they’ve seen since childhood (e.g., home bias),

• the stocks the “knowledgeable” coworker keeps repeating (often the one who got them into it in the first place),

• add a pinch of social-media echo chambers (two “must-listen” podcasts, three viral threads…),

and voilà: a perfectly functional availability bias.

Availability bias is the tendency to make decisions based on the information that requires the least effort to access.

Where do most people go for finance information? Mainstream financial media: social platforms, TV, YouTube, and the rest.

And how do mainstream finance media become (and stay) mainstream? By talking about the mainstream companies that are most likely to interest the largest audience: large caps.

The snake is really eating its own tail. But we can go further.

Once you add the impact of other powerful biases, like:

• confirmation bias (e.g., the 4th bullish thesis on a company mostly just reinforces what you absorbed from the first three),

• anchoring (your baseline for what a “good” business looks like, what kind of volatility is “normal,” and so on),

you get a classic Lollapalooza effect: investors are almost inevitably pulled toward the same large caps, as if by gravity.

But that’s only the perception side of the story. If the problem stopped there, small-cap avoidance among small active investors wouldn’t be so persistent.

To go further, we need to look at how people react to these perceptions.

I Perceive, Therefore I Become

Let me borrow a concept from social psychology and tailor it to investing: cultural blindness.

If availability bias is about the first environment you’re exposed to, cultural blindness is what you do afterward, when you more or less willingly lock yourself into that environment and become blind (or impermeable) to everything else.

In investing, it’s something we all do:

• when we subscribe to this a newsletter, a newspaper, a YouTube channel, etc.

• when we repeat a takeaway about a company that we first picked up through availability bias

• and more subtly, when an algorithm “refines” what it shows us based on our clicks, or even the extra second we spend staring at a thumbnail

And those are just our internal incentives to build an environment that feels comfortable.

But our environment can be persuasive too:

• Attention maximization: you talk about what everyone talks about (otherwise you’re talking to yourself.)

• Social defensibility: picking ideas you can defend with standard references so it sounds “reasonable” before the results show up.

• Shame asymmetry: a “mainstream” mistake gets diluted whereas an “obscure” mistake sticks.

• And so on.

Add thousands of other small, subtle, mostly unconscious forces, and the combined pull becomes almost inexorable.6

So inexorable that it ends up shaping our “preferences” everywhere (music, sports, political views, etc.), to the point where we often confuse repeated exposure with genuine taste. Or maybe genuine taste is just repeated exposure.

Either way, a pattern this recurring probably deserves a few minutes of thought.

Just to be clear: my goal isn’t to convince you to invest actively in small caps.

Everyone chooses their game, and that’s perfectly fine.

Think of this post as a message in a bottle: one more perspective, take it or leave it.

Before I cork it and toss it back into the ocean, let me slip one last note inside:

If I was running $1 million today, or $10 million for that matter, I’d be fully invested. Anyone who says that size does not hurt investment performance is selling.
The highest rates of return I’ve ever achieved were in the 1950s. I killed the Dow. You ought to see the numbers. But I was investing peanuts then.
It’s a huge structural advantage not to have a lot of money. I think I could make you 50% a year on $1 million. No, I know I could. I guarantee that.
Warren Buffett

Take care,

Flo

Subscribe now

“I’m a big fan of knowing the big ideas in pretty much all the disciplines … and then using those routinely in your judgments. That’s just my system.”

I’ve extracted 50 “big ideas” from 50 disciplines.

5 new ones every week.
You can find the first 5 here.
This week: #6 through #10.

Immunology: Clonal Selection

The Big Idea

Imagine we receive a radio message from an extraterrestrial species: “We’re declaring war. We’re going to wipe your planet out.” How do you prepare?

You don’t know the enemy, their weapons, or their strategy. You’re facing an “infinite” adversary in the sense that the space of possibilities exceeds what you can even imagine.

That’s exactly what your immune system has to do, constantly.

To do that, it pre-positions as much diversity as possible and lets reality do the selecting. This process is called clonal selection:

• The immune system maintains a gigantic library of immune cells (especially B cells and T cells), each carrying slightly different receptors.

• When a pathogen X shows up, most cells won’t detect it, but a few will partially match it.

• That partial match triggers a cascade that makes those cells proliferate and specialize.

• Some of the newly produced cells are kept and strengthen immune memory.

• Next time a similar pathogen appears, the response is faster and stronger.

In one line: clonal selection = variation → selection → amplification.

Here’s an excellent visual representation of the process:

Ontologically, it’s simply an immune learning process: the immune system learns how to defend itself against a new, unknown “enemy,” and stores what worked.

Interestingly, different types of learning tend to share the same shape, whether it’s human, immune, or machine learning.

The Mental Model

The process is so effective that it almost begs to be translated.

When you’re facing a problem that’s complex, uncertain, or evolving:

1. Generate variety: produce multiple candidates (hypotheses, options, prototypes, routines, etc.) with low unit cost and fast feedback.

2. Define a selection criterion: a metric that tells you a solution works “well enough.”

3. Let reality select: deploy candidates and observe which ones survive by the criterion.

4. Amplify: double down on what works, while reintroducing a bit of variety.

5. Store it: treat it as learning, and optimize the learning process itself (immune memory).

6. Repeat until the solution is good enough, or stop if the cost (time, attention, money) becomes too high.

Simple, but brutally effective.

I joined social media three months ago and knew absolutely nothing about it. Faced with that complexity, I applied exactly the method above:

• I opened accounts on a bunch of platforms.

• I tried many things in parallel (and still am), then focused on what worked best.

• This series is one way for me to inject variability during the amplification phase.

That said, don’t judge the method by my “performance.” Most problems come from the strategist more than the strategy.1

This isn’t gratuitous self-bashing, but if nature selected this learning process for the immune system over millions of years, it’s probably because it’s fundamentally extremely effective.

Machine Learning: Overfitting

The Big Idea

No wise teacher would use the exact test questions as training exercises. Students would have massive incentives to memorize the answers, and the teacher’s hard work would be wasted.

So teachers vary practice problems and test problems. Surprisingly (or not), that’s also what we do in machine learning.

What we really want from a neural network is discriminative power across the widest possible range of cases.

In general, when you train a neural network, you split the data into three sets:

• a training set,

• a validation set (used during training to tune and decide when to stop, but not essential for our purpose here),

• and a test set.

The goal is to prevent the network from becoming “too perfect” on the training set, and then failing on slightly different data (the test set), which would make it useless. That failure mode is overfitting.

There are many ways to reduce overfitting (e.g., dropout, deliberately forcing “forgetting”), but the most fundamental lever is simple: the quality, diversity, and representativeness of your data.

The closer your sample gets to the true population of situations the model will face as input, the less overfitting you should expect (all else equal).

But in machine learning, the overall process is relatively standardized, which makes optimization much easier than in biological learning.

In biological learning:

• We don’t have a clean, reliable “test set” metric the way we do in machine learning.

• Data arrives discontinuously and in a non-standardized way.

• Training is often discontinuous (e.g., learning how to react appropriately when someone wishes you happy birthday: one “training session” per year).

So the consequences of overfitting are more subtle and more pervasive, but we can still borrow a lot from machine learning to reduce its impact.

The Mental Model

Overfitting is a cognitive bias. And like most cognitive biases, its impact tends to shrink simply by becoming aware of it.

Overfitting’s best friend is precision. Your detector should go on high alert whenever something feels too clean, too specific, too perfectly fitted.

A classic example in finance: indicators or models that “predicted” the last 10 bear markets or recessions. They’re almost always built after the fact and tuned to the past, impressive in hindsight, fragile in the future.

Statistically, there’s an infinite supply of indicators. So you can always find some that fit the historical record perfectly. But statistically, (almost) all of those indicators won’t predict the future.

There’s no absolute rule, but these heuristics make the mental model usable in practice:

• Precision → Alert. The more precise a claim/model/theory is, the more suspicious you should become.

• Stress-test robustness. Automatically ask: does this survive changes in conditions? The more precise it is, the more it must.

• Separate discovery from proof. What helped you find the idea should not be what you use to “validate” it (keep training set and test set separate). If you’re using the same data for both, the odds of overfitting jump.

• Sometimes: don’t learn. Human unlearning is expensive. When the downside of internalizing a bad model is high, caution can beat cleverness: better to skip learning than to overfit and then pay the price to unlearn.

Subscribe now

Anthropology: Thick Description

“There is no reason to believe that beasts have thought… and that it is more probable that they are devoid of it.”

Anyone who has a pet would call the person who said that cold, detached, and irrational.

But your certainty probably wobbles if I tell you that this very same person also pioneered analytic geometry, while being one of the most famous philosophers of all time: René Descartes.

So how can a mind that is clearly brilliant say something that looks this absurd?

As usual, you have to zoom out: think in fields, not points.

The trap is judging an idea (or a behavior) outside its context. In anthropology, we often contrast two levels:

• Thin description: you describe the act “as-is,” isolated, stripped of its environment.

• Thick description: you describe the act together with the system that gives it meaning (norms, beliefs, tools, era, constraints…).

If you ignore the environment and focus only on the act itself, “irrationality” is often the only conclusion our personal lenses can offer.2

But once you integrate the constraints shaping that environment (knowledge constraints, tools, time, culture, incentives, etc.), many behaviors that looked irrational start to make sense.

Even though the concept comes from intercultural analysis, it applies extremely well to “everyday life”:

• Why did this person commit such an obvious fraud?

• Why buy this stock?

• Why sign this treaty?

If you stop explaining the world with “irrationality,” and instead look for the context and constraints that structured the decision, you very often get closer to the truth.

The Mental Model

Whenever a behavior (decision, statement, action, etc.) feels irrational, pause judgment long enough to answer six questions:

1. What is the real goal of this behavior?

2. Are there credible alternatives to reach that goal?

3. If yes, which ones, and what do they cost?

4. What does the actor risk (reputation, sanction, loss of face, etc.)?

5. What does the actor actually know, and what might they not know?

6. Am I sure I know all the important constraints that could be shaping this behavior?

Applied to the biggest financial crises in history, constraints usually explain far more than “pure irrationality”: liquidity constraints, leverage constraints, balance-sheet constraints, and so on.

And the biggest bull runs are often tied to the gradual, continuous relaxation of those constraints. Irrationality tends to matter most in the final act.

Information Theory I: Overconsumption of Complexity

The Big Idea

Promise made, promise kept.

Nothing is more fundamental than information. In some sense, any phenomenon can be modeled as a flow of bits:

• the erosion of your neighbor’s home’s walls,

• your neighbor’s cells in the middle of mitosis,

• the evaporation of the black hole at the center of our neighboring Andromeda galaxy.3

Nothing escapes that lens (yes, even the neighbors).

This is perfect, because the whole point of this series is to extract important ideas and generalize them to maximize their usefulness (and that’s why at least one more information-theory concept will show up later in the series).

But not all information is created equal.

In information theory, information is tightly linked to uncertainty: the core quantity (entropy) measures uncertainty, and “information gained” can be seen as uncertainty reduced.

If it doesn’t reduce the relevant uncertainty, whether it’s true, rare, or interesting is trivial.

So in practice, a piece of information has value only if it reduces the uncertainty that actually matters (i.e., the uncertainty that blocks a decision).

Ignoring this leads to the error of overconsuming complexity: hoarding non-decisive information while feeling like you’re doing rigorous work.

The fix is to focus on the marginal value of new information.

The Mental Model

We’re almost constantly foraging for information (visual, auditory, procedural, etc.).

So the process needs to be fast, cheap, discriminating, and repeatable:

1. Decision variable: State the decision type (act, wait, choose, quit, etc.).

2. Switching information: What single piece of info could drastically flip my decision?
   (e.g., “I turn down this job offer if I find a negative, non-anonymous review from a former employee.”)

3. Compression: If no decisive info comes to mind quickly, compress the problem into 2-3 variables.

4. Stopping rule: Set an exit rule in advance (e.g., “I stop after 30 minutes, no matter what”).

5. Marginal value test: For each new piece of info: does it reduce uncertainty on those variables enough to change the decision?

6. Exit: Stop when either:

   1. a decision is made + you can name the information that changed your mind, or

   2. your stopping rule triggers.

Across a lifetime, the marginal value of this reflex is enormous. It even makes you wonder why information theory isn’t taught in school in a practical (math-free way), through heuristics like these.

Motivation Psychology: The Expectancy-Value Trick

The Big Idea

Motivation psychology is a deep well of knowledge you can apply to everyday life.

I could have talked about feedback types and what they’re good for, about juggling extrinsic vs. intrinsic motivation, about emotion as a signal of motivation rather than the enemy, etc. But I chose a “simple equation” of motivation:

Motivation ≈ Goal value × Perceived probability of success4

Be kind to motivation psychologists. The field has roots in 19th-century physiology, where there was a strong urge to describe human body with equations, and that habit stuck around well into the 20th century.

Even if this equation doesn’t make mathematical sense, what matters are the concepts and the relationship between them:

• Your motivation for a task depends on the value of the goal and your perceived odds of succeeding.

• These are tendencies, not absolute rules. But often very pronounced:

  • High value + low probability → anxiety + avoidance

  • High value + high probability → motivated

  • Low value + high probability → apathy, procrastination, etc.

  • Low value + low probability → indifference

But that’s not all. There’s a crucial nuance that turns this into something actionable.

The Mental Model

I said value and perceived probability. So we’re talking about perception (as always, with living things).

The upside is that we can influence our perceptions.

Concretely, if you want to keep motivation high for longer, you have two levers (the examples may sound caricatured, but this is about perception; use whatever fits you best):

1. Increase perceived value

   1. Add meaning to the task → “My kids’ air quality will be better if I sort my waste properly.”

   2. Tie it to identity5 → “I’m someone who takes care of myself, so I’m doing this workout.”

   3. Strategic rewards/punishments → “I won’t look like someone who breaks their word if I finish this post by Sunday.” (Notice: it’s a negative reward, I’m avoiding a punishment.)

• Increase perceived probability of success:

  • Reduce task difficulty → go from a 1-hour run to 45 minutes

  • Reduce perceived difficulty by clarifying the next step → “analyze management’s track record” rather than “do my research”

  • Design the task to provide fast evidence of progress → break a complex task into a sequence of simpler sub-tasks with feedback.

Actively managing, and even “editing”, our perceptions is (probably) the biggest cheat codes in psychology.

This mental model is just one example among many.

10/50. 40 to go. I’m almost starting to feel nostalgic.

As much as I genuinely enjoy working on this project, the main goal is still for these tools to reach as many people as they can actually help.

So if you think this post could be useful to someone else, share it.

This project is 100% free, and it will stay that way.

Take care,

Masters of Compounding

PART 3 is already out:

• Stare Decisis

• Meta Mental Model

• Reservoir Thinking

• Pentesting

• State Capacity

I did my best to make it as valuable for you to read as it was for me to write.

Sources

1. Immunology: Clonal Selection

• As is often the case, it’s hard to point to a single truly “foundational” study behind a theory. More often, knowledge accumulates until a few people (or sometimes a single person) come along and formalize it into a coherent whole. That’s exactly what this book does. Maybe it’s because the field was much less developed at the time, but the book is very readable if you still have a solid memory of your biology classes: Frank Macfarlane Burnet (1959). The Clonal Selection Theory of Acquired Immunity. Vanderbilt University Press.

• For a more recent (and more comprehensive) approach to the topic: Hodgkin, P. D. (2008). “The golden anniversary of Burnet’s clonal selection theory.” Immunology and Cell Biology, 86(1), 15.

2. Machine Learning: Overfitting

• For a simple introduction to overfitting vs. underfitting and the evergreen bias–variance tradeoff: Google. Machine Learning Crash Course — “Datasets, Generalization, and Overfitting” (Overfitting / Generalization module). Google for Developers, last updated December 3, 2025.

• This paper is “old” by machine-learning standards, but it nails the core point: generalization is the real problem. And it’s exactly this kind of (sometimes uncomfortable) thinking that, in my view, enabled the field’s major breakthroughs, up to modern architectures like transformers: Zhang, C., Bengio, S., Hardt, M., Recht, B., & Vinyals, O. (2016). Understanding deep learning requires rethinking generalization.

3. Anthropology: The Thick Description

• The foundational text for the concept. The book is fairly old and you can feel it at times, but it’s still a worthwhile read: Clifford Geertz (1973). “Thick Description: Toward an Interpretive Theory of Culture”, in The Interpretation of Cultures.

• For a more recent critique of the concept and its limits: Susen, S. (2024). The Interpretation of Cultures: Geertz Is Still in Town. Sociologica.

4. Information Theory I: Overconsumption of Complexity

• The foundational text of information theory, and one of the greatest papers of all time. It’s fairly hard to read without math skills (even if it’s not very demanding), but the first 10 pages are really worth it, IMO. Claude Shannon (1948). A Mathematical Theory of Communication.

• For a detailed framework explaining why humans tend to over-consume information (and how they implicitly trade off value vs. cost when searching, reading, and “digging”): Peter Pirolli & Stuart Card (1999). Information Foraging. Psychological Review.

5. Motivation Psychology: The Expectancy-Value Trick

• The model described in this post is Expectancy-Value Theory. It’s neither the most comprehensive nor the dominant model today. But most modern frameworks can be seen as more developed versions of this core idea. For a deep dive into the foundations: Wigfield, A., & Eccles, J. S. (2000). Expectancy–Value Theory of Achievement Motivation. Contemporary Educational Psychology, 25(1), 68–81.

• For a broader view of the current state of motivation psychology: Urhahne, D., & Wijnia, L. (2023). Theories of Motivation in Education: an Integrative Framework. Educational Psychology Review, 35, Article 45.

For many investors, the answer is yes. The longer the period, the higher the odds that the answer becomes “yes.”

That’s because the distribution of opportunity costs is far more asymmetric than the distribution of capital costs.1

Translation: thousands of hours analyzing companies and sectors, writing theses, building valuation models. All that, only to destroy value in the end.

Incredibly frustrating, isn’t it? But what if it’s actually liberating?

This phenomenon stems from a very specific property: the quality of the output doesn’t directly depend on the quality and quantity of the inputs.

In our investing world: performance doesn’t depend on hours worked. When you think about it, that’s an extremely rare phenomenon in nature. But that’s exactly what makes investing so unique, and so liberating.

From that property comes an incentive with incredible power: simplify, simplify, and simplify.

If we take it a step further, it’s an incentive to build an investing process that is both:

• as simple as possible without sacrificing relevance, and

• explicit and repeatable, so you can refine it over time through feedback (without adding complexity).

That’s exactly how François Rochon thought about it, and acted. Result:

• 15.1% annualized for 30 years

• Worst calendar-year drawdown: -15.2% (2023)

Better yet: he shared his entire strategy in 2017 in a one-hour Talks at Google (and it has surprisingly few views).

The advantage of a simple system is that it takes five minutes to explain.

1. The Foundations: Invert? Always invert

“Index ETFs.” Some of you thought it so loudly while reading the intro that I heard you from the future.

Yes, it’s an excellent way to simplify things and earn “acceptable” returns. But no. François Rochon doesn’t want acceptable returns.

In fact, his core premise is precisely that he does not want the performance of his benchmark, the S&P 500. He’s aiming for 15% per year.

To get there, he studies the best by asking what he would have to do to achieve market returns… and then he does the opposite.

That leads him to 3 simple ideas, but ones he takes very seriously:

• To own a few selected companies

• To develop the right behaviors (rationality, humility, and patience)

• To be able to think for himself

2. The Art of Science & The Science of Art

As a trained engineer with a passion for art, François Rochon approaches investing through both lenses.

For him, the “science” side is the “Stock Selection Process”. And the advantage of an explicit process is that it’s easy to lay out.

Here is François Rochon’s Stock Selection Process (verbatim):

• Financial Strength

  • ROE > 15%

  • EPS growth > 10%

  • Debt/Profit Ratio < 4x

• Business Model

  • Market Leader

  • Competitive advantages

  • Low cyclicality

• Management Team

  • High level of ownership

  • Constructive acquisitions

  • Good capital allocation

• Market Valuation

  • 5-year valuation model

  • Try to purchase at half the estimated value in 5 years (≈15% annual return)

If you’re thinking: “Wait, how do you judge a competitive advantage? That’s not as measurable as ROE or low cyclicality,” congratulations: you’re exactly right. And you’ve just handed me a perfect transition.

That’s where art fills in the gaps left by science (in this context only!).

According to François Rochon, competitive advantage isn’t something you can fully objectify, it’s something you recognize when you see it.

His method is to find what he calls “masterpieces”: rare, unique businesses (pieces of music) where the presence of a competitive advantage (the quality of the work) is so obvious that almost no one can reasonably deny it.

Examples he gives: IKEA, GEICO, Apple, McDonald’s, Gillette, The Washington Post, Disney, National Cash Register Co., etc.

This system stays simple by drastically shrinking Rochon’s investment universe. In 2017, that universe was limited to 125 companies.

We’ve covered the “business” side of investing. Now let’s move to the main component: psychology.

3. His Behavioral Edges

If you spend any time on social media, you’ve definitely run into people giving all kinds of advice about “behavioral edge”: mindset, discipline, “controlling your emotions,” etc.

The problem is that this kind of advice is often more useful to the people who write it than to the people who read it. (And I include myself in that.)

However, none of the advice-givers out there has a track record as exceptional as Rochon’s over 30 years (and yes, I include myself again.)

So I’ll point out that even if Rochon’s advice sounds… basic, it’s probably deeper than it looks.

To avoid misrepresenting his thinking, I’ll quote François Rochon directly on each of his behavioral edges.

Humility

• We do not believe that we can predict macroeconomic events.

• One key element is to define our circle of competence2 and to know where are the limits of its circumference.

• We recognize our mistakes and try to always strive for improvement.

A small detail: François Rochon includes a “podium” of mistakes in every letter to his partners.

Beyond the humility and honesty that takes, it’s also a goldmine for any investor who wants to learn from other people’s mistakes. That will be a future post.

Rationality

• We try not to be affected when others make more money than us in stocks.

• We try to be impervious to stock market quotations in the short term

• We accept that we don’t know the future and focus on what we can have some control (our own process).

Even though he didn’t include it in the “rationality” section, this quote captures the essence of Rochon’s rationality for me:

“We believed that if we owned great companies run by great managers, things would work out in the end. They did. So we’re going to keep doing that.”

Thanks to François, there’s no need to wait 30 years to validate the idea. It worked, and it still works.

It seems reasonable to bet that it will keep working.

Patience

• Patience is not the ability to wait, but the ability to keep a good attitude while waiting.

• However, patience is not the same thing as being stubborn.

• To avoid that mistake as much as possible, the good attitude is to focus on what is happening to the company, not the stock.

From these three tendencies emerges a broader one, something many great investors preach, and Sir John Templeton captured perfectly:

“To obtain better results than the others, you must do something different from the others.”

4. I Am, Therefore I Think for Myself

This trait is inevitable: it’s baked into the very origin of François Rochon’s investing system.

When he asks himself what to do to avoid average returns, he imposes an implicit condition: act differently.

And to act differently, you have to think differently. Paraphrasing him a bit, that becomes: be impervious to fads and popular beliefs. Don’t be afraid to underperform your peers for a while.

François Rochon has held positions that underperformed for years. Positions that were hated by the public, before watching them explode in a matter of months.

Few investors have a mindset as independent and as consistent as François Rochon. Let me show you:

“Good businesspeople are like artists: they build with passion, they create something innovative and unique, and they aim to make it last, something that inspires others. My job, then, is to estimate what that remarkable business should be worth and try to buy it at roughly half of what I believe it can trade for in five years if things go as planned, about 15% a year.”

Sounds like a slightly trivial summary of the post, doesn’t it? This is an excerpt from something François Rochon wrote in 2005.

The video I used as a reference for this post is from 2017. He’s still applying the same strategy today.

To me, it’s the clearest sign that his performance came from his ability to think for himself consistently, for decades, without being stubborn in the face of reality.

And voilà. Full circle.

We’ve just seen the three pillars of his strategy:

• To own a few selected companies

• To develop the right behaviors (rationality, humility, and patience)

• To be able to think for himself

One last question to close.

What’s stopping you from following his system, or even following his positions directly, to get closer to his performance?

That’s the kind of question every investor should answer honestly.

Personally, I want to take advantage of the fewer constraints that come with a smaller capital base to aim for more than 15% a year. (At least until the next real bear market. I’m joking, but…)

That’s my answer. What’s yours?

I’ll leave you alone with your honesty.

(And a quick reminder: Part 2 of my series 50 Ideas to Think Like a High-Level Generalist comes out on Sunday.)

Take care.,

Flo

Share

In opportunity cost. In repeated mistakes. In blind spots that show up at the worst possible time.

Charlie Munger put it simply:

“The man who needs a new thinking tool but hasn’t yet acquired it is already paying for it.”

His fix is simple, demanding, and, most importantly, repeatable:

• Learn the big ideas from the big disciplines.

• Then reuse them everywhere: to decide, analyze, evaluate.

“I’m a big fan of knowing the big ideas in pretty much all the disciplines … and then using those routinely in your judgments. That’s just my system.”

So I decided to do the work: build my own Swiss Army Knife for thinking, in public.

To do that, I’m going to map the big ideas across the big disciplines, run them through a filter, then turn them into ready-to-use mental models.

Here’s how it’s going to work:

• 1 post a week for 10 weeks (on top of the usual weekly post).

• Each post: 5 disciplines, 5 big ideas, 5 mental models.

• And to go deeper: graded sources, from “I’m discovering” to “I want to suffer.”

I’ve been working on this project for weeks. It’s ready. And it starts now.

Seismology: Self-Organized Criticality

The Big Idea

In seismology, the Gutenberg–Richter law says two main things:

• The higher the magnitude of earthquakes, the lower their frequency.

• But this decline in frequency doesn’t fall off “normally” (in the sense of a normal distribution).

Translation: truly extreme earthquakes are rare, but clearly not as rare as a normal distribution would allow.

Distributions with that property show up in systems with equally unusual characteristics:

• They accumulate constraints slowly, but surely.

• They release them suddenly, in jolts.

In seismology, those constraints are created by the continuous motion of lithospheric plates, which builds tectonic stress in the Earth’s crust. Slowly, but surely.

That’s self-organized criticality1.

More rigorously: it’s when a system has the critical point as an attractor, due to its intrinsic dynamics.

It continuously drives itself toward a near-critical state, one that’s just waiting to flip into a more-or-less extreme event, triggered by any of a wide range of sparks.

If that definition feels familiar, that’s normal. There’s a good chance you’ve seen a “documentary” about it:

Financial markets are textbook cases of self-organized criticality.

For years, the system stacks constraints: leverage, mispriced risk, “crowded” positions, hidden correlations, and last but not least, illusory liquidity.

Then comes the spark (sometimes tiny) that triggers a cascade of causes and consequences that makes history:

Drop → forced selling & margin calls → liquidity dries up → further drops → again and again.

This kind of logic is widely distributed in the universe. You find it in:

• Avalanches: the more snow accumulates, the higher the odds that a few extra snowflakes are enough to trigger a break.

• Revolutions / social contagion: a local event can serve as the trigger for a dynamic that was already under extreme tension. A “recent” example is the Arab Spring in 2010, sparked after Mohamed Bouazizi, a street vendor, set himself on fire in public in Tunisia.

• Wildfires: “fuel” builds up for weeks (dried vegetation, low humidity, wind, etc.), then comes the spark (literally).

The Mental Model

Systems where extremes are structurally more likely correspond to fat-tailed distributions.

Being able to spot this kind of system completely changes risk management: the focus stops being “predict the trigger” (it can be too many things, at too many times), and shifts to quantifying fragility.

For that, here’s a 6-point checklist to diagnose this kind of system:

• Constraint: is pressure building vs dissipating?

• Feedback loops: does the system dampen shocks vs amplify them?

• Dependence on stability: does it “work” only when everything is going well?

• Coupling: does a break stay local vs spread as a cascade?

• Margins / redundancy: is there slack when something breaks, or at least a fallback route?

• Failure mode: is the breakdown gradual vs threshold-based?

This kind of analysis (deeper, I’ll admit) could have helped regimes avoid collapse, municipalities invest more in wildfire prevention, or funds (like Long-Term Capital Management2) avoid going under.

In practice: when fragility rises, the priority isn’t being right about the scenario. It’s reducing exposure to the cascade.

Because the only certainty is that it’s coming. Inevitably.

Artificial Life: Complexity and Pattern Can Emerge from Simple and Local Rules

The Big Idea

I’m going to do some proselytizing: this field is genuinely FASCINATING. If you’re even slightly curious about life, emergence, and where patterns come from, do yourself a favor and go down this rabbit hole. It’s almost a personal career regret… anyway.

Being a researcher in artificial life means building and analyzing computer simulations from a handful of rules, then watching how those structures evolve over time.

In its most rudimentary form, Conway’s Game of Life, you start with a grid of cells (on/off), and you add:

• local rules

  • a live cell survives if it has 2 or 3 live neighbors.

  • a dead cell becomes live if it has exactly 3 live neighbors.

• an initial state: the configuration at t = 0 (the starting distribution of cells)

Then you let it run.

And that’s where the magic happens (or rather reality):

From two or three ridiculously simple rules, you start seeing extraordinary configurations emerge3:

• stable patterns,

• structures that move,

• “organisms” that seem to reproduce, feed, or interact.

Make no mistake: nobody coded those “behaviors.” They emerge.

And this is the basic case. You can take it much further:

• move from a grid to continuous space,

• add variables (color, energy, resources),

• introduce randomness, selection, constraints, etc.,

and watch dynamics emerge that are infinitely richer, so rich that biological life starts to feel almost… banal.

Once you’ve gone deep enough, you won’t look at “simplicity”, or even “life”, the same way again.

The Mental Model

Complexity and patterns can emerge from simple, local rules.

Two central ideas fall out of that, mirror images of each other:

• A starting situation with very simple rules can quickly produce complexity that’s hard to wrap your head around.

• A situation that looks insanely complex and random can come from rules that are simple and deterministic.

In a field where the expected level of analysis often requires you to dissect situations that seem overwhelmingly complex (like finance and probably every discipline), it can be a lifesaver to remember that a few simple, decisive rules are usually hiding underneath that complexity.

Subscribe now

Climatology: Think in Fields, Not Points

The Big Idea

The concept of a field is central in meteorology.

Concretely, a field is a map where every point on the map is associated with a value, or several values.

Weather is “simply” the simultaneous evolution of multiple fields (pressure, temperature, geopotential height, etc.).

Thinking in “points” is like examining a single tree in the middle of a forest: you lose the overall structure, the gradients (the differences), and above all the dynamics that create the “whole.”

Thinking in “fields” is looking at the map instead of a few pixels: the overall shape, and how it deforms across space and time.

The Mental Model

The advantage of this big idea is that points are everywhere: prices, ratings, metrics, etc.

The two fundamental questions are:

• What field is hiding behind these points?

• What moves it and what makes it vary?

Reasoning in terms of fields is stepping up one level and focusing on the big trends (the fields, often invisible to our perception) that determine local variables (the points), the very things that saturate our attention and perceptions.

It also means accepting a bit less local precision as you move up a level of abstraction, in exchange for better global relevance.

Better to be vaguely right than precisely wrong.

Probabilities: Bayesian Thinking

The Big Idea

The universe is too complex for our brains, in a computational sense. We don’t have the time, the horsepower, or the data.

So we do the only sensible thing: we replace all-or-nothing with degrees of confidence.

In an environment where information is uncertain and arrives continuously (so: our universe), Bayesian probability is the reference tool:

• You start with one or more priors: what you already believe (or what you assume, for lack of anything better).

• You estimate the probability of a claim based on those priors.

• You update that probability with every new piece of information.

In a Bayesian frame, a “fact” is a belief attached to a slider between 0 and 1. Ideally, each of our beliefs about the world should be a set of sliders between 0 and 1, updated regularly.

If that feels too abstract and unusable in practice, understand this: it’s because of Bayesian probabilities that we can understand each other (among other things).

A baby doesn’t learn language because it has a grammar manual encoded in its genes. It learns like a Bayesian, through successive updates:

• it watches mouth movements,

• matches sounds to those patterns,

• tries to reproduce them,

• then adjusts its hypotheses based on feedback and the regularities it perceives.

In other words, its beliefs refine little by little, over months and years, as evidence accumulates, until language becomes coherent.

That doesn’t disappear in adulthood. Our unconscious is naturally Bayesian.4 In everyday language, that shows up as instinct and intuition.

Our conscious mind, on the other hand, is much less Bayesian, but we can train it!

The Mental Model

When you touch something as fundamental as information (that’s for a future post), the upside is that it transfers easily.

In practice, it’s just applying Bayesian principles:

• Make your prior / hypothesis explicit, and your confidence level.

• Make explicit what claim your belief is about.

• Attach a number or a range (ideally) between 0 and 1 (watch out for false precision → 0.743).

• Update as soon as new information arrives.

• Improve your estimates from the feedback reality gives you (sometimes very brutally).

The more we apply these principles, the more rational we become, the less we’re affected by bias, and the more successful we are, in general. It’s probably the best way to raise our Rationality Quotient.

Systems Theory I: Recognize the Story, Regardless of the Set

The Big Idea

In short, systems are organized complexity. More formally, a system is a whole made up of:

• elements

• that interact with each other

• with defined boundaries

• inputs and outputs (information, energy, capital, etc.)

• states (temperature, price, social status, etc.)

Here are two examples of systems:

No matter your domain, that definition should ring a bell: an organization, a technology, a product, a customer base, a portfolio, a family… these are all systems.

The real mental leap is to stop looking only at “things,” and start looking at the shape they share: interaction patterns, feedback loops, constraints, delays, flows.

That’s the famous “think outside the box.”

The earthquakes and financial crashes at the beginning of this post illustrate it perfectly. On the surface, they have nothing to do with each other. But if you look at them as systems, the common points jump out:

• stress accumulation,

• sudden release in bursts,

• and therefore a much higher probability of extreme events than “normal” intuition suggests.

Same story. Different set.5

The Mental Model

By recognizing these patterns, you can anticipate the dynamics before experience has to teach you the same lesson again, just in a different setting.

Here the framework is:

• Level up: describe the situation as a system by making its components explicit (focus on the story).

• Transpose: reuse what you learned from system A to illuminate system B (even if they look unrelated).

• Act based on that transposition.

The hardest part is obviously recognizing the patterns. But taking time to analyze a situation after the fact, again and again, is one of the best ways to build your personal library of patterns.

Some people call that wisdom.

And there we go: the first 5 mental models in our Swiss Army Knife. I already can’t wait to share the next 5.

As much as I genuinely enjoy working on this project, the main goal is still for these tools to reach as many people as they can actually help.

So if you think this post could be useful to someone else, share it.

This project is 100% free, and it will stay that way.

As for me, I’m going back to prepping the next one.

Take care.

Masters of Compounding

PART 2 is already out:

• Clonal Selection

• Overfitting

• Thick Description

• Overconsumption of Complexity

• The Expectancy-Value Trick.

I did my best to make it as valuable for you to read as it was for me to write.

Share

Sources

1. Seismology: Self-Organized Criticality

• An introductory book that’s surprisingly complete despite having no math. There’s a certain poetry to it (and to seismology in general): Susan Elizabeth Hough — Earthshaking Science: What We Know (and Don’t Know) About Earthquakes

• The foundational paper on self-organized criticality: Bak, P., Tang, C., & Wiesenfeld, K. (1987). Self-organized criticality: An explanation of 1/f noise. Physical Review Letters.

• The first direct application of self-organized criticality to earthquakes: Bak, P., et al. (1989). Earthquakes as a self-organized critical phenomenon. Journal of Geophysical Research.

2. Artificial Life: Complexity and Pattern Can Emerge from Simple, Local Rules

• Please, read this: Artificial Life: A Report from the Frontier Where Computers Meet Biology — Steven Levy.

• ISAL (International Society for Artificial Life) provides a complete online bibliography covering the full range of topics studied in artificial life. Pick what interests you.

• Probably my favorite paper on the topic, and one of my favorite papers in any field. A gem on the deep links between these computer simulations and biological life: Bartlett, S., & Wong, M. L. (2025). Lyfe: learning to learn better. Interface Focus, 15(6).

3. Climatology: Think in Fields, Not Points

• A great introductory book (a bit long, but very digestible): C. Donald Ahrens — Meteorology Today: An Introduction to Weather, Climate, and the Environment.

• For a deep dive into field-based reasoning in meteorology (you’ll have to hang on): Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz‐Sabater, J., ... & Thépaut, J. N. (2020). The ERA5 global reanalysis. Quarterly Journal of the Royal Meteorological Society, 146(730), 1999–2049.

4. Probabilities: Bayesian Thinking

• Probably the best source on Bayesianism without math: “Bayesian Epistemology” (Stanford Encyclopedia of Philosophy).

• How the brain is built in a Bayesian way: Knill & Pouget (2004), Trends in Neurosciences.

• For those curious about how a Bayesian framework can explain how infants acquire their native language: Romberg, A. R., & Saffran, J. R. (2010). Statistical learning and language acquisition. Wiley Interdisciplinary Reviews: Cognitive Science, 1(6), 906–914.

5. Systems Theory I: Recognize the Story, Regardless of the Set

• The best introductory book on the topic, in my opinion. A bestseller I’d recommend to anyone, read it, then reread it: Donella H. Meadows — Thinking in Systems: A Primer

• A massive tome on applying systems thinking to economics and business. Very long, very interesting, and absolutely worth it, but at times you really have to hang on: John D. Sterman — Business Dynamics: Systems Thinking and Modeling for a Complex World

Share

The only way to avoid the “volatility bill” is to be able to afford it.

Let’s use some remarkable data to show why.

The Man Who Had Only a Hammer

Picture the scene. You graduate with a master’s in economics. You’ve been handed a powerful tool: math. Naturally, you try to apply it everywhere.

Problem: “To the man with a hammer, every problem looks like a nail.”1

And when what’s in front of you doesn’t look like a nail? More often than not, you treat it like one anyway.

In finance, standard deviation (and its sister, variance) is the basic building block behind a whole toolbox: optimization, portfolio models, risk measures, allocation, and so on. It’s clean, computable, and it snaps neatly into elegant frameworks.

The central challenge of applied mathemathics, though, isn’t producing equations. It’s producing equations that stay close enough to reality.

When the model and reality don’t line up, you have two options:

1. adapt the tool to reality, or

2. adapt reality to the tool.

The second option is the easiest, the most comfortable, the most wrong… and the most common. It gives you “clean” results: easy to publish, present, and conclude from.

In our case, the guilty assumption is that risk = volatility.

In other words: we call “risk” the dispersion of returns around their average, measured by standard deviation.

“According to the academicians […], risk equals volatility. […] I take great issue with this definition of risk.”
Source: Oaktree Capital Management, memo “Risk” (2006)

And the nice thing about this kind of assumption is that it has a weak spot: find just one clean counterexample, and it collapses.

So that’s exactly what we’re going to do.

(Everything in this post stays in a long-term investing context.)2

Q.E.D.

Ontologically, risk is simple: an event with some probability X of happening, and, if it happens, some damage Y.

If I can show that:

• the “risk event” occurs with 100% probability, and

• when it occurs, it doesn’t cause damage,

then it becomes pretty hard to keep claiming that risk = volatility.

So, let’s go back to the future past.

We’re in 1927.

You’ve invented a time machine, but it still has two big constraints (you’ll have to build the v2):

• it can only go five years into the future, and

• it takes five years to recharge.

You’re smart. You buy the Grays Sports Almanac Wall Street Journal pages that list prices for the 500 largest companies.

You come back to the present with something absurdly valuable: the next five years of returns for each of those 500 companies, in black and white.

What’s the strategy? Go long the winners? No. You’re greedier than that: you go long the best performers and short the worst.

You do that every five years, all the way through 2016.

Now try to guess your CAGR over 90 years.

Seriously, take a few seconds. I think anchoring bias is about to do some work for me.

Here’s what you get:

• Long the winners: 29.37% CAGR

• Long the winners + short the losers: 46.23% CAGR

At this point you might be thinking: “What?! Only 46.23% CAGR even when I literally know the future? What a joke. Not even worth building this damn machine.”

You might be able to break the laws of physics by time traveling (especially the “going back” part), but you can’t break the laws of economics. (I’m kidding, kind of. Economics is downstream of physics anyway, so it’s the same thing.)

Because $1 compounding at 46.23% for 90 years turns into $713,300,000,000,000. $713T.

Put in just $100 back then and you’d end up richer than today’s entire global economy, which, obviously, makes no sense.3

Interesting. But what does this have to do with risk?

Simple: even with a literal crystal ball, even with returns that are basically divine, even going long the best and short the worst… you still would have experienced 50% drawdowns.

That part blows my mind: you know the future, you take the most profitable possible positions on both sides (long and short), and you still eat 50% drawdowns (and 75% if you’re long-only).

So let’s summarize:

• You know you’re going to win. It’s inevitable. There is no “risk” in the sense of being wrong.

• Volatility still shows up, repeatedly, and violently. And yet your performance is extraordinary.

So we’ve found a case where volatility:

• occurs with 100% probability, and

• when it occurs, it doesn’t cause damage because you are 100% sure you’ll win.

Volatility is present even in a world where there is literally no risk of being wrong.

So in long-term investing, risk is not volatility.

Q.E.D.

Subscribe now

So What is Risk ?

The most useful approach is to ask people who’ve actually practiced long-term investing for decades, people who’ve experienced risk firsthand:

• “Risk to us is the risk of permanent loss of capital.” — Charlie Munger

• “The risk of an investment is described by both the probability and the potential amount of loss.” — Seth Klarman

• “Risk comes from not knowing what you’re doing.” — Warren Buffett

• “We define risk as the permanent loss of value per share.” — Nick Sleep

• “The biggest investment risk is not the volatility of prices, but whether you will suffer a permanent loss of capital.” — Li Lu

I could keep going.

What they’re really talking about is permanent loss, probability, uncertainty, and understanding. If you compress all of that into one sentence:

Risk is anything that can make you lose money irreversibly, or force you to sell before you’re right.

Let’s unpack that definition.

Permanent Loss: The Real Risk

The most fundamental risk is the one you can’t come back from. No recovery. No do-over. It can come from:

• A broken thesis → questionable management, disruption, the moat disappearing, bad capital allocation, etc.

• A weak balance sheet → too much debt, dilution risk, covenants, and so on.

• A bad entry price → paying too high a multiple, wrong assumptions on growth, no margin of safety.

In all of these cases, volatility is a cause of risk, not risk itself. It’s the market warning you before the final bill shows up.

Being Forced to Sell

This is probably the factor most standard models forget to account for.

That risk can come from:

• Leverage on your capital → volatility triggers a margin call and you sell at the worst possible time.

• Lack of liquidity → you sell to fund life, taxes, unexpected expenses, etc.

• An asset that’s too illiquid → sometimes you can sell, but only at a brutal price. Other times there’s no bid at all, so you can’t sell even if you want to.

Constraints are the market’s most dangerous silent killer. Without constraints, a drawdown is just discomfort. With constraints, it can be a death sentence.

Minimizing constraints as much as possible is a rule every great investor has followed on their journey.

Not Knowing What You’re Doing

As always, Buffett’s words carry more weight than mine:

“Risk comes from not knowing what you’re doing.”

The subtext is simple: if you don’t understand what you own, you can’t tell the difference between:

• a “normal” drawdown, and

• a drawdown that reflects real deterioration, i.e., actual risk of permanent loss.

That uncertainty feeds a whole chain of biases that almost inevitably lead to value-destructive decisions (panic selling, averaging down blindly, etc.).

Volatility Is a Tax on Those Who Can’t Afford It

Volatility isn’t risk. But volatility is a:

• horizon test

• sizing test

• liquidity test

• understanding test

In other words, volatility is a tax you pay… but only if you don’t have the mental and financial infrastructure to pay it. You pay the tax precisely when you can’t afford to pay it.

As long as you can hold, financially and behaviorally, volatility becomes what it really is: an intrinsic feature of equity markets. A “bad” stretch to live through, and one that will eventually pass.

Risk, the probability of a permanent loss of capital (because of constraints or a broken thesis), is more pernicious. It can disguise itself as volatility. But only people who’ve done the work can recognize the disguise. Most of the time, anyway.

Risk Isn’t Bad. It’s the Game.

In an uncertain world, trying to “eliminate risk” doesn’t make sense. The only serious question is: what is this risk paying me?

Risk is always there. Upside isn’t. You can take very real risk… for zero upside, or a ridiculous payoff.

The goal isn’t to “reduce risk to zero.” The goal is to take bets where, on average, and over time, the possible gains dominate the possible losses.

In other words: take risks that are well-paid, with a structure that lets you survive long enough for the statistical edge to do its job.

In investing, the winners are the ones who don’t get taken out of the game.

Writing here is one way I increase my own odds of staying in the game.

I hope reading it helps increase your odds of staying in the game too.

Flo

Share

“I came to law school having learned the method of learning, and nothing has served me better in my long life than continuous learning.” — Charlie Munger

That is a massive claim. Not because Charlie loved learning. Plenty of people do.

But because he’s saying something far more specific: in his view, the highest-ROI skill of his entire life was learning how to learn.

That should get our attention.

Learning is one of the few processes we can run almost every day, for the rest of our lives, at very low cost, with absurd upside if done well.

It would be stupid not to optimize it.

The problem is that Charlie never really left a clean, step-by-step instruction manual. The good news is that modern learning psychology did most of that work for us.

This post has a simple goal: to lay out the best method of learning for most people in most cases, what works, why it works, and how to apply it.

Here’s how to learn according to science.

Subscribe now

We Need Filters

If I ask you how to run in the best way, you’ll probably tell me it depends on a lot of things (terrain, race type, body shape, goals, etc.).

Learning works the same way. There’s no such thing as a perfect learning method.

So we’re going to filter learning techniques and keep the ones whose effectiveness doesn’t swing wildly across these four dimensions:

• Conditions: solo vs. group, dosage, timing, format (reading vs. listening, open vs. closed book, etc.)

• Learner profile: age, level, prior knowledge, abilities.

• Material: vocabulary, concepts, math problems, complex texts.

• Test: Criterion tasks: what you’re tested on → memory, comprehension, problem solving, transfer.

That’s exactly what Dunlosky & al. (2013)1 did. After reviewing a large body of evidence, they rate each technique’s relative utility:

• High utility

  • Practice testing (testing yourself)

  • Distributed practice (spacing)

• Moderate utility

  • Elaborative interrogation

  • Self-explanation

  • Interleaved practice

• Low utility

  • Summarizing

  • Highlighting/underlining

  • Rereading

  • Imagery for text

  • Keyword mnemonic

From here, the system designs itself: build the whole machine around the two high-utility pillars, then plug in the moderate-utility modules where they actually pay off (everything else stays secondary).

The Most Effective Method in Most Cases

Let’s take a few seconds to thank the authors of that paper, who probably spent thousands of combined hours grinding through scientific studies to arrive at this result.

In my opinion, their contribution to humanity through this work is massive, and I’m happy to be a small link in the chain that passes these ideas along.

Let’s begin.

1. Define how you’ll know you know

As with any task, you have to start by making explicit what “done” looks like.

So the first step is to define a proof, something that demonstrates you learned what you wanted, the way you wanted. For example, if you’re learning a:

• Concept → you can explain it with no notes, then give two counterexamples

• Procedure → you can redo the exercise cold, with no template.

• Domain → you can solve novel cases (not the ones you’ve already seen)

If you don’t define that proof, you’re very likely to optimize a proxy instead of the learning itself. It’s the finance equivalent of improving a ratio without improving the business.

2. Break the material into testable units

Your brain isn’t wired to evaluate “the whole” in isolation. It judges the whole by summing up the parts.

So except for the simplest cases, you have to convert the learning material (a course, a book, a video, etc.) into items: discrete things you want to be able to reproduce on demand.

Almost all items fall into four buckets:

• Facts (definitions, formulas, dates, vocabulary)

• Concepts (relationships, causes, limits, implications)

• Procedures (algorithms, methods, steps)

• Discrimination (choosing the right method among several similar ones)

This classification is what lets us move to the next step.

3. Turn every item into a question

Your brain remembers what you force it to retrieve. Turning each item into a question is how you optimize the material for the actual learning process.

In practice: write questions, not notes. Using the same four buckets:

• Fact → “What is X?”

• Concept → “Why does X imply Y?”

• Procedure → “How to solve / derive / apply ?”

• Discrimination → “In this case, which method and why?”

Over a long life of learning, your “question bank” is your main intangible asset.

4. Test yourself for real, using active recall

In learning psychology, we distinguish two ways of “retrieving” knowledge:

• recall (the answer comes to mind),

• recognition (you spot the right answer among others).

Recognition is much easier than recall. Recall requires a stronger memory trace, and more cognitive work to pull it out.

So here, forget recognition. Focus on recall, through practice testing:

• No information source available.

• One explicit question, with a clear answer you must produce, even if it’s imperfect.

  • You answer even if you have no idea, because the error will make the correction stick

• Then you correct by comparing your response to the right answer

Change one of those conditions and practice testing becomes less effective (that’s not what you want.)

I’ll go further: if you keep only one method from this entire post, make it this one. Practice testing has huge advantages:

• It works for most people, in most cases

• It’s cheap in time and easy to set up (especially with GenAI)

• Its effects show up across many retention intervals → from hours to months

It’s the Swiss Army knife of learning. Even so, it’s still more powerful when it’s not used alone, but embedded inside the full system.

5. Feedback. Always feedback.

The core of the process is practice testing: test + feedback.

But feedback is the real lever. In many cases, a weak test + strong feedback beats a strong test + weak feedback.

So you should overinvest in feedback quality. Every time your brain produces an answer and then receives feedback, it prunes some neural connections, strengthens others, and builds new ones.

Non-negotiable rules for high-quality feedback:

• Every answer gets corrected with one clear, explicit “best answer.”

• Every recurring mistake becomes its own item.

  • If you often get one part wrong, that part becomes a separate question with its own feedback.

• Retest quickly after an error.

  • Anywhere from a few minutes to 2–3 days max, depending on how “serious” the mistake was.

One important detail: feedback doesn’t have to be immediate. Delayed feedback can still be effective.

6. Space out your sessions

Spacing your learning sessions is a simple way to compound learning: for the same total study time, spreading practice over time is associated with better long-term retention than massed study (cramming)

The reason is biological (as it almost always is). When you do practice testing with an error + feedback, you trigger cascades of biological processes that take days and weeks to play out, ultimately rewiring your neurons and connections (in other words, learning). You can’t outrun your biology.

Spacing your sessions creates a kind of Lollapalooza effect2: multiple powerful forces (deliberate practice, motivation, biological rewiring, etc.) aligned in the same direction, toward learning.

A simple template (adjust to your time horizon):

• After creating an item: D+1, D+3, D+7, D+14, D+30

• If the item is easy: stretch the intervals faster

• If it’s fragile: shorten the intervals until it stabilizes

The goal is simple: get the same item right multiple times, across increasing intervals.

For implementation, I recommend Anki, a spaced-repetition flashcard tool that’s literally built for this and highly customizable to your goals (not a paid partnership). Some GenAI tools now offer similar systems as well.

7. Graft the “moderate-utility” modules in the right place

Practice testing + spacing is already an excellent method on its own. But under the right conditions, you can make it even better with a few add-ons.

Module A - Elaborative interrogation (“Why?”)

You take a statement and force your brain to generate a cause by answering: “Why?” (some would call it the most important question).

This was the most surprising part of the paper for me: elaborative interrogation isn’t that useful, in general.

The problem is cost (and it’s the main problem with most other techniques, too).

It can demand a lot of cognitive effort for limited return: you often need to go one “why” deeper, and you can end up spending your time inventing explanations instead of learning the material.

So the opportunity cost versus practice testing + spacing can be high.

It’s mostly useful for discrete facts/concepts that aren’t too complex. The rules are simple:

1. Identify a testable claim.

2. Ask: “Why?”.

3. Answer in 3/4 sentences max.

4. Get feedback: do a quick check to confirm your “why” actually holds.

Module B - Self-Explanation

Here you explain your own reasoning while you learn/solve, step by step.

More formally: you make explicit the intermediate steps of your information-processing, which helps link new knowledge to what you already know (and exposes gaps).

Its “moderate” rating comes from three issues:

• Limited scope: it mostly applies to procedures and some concepts.

• Time cost can explode (so does the opportunity cost).

• It’s easy to slip into the illusion of self-explanation by paraphrasing (chauffeur knowledge vs. Planck knowledge).

One key point: self-explanation works better during learning (concurrent self-explanation) than after the fact (retrospective self-explanation).

A simple method:

• At each step of a procedure or chain of reasoning, answer (briefly, in your own words) questions like:

  • “Why is this step valid?”

  • “What rule / principle did I just use?”

  • “What would change if assumption X were false?”

Module C - Interleaving (Mixed Practice)

This isn’t about alternating learning techniques. It’s about alternating the problems themselves.

I mention it because it’s directionally right, but it has three important limitations:

• It only works well in domains with a large variety of problem types. The strongest results show up in math learning.

• The literature is smaller and less decisive than for the other techniques.

• Interleaving only helps once you have baseline competence in each individual task, which is not always obvious in solo learning.

Now that we’ve laid the foundations, we can finally build the full system.

The Full System — Summary

• 0) Pre-setup

  • Define your goal + time horizon: what do you want to know, and by when?

• 1) Define the proof (how you’ll know you know)

  • Concept → explain it with no notes + give 2 counterexamples

  • Procedure → redo the exercise “cold,” with no template

  • Domain → solve novel cases (not the ones you’ve already seen)

• 2) Break the material into testable units

  • Facts (definitions, formulas, vocabulary)

  • Concepts (links, causes, limits, implications)

  • Procedures (methods, steps, algorithms)

  • Discrimination (choosing the right tool among similar options)

• 3) Turn every item into a question

  • Fact → “What is X?”

  • Concept → “Why does X imply Y?”

  • Procedure → “Solve / apply / derive.”

  • Discrimination → “In this case, which tool—and why?”

• 4) Active recall (practice testing)

  • Zero support (notes/book closed)

  • One question → one produced answer, even if imperfect

  • ALWAYS correct after (compare to the best answer)

• 5) Feedback (the most important part)

  • Every answer is ALWAYS corrected with a clear best answer

  • Every recurring mistake becomes its own item (or sub-item)

  • Retest quickly after an error (minutes to 1-3 days, depending on severity)

  • Useful note: feedback can be delayed, it doesn’t need to be instant

• 6) Space it (distributed practice) — calendar-based

  • Rule: for the same total time → spacing > cramming for long-term retention

  • Template: D+1, D+3, D+7, D+14, D+30 (adjust to your horizon)

  • If easy: stretch intervals faster / if fragile: shorten until stable

• 7) Add the “moderate-utility” modules in the right place (optional but recommended)

  • Module A → Elaborative interrogation (“Why?”): for simple facts/concepts; short answer; quick verification

  • Module B → Self-explanation: for procedures/reasoning; justify each step (“why valid?” “what rule?”)

  • Module C → Interleaving: mix problem types to train discrimination (after baseline competence)

• 8) What you reduce by default (low utility)

  • Rereading / highlighting / summaries / mnemonics / imagery.

  • You can use them only as prep for questions but never as the core of learning

• 9) Repeat for life.

To close, let me remind you that cognitive activity at any age is associated with a longer healthy life expectancy, and the association tends to be stronger when that cognitive activity is more intense.3

I wish you a long life filled with learning and reading.

Take care,

Flo

I spent hundreds of hours reading and listening to Charlie Munger to extract the mental models that, in my view, are the real gears behind his success.

I found 40.

Here they are.

Subscribe now

1. Be a Swiss Army Knife

Each discipline has a lot to offer. Knowing the big ideas from each one and making them your own gives you a more integrated view of the world and a more general way of approaching problems. The trade-off between the cost of not having these tools and the cost of acquiring them makes it an obvious choice.

“The man who needs a new thinking tool but hasn’t yet acquired it is already paying for it.”
Charlie Munger

2. Deserve It Until You Get It

The best way to get what you want is to act in a way that makes you worthy of it. Time will inevitably do its work.

It’s such a simple idea. It’s the golden rule, so to speak: You want to deliver to the world what you would buy if you were on the other end. […] By and large the people who have this ethos win in life and they don’t win just money, not just honors. They win the respect, the deserved trust of the people they deal with, and there is huge pleasure in life to be obtained from getting deserved trust.
Charlie Munger

3. Manage Happiness Through Expectations

Happiness is mainly the result of the gap between the outcomes of our actions and our expectations. By lowering our expectations, we can reduce a lot of unnecessary suffering (for example, by taking fewer risks to reach them).

“A happy life is very simple. The first rule of a happy life is low expectations. It’s one you can easily arrange. And if you have unrealistic expectations, you’re going to be miserable all your life.”
Charlie Munger

4. Independent Thinking to Counter Self-Serving Bias

You are (almost) the only person naturally incentivized to care about your success. Everyone who gives you an opinion or a business idea—even experts—has incentives that are not necessarily aligned with yours. Independent thinking is the only way to align your decisions and your incentives with your own success.

“You must have the confidence to override people with more credentials than you, whose cognition is impaired by incentive-caused bias or some similar psychological force that is obviously present.”
Charlie Munger

5. See the World Through the Lens of Incentives

Human behavior is driven by classical and operant conditioning. Over the long term, it is rewards and punishments—in other words, incentives—that shape these patterns of conditioning, and therefore behavior.

“Show me the incentive and I’ll show you the outcome.”
Charlie Munger

6. Reduce Competition Through Specialization

Those who perform the best are often the most specialized. The more specialized you are, the less competition you face and the more you’re rewarded for it.

“Specialization is the safest way out for most people, and for that reason the surgeons know more and more about less and less, and that’s what gets rewarded.”
Charlie Munger

7. Time Is Limited, Use Filters

The rarest resource we have is our time. It is essential to save it, especially in an age filled with noise. Use as many filters as possible to categorize as much information as possible, as quickly as possible (for example, the “too hard” pile).

“We have three baskets for investing: yes, no, and too tough to understand.”
Charlie Munger

8. Two Kinds of Knowledge

Not all knowledge is equal. Not all people who share knowledge are equal. Look for those who have a deep understanding of their subject and can adapt what they know to each situation, rather than someone who simply repeats what they think they know.

“In this world we have two kinds of knowledge. One is Planck knowledge—the people who really know. Then we’ve got the chauffeur knowledge—they have learned to prattle the talk and they have a big head of hair […] but in the end, it’s chauffeur knowledge.”
Charlie Munger

9. Checklist Against Complexity & Bias

Any process where error is too costly uses checklists (nuclear safety, aviation, investing, etc.). It’s a method that’s too simple and too effective at limiting potentially very costly mistakes not to be used.

“I’m a great believer in solving hard problems by using a checklist. You need to get all the likely and unlikely answers before you.”
Charlie Munger

Subscribe now

10. The Iron Prescription

Refuse the victim posture, even in an unfair world. In his early 50s, a botched cataract surgery left Charlie Munger blind in his left eye. Doctors warned he could lose vision in his remaining eye. Despite the unfairness of it, he didn’t waste time complaining and simply started taking Braille lessons—a pure display of pragmatism. Fortunately, the condition later eased, and he kept sight in his right eye.

“Whenever you think that some situation or some person is ruining your life, it’s actually you who are ruining your life. It’s such a simple idea. Feeling like a victim is a perfectly disastrous way to go through life. If you just take the attitude that, however bad it is in any way, it’s always your fault and you just fix it as best you can—the so-called ‘iron prescription.’”
Charlie Munger

11. Win-Win-Win

Value creation is more likely when all the players have an interest in keeping the system going. Here, the key actors are:

• Customers

• Companies (management, employees, etc.)

• Investors

The point is not that opportunities exist only in this kind of scenario, but that you can quickly and easily eliminate a large number of potential companies and focus on the signal when all types of actors win.

I had dinner with Charlie. I said, “Charlie, have you heard of a company called Credit Acceptance?” He said no. So I said, “Let me describe the business to you,” and I spent less than a minute explaining Credit Acceptance—exactly the way I just explained it to you.

He cut me off and said, “We’re not interested in that kind of business.” That really surprised me, so I asked why. He said, “We don’t want to do predatory lending. We don’t want win-lose business models.”

[…] Charlie’s standard was basically: you can’t make a lot of money by taking the high road. So he immediately rejected Credit Acceptance.

Source: Mohnish Pabrai’s Guest Lecture at University of Nebraska, Omaha on October 15, 2024

12. The Inversion Process

In most problems where uncertainty reigns, it can be very hard to reach an extremely desirable solution. In this kind of situation, an excellent heuristic is to invert the problem and avoid undesirable outcomes rather than trying to find the extremely desirable ones.

13. Probability Against Uncertainty

Uncertainty is a component of almost every environment, and definitely of investing. Probabilities are, among other things, a tool for measuring that uncertainty. Not thinking in terms of probabilities means dooming yourself to repeat the same mistakes over and over, sometimes with catastrophic consequences.

14. Opportunity Cost

Every limited resource (time, attention, money, relationships, etc.) that you spend on one thing is no longer available for all the others. Not taking these potential alternatives into account in your final decision means accumulating hidden costs.

I think this mental model of always thinking in terms of opportunity cost is the most important on this list in investing, and probably in life (although in a way, it is intrinsically linked to probabilities).

Opportunity cost is a superpower, to be used by all people who have any hope of getting the right answer.
Charlie Munger

15. Six Checks to Have a Better Life

It’s simple:

• You don’t have a lot of envy.

• You don’t have a lot of resentment.

• You don’t overspend your income.

• You stay cheerful in spite of your troubles.

• You deal with reliable people.

• You do what you’re supposed to do.

16. Vaguely Right > Precisely Wrong

The opportunity cost of precision (time, attention, etc.) is often much higher than the opportunity cost of being vague. When you add the uncertainty about actually being right and about the gains or losses tied to it, you often end up concluding that the opportunity cost of keeping things simple makes being vaguely right better than being precisely wrong.

“It isn’t that we were so good at doing things that were difficult. We were good at avoiding things that were difficult, finding things that were easy.”
Charlie Munger

17. Learn to Learn

Learning is something you’ll be doing all your life (hopefully). You’d better read the few pages of your brain’s “manual” that science has discovered so far so you can get better at it as quickly as possible.

“Without the method of learning, you’re like a one-legged man in an ass-kicking contest. It’s just not going to work very well.”
Charlie Munger

18. Learn to Unlearn

Our brain is wired to pay attention to information that aligns with our existing beliefs. This tendency limits our ability to unlearn false beliefs and therefore reduces the quality of our decisions.

“One should recognize reality even when one doesn’t like it, especially when one doesn’t like it.”
Charlie Munger

“It’s not the learning that’s so hard, it’s the unlearning.”
Charlie Munger

19. Make Friends with the Eminent Dead

It’s estimated that around 117 billion humans have lived on Earth. Only about 8.2 billion are alive today. A large share of existing wisdom was produced by people who are now dead. Charlie Munger himself is proof of that: he’s been among the eminent dead since late 2023.

Read the dead. Statistically, it’s more likely to be worth it.

“The man who doesn’t read good books has no advantage over the man who can’t read them.”
Mark Twain

20. Lollapalooza Effect

For the sake of simplicity, humans have classified the concepts they discover or invent into categories: physics, psychology, economics, etc.

However, the world is a continuum of structures and phenomena with no clear boundaries. Many phenomena arise from several disciplines, and many interact with each other across these borders.

The Lollapalooza effect is precisely when several forces (psychological, economic, social, environmental, etc.) act simultaneously in the same direction, creating an impact far more powerful than the sum of each force taken in isolation. In a sense, you move from addition to multiplication, leading to consequences that can be dramatic or exceptional.

“The lollapalooza effects can make you rich, or they can kill you.”

This was the first half. The other 20 are here. They're the better half.

Take care,

Flo

Subscribe now

This post was written in collaboration with the excellent , whose work I strongly encourage you to follow. If you enjoy mine, there’s a very high probability you’ll enjoy theirs too.

Join The Pursuit of Compounding

Every investor loves a great compounder: a business that can turn 1$ into 1.30$ year after year.

Look closely and you’ll almost always find a high ROIC (Return on Invested Capital) behind it.

But as investors, what we really care about are two things:

• whether today’s great compounders will stay that way, and

• which companies are on their way to becoming the next ones.

For that purpose, ROIC alone doesn’t help much.

Fortunately, there’s a ROIC-derived metric that’s much better suited to understanding the present and anticipating the future: ROIIC (Return on Incremental Invested Capital).

In this piece, we compare ROIC and ROIIC, and show how to use them in practice to judge the quality of a business.

ROIC

Definition

ROIC measures the return a company generates on the capital actually deployed to run and grow the business:

The numerator is NOPAT (Net Operating Profit After Tax):

The denominator is Invested Capital, which represents the capital employed in the company’s core operating activities:

Advantages

ROIC is one of the most widely used performance metrics, and for good reason:

1. Strong Indicator of Operating Quality

In the brutal world of creative destruction, sustaining high returns on invested capital over time is almost always a signal that something special is going on:

• disciplined capital allocation,

• a durable moat,

• pricing power,

• etc.

Consistently high ROIC is rarely an accident.

2. Direct Link to Value Creation

ROIC allows you to quickly assess whether a company has created value or destroyed it over its life:

• If ROIC > opportunity cost of capital (or WACC [Weighted Average Cost of Capital], for the purists), the company is creating value.

• If ROIC < opportunity cost of capital, it’s destroying value.

The comparison is conceptually simple and very powerful—on paper.

In practice, estimating the true opportunity cost of capital is hard for investors, which is why most people fall back on using an estimate of the WACC instead.

3. Easy Comparability Across Companies

Outside of a few specific sectors (banks, for instance, where ROE is usually more relevant), ROIC is broadly comparable:

• across sizes (small vs large caps),

• across industries,

• across business models.

This makes it a useful metric to rank and filter businesses on economic quality.

Here’s the distribution of ROIC by industry:

4. Aligned With Long-Term Investing Logic

Over the long run, one of the main drivers of shareholder value creation is a company’s ability to reinvest at high ROIC for many years. ROIC is essentially a snapshot, at a given point in time, of how effectively the company has deployed capital over its life so far.

The higher and more durable the ROIC, the more powerful the compounding engine for a long-term investor.

Drawbacks and Limits

A simple metric almost always comes with important limits. And ROIC is no exception.

1. Significant Sensitivity to Accounting Conventions

Here, both the numerator and the denominator are called into question.

Operating income depends on accounting standards related to:

• revenue recognition (subscriptions vs one-shot sales, cut-off, etc.);

• the treatment of certain expenses (capitalization vs immediate expensing);

• the categorization of exceptional or non-recurring items (restructurings, impairments, gains/losses on asset disposals, etc.).

It’s even trickier for invested capital:

• whether R&D expenses are capitalized or expensed;

• goodwill and intangible assets (especially for companies that do a lot of M&A and see their ROIC go down after each acquisition, even if their organic business hasn’t necessarily changed);

• how items are classified as operating vs non-operating, which changes what is included in “invested capital”.

Changes in accounting rules can also distort reported ROIC. The most relevant recent example is IFRS 16 on lease accounting, which artificially pushed assets up, and therefore invested capital.

That said, this sensitivity to accounting conventions is shared by almost all metrics. It is a drawback, yes, but a nearly ubiquitous one.

2. It Can Distort the True Performance of Companies with an Asset-Light Model

In asset-light models, invested capital is structurally low. With a smaller denominator, ROIC is mechanically higher.

But this does not necessarily reflect the company’s ability to reinvest at such returns in a sustainable way.

It’s quite common to see asset-heavy businesses become more asset-light in times of crisis (reduction of spending, closure of unprofitable stores, etc.), and ROIC tends to increase sharply.

In such cases, the business may be in distress, but the changes in ROIC tend to suggest that the business is improving.

3. It Mixes the Recent Past and the Distant Past

ROIC weights all historical capital similarly. A recent bad (good) capital allocation decision can be completely drowned out by many years of good (bad) capital allocation.

This becomes a major issue when a company is changing or developing a new business model, or going through a deep management transition.

It can take years before ROIC really reflects the impact of those more recent good or bad decisions.

Conversely, a young company can show ROIC figures that are totally out of proportion with its likely future economics, especially if it does a lot of M&A.

Yet for an investor, what matters is the present and the range of futures that can emerge from it.

And that’s exactly what ROIIC is designed to capture.

ROIIC

Definition

ROIIC (Return on Incremental Invested Capital) is the same concept as ROIC, but it focuses specifically on the change in invested capital and NOPAT over a defined period.

In other words, ROIIC measures how much additional NOPAT a company generates from the incremental invested capital over a given period (a quarter, a year).

Advantages

The power of ROIIC comes from the fact that you can choose the analysis period.

By adjusting the time window, you can:

• break a company’s history into coherent blocks (reflecting specific capital allocation phases), and

• measure, for each of them, the incremental returns generated by the new capital committed during that period.

As a result, instead of a single average return reflecting the company’s entire history, as with ROIC, ROIIC allows you to isolate exactly what you want to study:

• before / after a business model change

• before / after a major investment cycle

• before / after the arrival of new management, etc.

In general, the best predictor of the near future is the recent past. This ability to zoom in on the recent past is precisely what makes ROIIC so powerful.

In the current context of the AI revolution, this is particularly useful. Many companies are shifting toward massive AI-related CapEx, without it yet being clear whether these bets are rational or not.

By adjusting the period, you can:

• look at “pre-AI” ROIIC,

• then at ROIIC during the years when AI CapEx accelerates,

• and compare the quality of the dollars invested before and after this strategic shift.

This is exactly what we will do later with Meta: use ROIC and ROIIC to assess the true economic value of its AI investments.

However, here too, new advantages come with new drawbacks.

Drawbacks and Limits

The ability of ROIIC to focus on specific periods simply amplifies all the weaknesses already present in ROIC.

With ROIC, any impact on NOPAT or invested capital is diluted within a large set of data.

Here, we use a smaller slice of that data. The impacts themselves, however, do not necessarily shrink. As a result, their impact on the math becomes much larger. Here’s an example.

For ROIC:

• Total invested capital for ROIC = 100M

• NOPAT = 30M

• ROIC = 30 / 100 = 30%

For ROIIC:

• Incremental invested capital for ROIIC = 10M

• Incremental NOPAT = 3M

• ROIIC = 3 / 10 = 30%

Now assume a change in accounting standards that requires recognizing an additional 10M of invested capital (including over the period used to compute ROIIC).

New ROIC:

• Total invested capital = 110M

• NOPAT = 30M

• ROIC = 30 / 110 = 27.2%

New ROIIC:

• Incremental invested capital = 20M

• Incremental NOPAT = 3M

• ROIIC = 3 / 20 = 15%

We can clearly see that ROIIC is artificially depressed due to a purely accounting change.

If one still wants to use ROIIC under these conditions, one must make judgment calls with the objective of staying as close as possible to the company’s underlying economic reality.

And this brings us to the main “flaw” of ROIIC: it does not solve the problems of ROIC, it merely transfers them. More precisely, ROIIC fixes the “average vs incremental” issue, but at the cost of worsening the “accounting vs economic” issue.

It is therefore a tool with very little room for error, requiring a deep understanding of the situation, as well as of where and when to apply it, in order to produce results that are both meaningful and interpretable.

With this framework in place, let’s move on to practice with Meta.

ROIC/ROIIC Framework Applied to Meta

A hot topic among investors in recent years has been the return on AI investments at the hyperscalers in the S&P 500. Not only have AI investments been growing at a rapid clip, but they are expected to continue to grow into 2027.

In October 2025, Meta reported record Q3 results, beating both top and bottom line estimates, and yet the shares sold off nearly 10%. The company also raised its 2025 guidance for capital expenditures, which will now come in the range of $70 billion to $72 billion, where the prior outlook was between the range of $66 to $72 billion.

Over the past five years, Meta’s CapEx has already grown at nearly 35% annualized, while becoming an increasingly larger part of its Operating Cash Flow.

“…being able to make a significantly larger investment here is very likely to be a profitable thing over, over some period,”
Mark Zuckerberg — Founder, Chairman & CEO of Meta — Q3 2025 Conference Call

Is Zuck lighting money on fire? Is this yet another example of the AI bubble? Or is Meta laying the groundwork for future growth?

For this case study, we will apply the “Operating Approach” framework of Michael Mauboussin to calculate ROIC and ROIIC, using NOPAT.

Michael Mauboussin’s framework focuses on isolating the operating performance of the firm from its financing decisions and non-operating assets. This provides a clear view of the “cash-on-cash” return of the business operations.

Further, we’ll examine it against the approach of John Huber, who uses Net Income instead of NOPAT, with different adjustments to the Invested Capital Base. The overarching goal is to assess Meta’s return on its AI investments from 2020 to 2024.

Capital Light to Capital Intensive

Meta entered 2020 as a highly efficient, asset-light social media (software) utility benefiting from a pandemic-induced digital advertising boom.

By 2022, it faced an existential crisis of capital efficiency, characterized by rising expenses, a degraded advertising signal due to Apple’s iOS privacy changes, and a skeptical market reaction to its massive Reality Labs investments.

“We continue to see a lot of opportunity across our investment priorities and remain committed to dedicating additional talent and capital toward these areas while ensuring our investment plans are appropriately calibrated to the operating environment”
Mark Zuckerberg — Q1 2022 Conference Call

However, the subsequent fiscal years of 2023 and 2024 demonstrate a profound turnaround — stemming from the “Year of Efficiency” — that morphed into an AI-driven renaissance.

We will show that this has validated the heavy capital deployments of the prior cycle, evolving Meta into a hardware-intensive digital utility.

Return on Invested Capital Calculation for Meta

NOPAT Calculation for Meta

Meta has both tangible (hardware) and intangible (code, intellectual property) assets. We need to properly account for this in the NOPAT calculation. As Mauboussin states:

“We want to put all companies on the same footing in order to measure ROIC consistently. The way to do that is to put intangible assets on the balance sheet and amortize them over their useful lives. We will call the line item “capitalized intangibles, net.” This way we treat all investments alike.”

Thus, in doing the NOPAT calculation, we will make a reasonable attempt to estimate the amortization of the intangibles.

Table 1: NOPAT Calculation for Meta from 2020 to 2024

Between 2020 and 2024 Meta experienced substantial volatility in its NOPAT. The jump in 2021 reflects the pandemic pull-forward. The collapse in 2022 (-39.9%) was driven by a perfect storm:

• the revenue decline (weaker demand for ads),

• combined with a massive OpEx ramp (headcount added in anticipation of growth that didn’t materialize),

• and Apple’s App Tracking Transparency (ATT) headwinds.

Is it any surprise Meta shares were down ~75% in 2022?

The NOPAT recovery in 2023 and 2024 was equally dramatic, driven by the “Year of Efficiency” cost cuts, as well as the accelerating monetization of Reels and Advantage + AI tools.

Invested Capital Calculation for Meta

The Invested Capital base expanded from $79.5 billion in 2020 to $163.8 billion in 2024, even after stripping out the non-operating assets and adding back PP&E, Operating Leases, Goodwill, and Intangibles.

Table 2: Invested Capital Calculation for Meta from 2020 to 2024

The defining feature of this table is the explosion in Net PP&E, which nearly tripled from $45.6 billion in 2020 to $121.3 billion in 2024. This represents the physical manifestation of Meta’s strategy: data centers, servers, and network infrastructure.

Note also the shift in Net Working Capital to a negative figure from 2022 onwards. Meta is employing a Negative Net Working Capital strategy, financing its operations using its suppliers’ balance sheets (Accounts Payable), all while collecting cash from advertisers relatively quickly.

This negative working capital acts as a source of funds, partially offsetting the massive PP&E spend.

That’s not the only creative use of their balance sheet - in the fall of 2025 Meta announced a Joint Venture with Blue Owl to build the Hyperon data centre. However, this applies to Fiscal 2025 and so we’ll only have to consider this when we calculate the ROIC and ROIIC once the full year has been reported.

We can now calculate Meta’s ROIC.

ROIC Calculation for Meta

Table 3: ROIC Calculation for Meta

Despite the doubling of the capital base from 2022 to 2024, the Return on Invested Capital (ROIC) has demonstrated remarkable resilience, rebounding from a nadir of 21.0% in 2022 to 29.6% in 2024.

This suggests that the “denominator problem” — the rapid expansion of the asset base — has been matched by an equally potent expansion in the numerator (NOPAT), driven by AI-enhanced monetization:

“advertisers are increasingly just going to be able to give us a business objective and give us a credit card or bank account and like have the AI system basically figure out everything else that’s necessary including generating video or different types of creative that might resonate with different people that are personalized in different ways, finding who the right customers are.” Mark Zuckerberg — Q3 2025 Conference Call

This ROIC analysis provides quantitative evidence that the aggressive capital deployment of 2022, which was heavily criticized at the time, was not value-destructive. Instead, it led to the explosive NOPAT growth seen in 2023 and 2024.

We can see that the investments made in GPU clusters and data center re-architecture sowed the seeds for resurgence in 2023 and 2024. The GPU clusters bought in 2022 began powering the discovery engine, increasing time spent on Reels, which in turn attracted ad dollars.

This also contradicts the narrative that capital intensity inevitably degrades ROIC. Based on this analysis, the lag time for AI infrastructure CapEx to start generating returns appears to be approximately 18 to 24 months.

The recent past looks great for Meta! What do the incremental returns look like? Is Meta maintaining its status as a great compounder?

ROIIC Calculation for Meta

Let’s go one step further and calculate the ROIIC according to Mauboussin’s framework.

Table 4: ROIIC Calculation for Meta.

The -31.5% ROIIC in 2022 is an example of the J-Curve that can occur in capital investments. Meta deployed $36.1 billion in incremental capital, yet its NOPAT significantly contracted. In a vacuum, this looks like value destruction.

However, in the context of Meta, it represents the lag from when the capital investments occur, versus when they begin to generate NOPAT. Data centers built in 2022 do not come online instantly; GPU clusters require configuration and software optimization before they can drive ad ranking improvements.

Over the entire period (2020-2024) Meta, however, generated $36 billion in new NOPAT on an $133 billion expansion of their invested capital, for a ROIIC of 27% over the entire period.

This implies that the internal rate of return (IRR) on Meta’s AI infrastructure is exceptionally high. This “ever-escalating spend” is not vanishing into a black hole where capital never returns — it is being converted into NOPAT at a rate of about 27 cents on the dollar annually.

Huber vs. Mauboussin

John Huber’s framework provides a different lens. He instead looks at Net Income (which includes interest income) and Tangible Capital (excluding Goodwill and Intangibles), to assess the compounded return.

By stripping these out, he instead focuses on the financial asset return (including interest income from Meta’s massive cash pile). Huber assesses the value creation from all tangible parts of the business, not just from the underlying operations.

Table 5: Tangible Capital Calculation for Meta

If we take the change in Net Income divided by the change in the Tangible Capital base, we can calculate Huber’s ROIIC for META.

Table 6: Huber ROIIC Calculation for Meta

Huber’s ROIIC metric suggests Meta is compounding at rates exceeding 45% on incremental invested tangible capital, contrasted to Mauboussin’s 27% on incremental invested total capital.

Figure 1: Calculated ROIIC for Meta using both Mauboussin and Huber methodology

Huber’s use of Net Income captures the benefit of Meta’s fortress balance sheet. In 2023 and 2024, interest rates surged. Meta, holding roughly $70+ billion in cash and marketable securities, generated substantial Interest Income.

Mauboussin’s NOPAT excludes this to focus on operating purity. Huber’s approach implicitly argues that this interest income is part of the shareholder’s return stream.

Further, Huber’s exclusion of Goodwill / Intangibles makes the capital base smaller. Since Meta’s recent growth has been organic (CapEx) rather than acquisitive (M&A), the change in capital is similar, but the base efficiency looks higher under Huber’s lens. However, much of the assets created by R&D are intangible in nature, and so Huber’s approach accounts for the Net Income increase in these assets, without including them in the invested capital base.

Which metric you focus on depends on what the goal is:

• If the goal is specifically to look at the returns on CapEx spend, then Mauboussin’s approach provides a clearer picture.

• If the goal is to assess the productivity of all tangible assets (including cash), then Huber’s approach is superior.

Meta has guided for 2025 CapEx to be in the range of $70-$72 billion, which is a staggering increase from the ~$37B in 2024. The payoff for the 2025 spend should be judged by the NOPAT in 2026 and 2027, given the 18-24 month delay.

What About Reality Labs?

Mauboussin’s invested capital base calculation includes all of the Reality Labs investments, while Huber’s invested capital base would exclude some (intangibles).

This segment acts as a massive drag on consolidated returns, effectively a venture capital bet funded by the core business. If we isolate the Family of Apps (FoA), the ROIC likely exceeds 60%, positioning it among the highest-quality businesses in history.

However, the stability of the consolidated ROIC above 40% implies that the core business is robust enough to subsidize this bet without destroying aggregate shareholder value, provided the spread between ROIC and the WACC remains wide.

We view Reality Labs as optionality. Yes, the returns would be even higher without this loss-making segment, although it could also spawn the next generation growth segment for Meta.

Closing Thoughts

ROIC and ROIIC are not panaceas and do have their limitations.

Even so, they are among the best tools we have as long-term investors to measure value creation objectively.

More specifically:

• ROIC is a good reflection of a company’s economic quality over its entire life.

• ROIIC measures the quality of recent capital-allocation decisions and gives a glimpse of what the company is in the process of becoming.

• Looking at the two side by side helps you judge whether its value-creation trajectory is improving, deteriorating, or broadly stable over time.

Meta’s trajectory from 2020 to 2024 offers us a textbook case study of theory in action. It shows us that Meta’s historical—and incremental—investments have created immense value. It appears that, overall, their AI investments are indeed paying off and investor concerns about their spend are unfounded.

The case study also highlights a methodological approach for tracking invested capital returns as a corporation shifts its strategy (for example, from capital-light to capital-intensive.)

Further, we can see how adjustments in working capital (Mauboussin vs Huber) are required depending on what specific information the investor is trying to parse out.

Each company is a unique case, and there are no universal rules.

That’s the good news: precisely because there isn’t a fixed playbook, time, experience, and a stubborn curiosity about how businesses really work can compound into a real edge.

Subscribe to The Pursuit of Compounding

Subscribe now

Disclaimer: We are private investors, not financial advisors. This post is for educational purposes only and does not constitute financial advice. Meta Platforms, Inc. (META) is a stock owned by The Pursuit of Compounding. Always conduct your own due diligence before making any investment decisions.

Nature is too complex for our limited brains to fully grasp. So we cheat a little. We use models: simplified versions of reality. Not to tell “the truth,” but to make thinking possible.

George Box put it perfectly: “All models are wrong, but some are useful.” Useful because they capture just enough—and stay simple enough.

The price-to-earnings ratio (P/E) ticks those boxes. It claims to get a handle on something profoundly complex—the value of a business—by using something simple on the surface: a price divided by earnings.

However, its apparent ease of use has “democratized” it in a way that has distorted its original purpose—to the point that the P/E is probably the most misunderstood ratio in finance.

Not All Earnings Are Equal

Take two companies. Both trade at a P/E of 15, have the same market cap, and posted the same EPS last year.

On paper, they look comparable. But suppose I tell you that:

• Company 1 sells software, with recurring revenue, strong customer stickiness, high margins, and little CapEx.

• Company 2 makes steel, with a heavy balance sheet, pronounced cyclicality, volatile margins, and a lot of CapEx.

At that point, it’s obvious that “P/E = 15” is not describing the same reality.

And that’s normal. The P/E is not a direct measure of “value.” It’s a price divided by an accounting profit per share at a point in time. And accounting profit is probably one of the most unstable objects in finance.

It depends on:

• accounting standards (revenue recognition, depreciation, provisions, stock-based compensation (SBC), impairments),

• capital structure/allocation (debt levels and therefore interest expense, buybacks, dilution),

• management choices (provisioning policy, investment timing, capitalizing vs expensing, one-offs, deferred revenue),

• and the nature of the industry (cash cycle, revenue cyclicality, capital intensity, pricing power).

A P/E only becomes interpretable once you’ve answered the question: “What kind of E is this?”

And that’s where most investors get it wrong.

Fix the E, Not the P.

A common way to use the P/E is to hold P fixed and let E float. Do that, and you can bend “reality” however you want, along with all the usual biases: cherry-picking, convenient adjustments, mixing up cyclical with durable, and so on.

From a valuation standpoint, it’s like shooting an arrow at random… then drawing the bull’s-eye around where it landed.

The opposite approach—the relevant one—is to fix E. That pins down at least one reality, and therefore a set of implicit assumptions.

For example, if you assume +10% EPS growth per year, you’re imposing a constraint: what combinations of price/volume/margins/CapEx/dilution can actually produce that +10%? And which ones can’t?

The ratio only becomes useful at that point: once E is set, P simply tells you what the market is willing to pay for those scenarios. Then, with what you know about the company, the industry, the management team, etc., you can narrow the scenario set and refine their probabilities.

For instance, if the company is already running at full capacity, that +10% can’t come from volume. It has to come from higher prices, a better product mix, or higher margins.

The P/E Is a Price on Assumptions

The P/E isn’t a verdict. It’s a starting point—then you work backward to uncover the assumptions baked into it.

In other words: what has to be true for 15x to be rational? And if I don’t believe it, what is the ratio actually telling me?

To make that concrete, here’s the checklist I force myself to fill out every time I lean on the P/E (which is less and less often):

• Which “E” is the right one, and why?
  TTM / Forward / Adjusted / Normalized. Make the choice upfront, justify it, and stick to it to limit bias.

• What are the one or two main drivers of growth?
  Price / Volume / Margins / Mix / M&A / Net buybacks. If the business analysis doesn’t give a clear answer here, it’s better not to use the P/E at all.

• What does that growth cost?
  CapEx, working capital, R&D, promotions, SBC, etc. The answer is rarely obvious (and often a mix), but every component has to be backed by something concrete: past numbers, management guidance, explicit decisions.

• Does the story still work per share?
  Growing total earnings at the expense of EPS is a real risk. The share count is part of the model.

• What would automatically invalidate my scenario?
  It should almost always be a specific, binary fact tied to a timeframe. Example: operating margin doesn’t improve over the next four quarters.

After you’ve answered those questions, there’s only one thing left to do: estimate how likely the path is that would make your E true.

In other words: does your E happen under conservative assumptions (so you have a margin of safety), under ordinary ones, under heroic ones, or under assumptions that are basically impossible?

Take a simple example. You fix EPS growth at +10% per year.

• If the company already has proven pricing power, recurring revenue, and doesn’t need much CapEx to grow, then +10% is a pretty ordinary (even conservative) assumption.

• But if that same +10% requires volume growth, margin expansion, and CapEx funded by issuing new shares, then +10% becomes a heroic case.

Same E. Very different probability. So the acceptable price can’t be the same.

The P/E isn’t a shortcut to value. It’s a shortcut to a story. And that shortcut is only useful if you’re willing to read that story and spell out exactly what would make it true… or false.

But in the end, the P/E is still anchored to an accounting concept—earnings—that can be shaped too easily:

“Common yardsticks such as dividend yield, the ratio of price to earnings or to book value, and even growth rates have nothing to do with valuation, except to the extent they provide clues to the amount and timing of cash flows into and from the business.”

Warren Buffett — Letter to Shareholders (2000)

That’s why I tend to prefer other metrics, especially those built on enterprise value (EV) and cash flows, when assessing valuations. But I’ll save that for future posts.

Thank you for your time and attention.

Take care,

Flo

Subscribe now