Psychology of memory, Sotheby’s, reflections on brands, LLM slop writing, the Cuban endgame, Blank Street, Dostoevsky, flight turbulence, and more. Enjoy!

(1) Some People Can’t See Mental Images. The Consequences Are Profound.

Not everyone has “mental imagery.” This condition, called aphantasia, results in a lack of visual memories—the inability to picture the face of someone, recall the imagery of past experiences, and so on. Mental visualization influence our emotions, so people with aphantasia have weaker emotions because they can’t visually relive past joys or traumas. On the other hand, extremely strong visual memories (called hyperphantasia) are an emotional amplifier. Aphantasia shows the deeper connection between mental visualization, memories, and emotions: “Many could remember very little about their own lives. Even for the events they did remember they could not muster the feeling of what those had been like. They knew some things had made them happy and others had made them sad, but that knowledge was factual, and it did not evoke any emotions in the present. We believe that our selves are made up largely of memories, and that the loss of those memories would be a self-ending catastrophe. But he knew now that there were also thousands of people just like him, who had work and marriages and ideas and thwarted desires and good days and bad days. All they lacked was a past.” They have the eternal sunshine of the spotless mind.

(2) How A Billionaire Owner Brought Turmoil And Trouble To Sotheby’s

The best parts of this article are the mechanics of Sotheby’s auction business. The auctioneer’s job is to dramatize the possibilities of a sale. For major consignors, Sotheby’s waives the seller’s fee and brings in third-party guarantors to backstop the price.. Private sales, where transactions are brokered outside the auction room, have become an increasingly important revenue line. The business model runs on expensive relationships and theatrical generosity, which is exactly where Drahi’s cost-cutting instincts collided with reality. Auction economics reward lavish spending on client relationships, so penny-pinching with a “relationship” business tends to be short-sighted. Unfortunately for Drahi, it sounds like cable-style cost-cutting doesn’t work with cultural assets.

(3) The Brand Age

The latest from Paul Graham: “Brand is what is left when the substantive differences between products disappear. But making substantive differences between products disappear is what technology naturally tends to do.” He covers the story of the Swiss watch industry, which differentiated itself with thinness and the accuracy of mechanical movements until the quartz-based watches became cheaper, thinner, and more accurate. After the “quartz crisis”, Swiss watches shifted into brand marketing, distinguishing themselves as luxury items and selling scarcity as the core value proposition. Watches turned into men’s jewelry, male equivalents of Cartier bracelets. Graham is critical of brands: they shift the focus away from technical quality, cater to the base psychology of customers (wanting to feel superior with status symbols), and they’re “a boring problem.” Not to relate everything to AI, but… if everyone can build the same software, does software become brand-focused? Or is software already brand-focused? How much hate for Microsoft Teams is due to the connotations?

(4) Against Sloppypasta: Don’t Paste Raw LLM Content at People

Until LLMs, writing long documents was “proof of work.” The process of writing, both the time required and the reflection that it produces, was implicit evidence that the author had thought deeply about the problem. LLMs have created the novel issue of “sloppypasta,” where anyone can instantly generate a mass of thoughtful-sounding text in response to anything. I think this will become a real issue in companies… AI writing long memos and emails, people asking AI to summarize them, and everyone outsourcing their thinking to LLMs. The article makes an excellent argument that sloppypasta is bad for both the creators and recipients: creators lose credibility and understanding, while recipients take an asymmetric burden of reading, vetting, and acting on the information. Lots of great quotes in the article, but the best argument for writing yourself (and only using AI for research/assistance) is to reduce personal brain rot: “Use AI to accelerate your work or improve what you send. Don’t use it to replace thinking about what you’re sending.”

(5) Cuba’s Broken Economy Leaves It At Donald Trump’s Mercy

Good profile of Cuba’s economic crisis. The embargo is strangling Cuba’s economy: GDP shrank 7% last year, population has declined over 20% since 2020, and foreign reserves are exhausted. Their main problem is a lack of fuel: the Russian embargo and Venezuelan coup eliminated their access to oil (which powers the electricity grid), resulting in constant blackouts and driving most factories to a halt. The additional pressure on exports, especially tourism and doctors sent to other countries, have cut off remaining sources of FX. Trump and Rubio are not calling for regime change just yet—demands currently focus on liberalizing the private sector (and probably putting Trump’s name on the Havana Libre). What do the locals think? “Not everyone wants a McDonald’s on the Malecón, Havana’s seafront promenade, but they do want an economy that functions.” Capitalism’s final victory will be the Malecón McDonald’s.

(6) Blank Street: Coffee, Counter-Positioning, and Collateral Damage

Blank Street is a venture-backed coffee chain that launched in NYC in 2020. Their model is classic disruption (in the Clayton Christensen sense): counter-positioning against the higher-priced and large footprint Starbucks locations with smaller stores focused on a limited selection of cheaper, fast turnaround, grab-and-go coffees. Blank Street is popular with investors for their unit economics, but their defensibility is an open question: “It would be painful for Starbucks, with its existing footprint and cost base, to compete with Blank Street on price, [but] there are not many barriers to entry preventing another company from replicating its model. [They are] a SoftBank Vision Fund 3 away from having direct competitors. Attractive unit economics are a necessary but insufficient condition for building a sustainable business. Electric scooter rental businesses like Bird and Lime had positive unit economics.” I’d probably bet on Blank Street’s model.

(7) Review of Demons by Fyodor Dostoevsky

For those just getting into Russian literature… Dostoevsky’s timeless mirror of society: “The ideas don’t matter, because at the end of the day they’re pretexts for desires—the desire to dominate, the desire to obliterate the world, the desire to obliterate the self, the desire to negate. […The] desire for status came first and wrapped itself in liberal politics in order to reproduce and advance itself […their] radical politics serve only as a mechanism of self-justification and a lever to pull. It’s not a novel about people, and it’s also not a novel about ideas. It’s actually a novel about desires, motives, urges, and the ways in which we construct stories to make sense of them. But where do desires come from?”

(8) A Love Letter to Novels

Fiction and feelings: “When read at the right time, a novel really can change you, the way the kindness of a stranger might alter the course of your day or the words of a mentor might shift your perception of a problem you thought was impossible to solve. Many people will pick up a book during a dark period in their lives and emerge from their reading experience transformed. For example, the emotions and events depicted in a novel [can] help you realize that you’re not alone, as famously noted by James Baldwin: You think that your pain and your heartbreak are unprecedented in the history of the world, but then you read [a book]. Dostoevsky and Dickens taught me that the things that tormented me most were the very things that connected me with all the people who were alive, or who ever had been alive. […] Timing is crucial. It doesn’t matter how well-written or revered a book might be; you have to read it at the right time for it to truly change you. […] Having a personal library with unread books […] is partly what allows someone to choose the right books at the right time. […] The stories are how we make sense of our lives.”

(9) Get Married Young

“The happiest young women (22-35) today are not footloose and fancy free, they are married moms. The ones least likely to be happy are single and childless. Data from the General Social Survey indicates that 41% of young married moms (22-35) are “very happy” compared to just 14% of their female peers who are single and childless. […] What about young men? […] Young married men (22-35) who are married with children are almost three times as likely to be “very happy” with their lives compared to their peers who are single and childless. 14% of young men who are single and childless are “very happy” compared to 37% of their peers who are married fathers. […] These data suggest young men and women who take their cues from pop culture—shows like “Friends” that celebrate single life in New York City—may be in for a rude surprise.”

(10) Why Your Next Flight Is Likelier to Hit Turbulence

Clear-air turbulence over the North Atlantic has increased 55% since 1979. We don’t know why, we can’t see it on radar, and we can’t predict where it will come next. The article has raises lots of alarms about turbulence, but the combined statistics made me less concerned about the risk: only 1 in 3,000 US flights encounter severe turbulence, nearly all of the 40 annual US deaths from turbulence occur on small private aircraft, and 80% of the remainder are flight attendants (often because they aren’t wearing seatbelts). The odds of dying on a US flight are only 1 in 46 million, and turbulence risk continues to decline because of engineering—sensors measure wind force in real time and auto-adjust the flaps/spoilers/ailerons to counteract turbulence, aggregated real-time data from flights can flag turbulence zones to other planes, and more. But moderate turbulence is still unsettling… so if you’re flying over Colorado or Chile (or any mountain ranges), buckle up!

Miscellaneous

Foreign policy in the Marco Rubio era. Buying a Romanian election with TikTok gifts. Explaining the global success of Pokemon. Reflections on the Reynolds Pamphlet. The expansion of Heathrow Airport. Solar panels in Cuba. How Swig’s dirty soda became a national obsession. AI is accelerating GMO fruits. Cost estimates for datacenters in space. AI could be reducing energy costs by increasing non-peak utilization. Tricolor accounting fraud. Inside the attempted coup in Benin. AI breakthrough in theoretical physics. AI assistance in building military target banks. New prime minister of Nepal. Declining cost of entrepreneurship. The expanding NVIDIA empire. The parsley test.

AI, SaaS, Napoleon, Flightradar24, Minneapolis, Filipino fast food, origins of finance in Bruges, diamonds, Anna Karenina, and Dante’s ninth circle. Enjoy!

(1) Something Big is Happening

Most people don’t realize how fast AI is improving: “The models today are unrecognizable from what existed even six months ago. [Let] me make the pace of improvement concrete: [in] 2022, AI couldn’t do basic arithmetic reliably. It would confidently tell you that 7 × 8 = 54. By 2023, it could pass the bar exam. By 2024, it could write working software and explain graduate-level science. By late 2025, some of the best engineers in the world said they had handed over most of their coding work to AI. On February 5th, 2026, new models arrived that made everything before them feel like a different era. If you haven’t tried AI in the last few months, what exists today would be unrecognizable to you. […] Nothing that can be done on a computer is safe in the medium term. If your job happens on a screen (if the core of what you do is reading, writing, analyzing, deciding, communicating through a keyboard) then AI is coming for significant parts of it. The timeline isn’t “someday.” It’s already started.” Based on my own experience, I’d agree.

(2) 10 Years Building Vertical Software: My Perspective on the Selloff

Public market software is selling off due to the fear that AI will eliminate their moats (creating competing software is easier, switching costs are lower because unstructured data is easier to re-structure, etc.). This article covers the various moats of vertical market software (VMS) and analyzes what changes, if anything, with AI. The moats that are lost: muscle memory; business logic in custom workflows (“LLMs turn all of this into a markdown file”); engineering talent (is this really a moat?); and bundling. The moats that remain: proprietary data (a consensus view); regulatory lock-in because “HIPAA doesn’t care about LLMs” (I think this is underrated even in unregulated markets… good luck getting F500 adoption of vibe-coded Salesforce); transaction-level embedding (e.g., payment processors); and network effects. Regulatory lock-in is the most durable moat remaining: “If your moat is regulatory certification, compliance infrastructure, and deep integration within mission-critical workflows, LLMs are barely relevant to your competitive position. […] Healthcare EHR systems with HIPAA compliance and FDA validation. Life sciences platforms with regulatory lock-in. Financial compliance and reporting infrastructure.” Everyone hates regulation until it protects their special interest!

(3) 18 Brumaire Year VIII (November 9, 1799)

Napoleon’s seizure of power in France was a messy affair. This article covers the play-by-play of 18 Brumaire (the revolutionary calendar date for his coup). Napoleon had coordinated with Sieyès and his brother Lucien (the president of the Council of Five Hundred) to fabricate a Jacobin terror plot, justifying the government’s relocation to suburban St. Cloud, where military control was easier than in Paris. When legislators realized that a coup was in progress, they physically attacked Napoleon on the floor and voted to have him executed. Lucien saved the coup at the last moment by sending in Napoleon’s grenadiers to “restore order.” By midnight, the deputies dissolved the Directory and named Napoleon as First Consul. Historians often see Napoleon’s rise as inevitable, but in reality, the outcome was a coin flip between execution and empire.

(4) Flying Under the Radar with Flightradar24

Flightradar24 is a popular app for tracking flights. What began in 2006 as a hobbyist website for plane enthusiasts has grown into a huge, profitable, privately-owned mobile app (recently valued at $1.5 billion). Flightradar24 has a unique data advantage with over 40,000 ADS-B receivers, mostly owned by hobbyists, that share real-time data with them, and which is monetized both as a consumer-facing app and a B2B data-stream. The article has a lot more info about the nuances of their business model—creating content for SEO, using their newsletter to get new transponder volunteers, partnering with easyJet, the impact of Apple enabling subscriptions on iOS (and the tradeoffs vs ad-supported monetization), and more. Not bad for a company that is “just an app.”

(5) The Battle for Minneapolis

Reflections on the ICE showdown in Minnesota: “When you grow up in Minneapolis, then you leave it, two truths become apparent: winters everywhere else seem fake, and no other city’s politics seem as doggedly progressive, not even New York’s.” They’re fighting a very different kind of ice this winter… “Thousands of Twin Cities residents have started observing ICE, from cars or on foot, since the surge began. Others volunteer to drive workers to their jobs, monitor school drop-offs and pickups, or deliver groceries to people too scared to leave their homes.” And some heartwarming, Minnesota nice sentiment: “I don’t care about your documentation status. You’re my neighbor. […] The people they rounded up, they can drive in the snow, they have good coats: they’re Minnesotans.” Quite a messy month in the homeland.

(6) Can Jollibee Beat American Fast Food at Its Own Game?

The global fast food industry is almost entirely American: McDonald’s, KFC, Subway, Starbucks, etc. Remarkably few chains flow the other direction. Jollibee from the Philippines is a rare non-US contender. The menu (sweet spaghetti, peach-mango pies, garlic-seasoned fried chicken) is catered towards Filipino tastes, and its 80 US locations are in cities with Filipino diaspora. While they have an ongoing challenge of adapting to US consumer tastes, their wider portfolio is growing. Outside of the main brand, Jollibee’s parent company operates over 10,000 stores globally with 19 brands, having acquired Smashburger and Coffee Bean & Tea Leaf. But it’s a long road ahead for mainstream awareness: even 500 locations is less than McDonald’s in Ohio alone.

(7) The Innkeeper of Bruges

Bruges was among the most important cities in medieval Europe. The first “modern” financial exchange (focused on FX, commodities, and futures) emerged organically in Bruges among a group of Hanseatic merchants at an inn run by the Van Der Beurse family (whose name is the origin of “bourse” in many languages). The pieces of a modern exchange emerged gradually. First, the inn itself acted as the trusted intermediary guaranteeing payments between parties (centralized clearing and settlement). Second, commercial courts called the Vierschaar were created to resolve disputes (binding arbitration). Third, to handle counterparty risk, a formal “protest system” for registering debts and defaults with notaries functioned as an early credit bureau, creating a shared blacklist that could follow the merchants’ reputations across Europe. Modern finance started in Bruges and spread worldwide.

(8) How The Diamond Industry Lost Its Sparkle

Diamonds aren’t forever after all. Chinese factories can mass-produce 3-carat lab-grown diamonds in 7 days using machines that cost around $280,000 each. A single factory in Zhengzhou can produce over 100,000 carats per month, and about 70% of the world’s lab-grown diamonds for jewelry now originate from Chinese factories (mostly in Henan). The pricing pressure is intense: “a three-carat lab-grown stone now sells for just 7 per cent of the price of a mined equivalent.” In fact, prices are falling so fast that China put a domestic price floor on diamonds at $15/carat, and the cost to polish and transport them to Antwerp is now more than the cost of production: “It costs $56 to polish one carat in China, more than it costs to produce the stone.” The pricing for lab-grown diamonds is at the price floor of $15/carat. Still higher than the resale value.

(9) Is “Anna Karenina” A Love Story?

“Deciding what to think of Anna is one of the central challenges with “Anna Karenina.” […] Anna does bad things [and] she underestimates just how bad the consequences of those things will be. […] Tolstoy doesn’t know exactly how to think about Anna’s role in her own downfall. He believes we make choices, and that our sense of free will is based on something real. But he also has a deep respect for the complexity and power of our circumstances, and he considers our personalities and psychologies to be “circumstances” too. […] She may have made an unwise choice, giving into desires she should have resisted because she underestimated how unyielding the world would be. We will never know what happened, exactly, just as Anna could not know. That is one of the dreadful lessons of Anna’s story: she could not distinguish between what she was choosing to do and what she was driven to do.” Unfortunately, her clarity came too late to avoid the consequences: “The light […] flared up more brightly than ever before, lighting up for her all that had been in darkness, flickered, began to grow dim, and was extinguished forever.” At least Levin had a happy ending.

(10) Dante’s Ninth Circle of Hell

Dante’s Divine Comedy is a medieval poem about sin and salvation. Its most famous section, Inferno, created the modern depiction of hell as a “fiery inferno” with nine descending circles of sins. But Dante’s lowest circle of hell was ice cold, alluding to the cold and calculating nature of the most unforgivable sin: “In the ninth circle, the innermost pit of hell […] entombed in ice, are the traitors, the people whose actions have shattered the foundations of belief and loyalty. […] The ice, impenetrable and unyielding, mirrors the hardened hearts of the traitors, incapable of regret or redemption. In contrast to the fiery images of hell in the earlier circles, which [depicted] a passionate, albeit misguided, pursuit of earthly pleasures, the icy grip of the ninth circle speaks to a profound lack of feeling and corruption. The act of betrayal, as is depicted in the ninth circle, is deeply dehumanizing. Both the betrayer and the betrayed are diminished by the act. […] It’s the place where hope dies and solely despair remains. Betrayal is the final and worst sin.”

(0) Miscellaneous

Paramilitary parallels with DHS/ICE expansion. Unprecedented shakeup in China’s military leadership. Rising importance of ship tracking software. Declining age gaps in relationships. Latest updates on the Congo. Charitable donor advised funds. Ideas for investing in Guyana. The future of spreadsheets with Claude for Excel. The rise of Moltbook. The main character is Elon Musk. Yet another Olympics cheating scandal. Relatable name and relatable content. What secure attachment looks like. Honesty in relationships. FT recommendations for NYC date spots. Very good pun. Call center scams with multi-district litigation. ICE’s clever data collection methods. Financial derivatives and market indices for AI GPUs. Quirky defeasance clause in EA bonds. How to buy a law firm. How to fix your life in one day.

Home retailer rollups, the economics of haggling, children, prenups, Russian history, Standard Oil, Excel, WhatsApp, the Bathtub Trust, and more. Enjoy!

(1) CSC Generation is a Billion-Dollar Home Brand. Why Haven’t You Heard of It?

CSC Generation is a rollup of distressed home retailers. Their strategy (according to this article) is focused on making better use of customer data for marketing, shifting retailers to online sales, and generally improving performance of fallen angel brands. The article covers the story of two failed takeover attempts of public companies, and the broader challenge of the “change-averse old boy’s club culture […] that was a real obstacle to deals getting done.” CSC seems like they’re doing just fine, with over $1 billion in revenue in less than a decade. “I asked [the CEO] what his long-term goal was for CSC, and if he had a model for what he was trying to build. He did, and maybe unsurprisingly it was a company I’d never heard of: Constellation Software.”

(2) A Theory of Haggling

Haggling is a means of price discrimination. Some people are willing to trade some unit of time for some amount of cash savings. This is why haggling is more common in poor countries. The back and forth is both sides trying to estimate the surplus from the transaction and capturing more of that. “You’ll expect more haggling in a context where information is scarce, and one where the return on time spent haggling is high for both sides.” This also shows up indirectly in richer countries: we’re haggling every time we ask a Chipotle employee to add more chicken to our bowls. In B2B SaaS procurement, volume tiers, multiple meetings, and questions about use cases are sizing up a customer’s willingness to pay. Markets are automated haggling at scale.

(3) Making 13x on a Preferred Stock

In 2007, Goldman took private the hotel chain Equity Inns for $2 billion. Equity Inns had another $145 million of preferred stock trading in public markets, which Goldman didn’t acquire. Instead, they paused the preferred dividends, then stopped financial reporting with the SEC (as a private company), to tank the shares and buy them back cheaply. But Goldman forgot a technicality with SEC reporting requirements: any company with more than 300 “registered” shareholders (shares held directly in their name, not a broker’s “street name”) is required to make public filings of their financials. One shareholder bought many of the preferred shares at 8% of par, divided them into over 300 trust accounts, then informed Goldman of its legal filing requirements. That, plus some shareholder lawsuits over contractual breaches in the treatment of the preferred shareholders, resulted in a settlement buyout of the shares that netted a 13x return over this 6 year saga.

(4) Book Review: Family Unfriendly, by Timothy P. Carney

Why have parenting burdens grown, even as the average family size has declined? This book review traces the shift from quantity to “quality” in parenting: parents became chauffeurs, kids became résumé-building projects, and childhood transformed into a scheduled march towards college admission. Parenting has turned into a status competition based on visible investments (organized sports, academic camps, structured activities), with old-fashioned and unstructured parenting being seen as low status. It seems like the issue is particularly bad in the US, where a culture of extracurriculars, combined with car commuting between them, makes having children much more time-consuming than it used to be.

(5) How Did Russia Get So Big?

Russia was built with the commercial logic of a trading empire. This article shows how Moscow repeated the same playbook over centuries: secure the resource, secure the route, secure the port. First they eliminated their main competitor, Novgorod, to dominate northern trade routes from Arkhangelsk. The next step, conquering Kazan in 1552, gave Moscow control of the Volga River, the main trade route for Siberian fur. Traders followed rivers east in the search for sable furs, soldiers came with them, and forts would spring up along the way. This process repeated across the continent until Russia reached the Pacific. When they needed better ocean access, they took Vladivostok from China in 1860. The Trans-Siberian Railway later tied it all together.

(6) History of Standard Oil: The Cleveland Massacre

Standard Oil’s dominance started with the rapid consolidation of all of its competing refiners in Cleveland over a one-month period in early 1872. Standard Oil went from 10% of the market to 90% in one month. How did they acquire their 25 Cleveland competitors in one month? It wasn’t Rockefeller’s charm… it was an offer they couldn’t refuse. The creation of the South Improvement Company, a secret pact with the railroads over freight rebates, gave Standard Oil an immediate, crushing advantage on transportation costs that scared his competitors into either selling today, or going bankrupt tomorrow. In the end, the SIC was shut down by the Ohio legislature before it even started operations, but the threat was enough to consolidate the market at fire-sale prices.

(7) Why Millennials Love Prenups

Prenups aren’t just for billionaires. They’re especially popular with younger generations: 47% of Gen Z wants prenups. The growing popularity reflects, in part, the normalization of divorce. But just as important, courts are more willing to uphold prenups than in the past: the combination of rising divorce frequency and increasing financial independence for women has made judges less likely to toss out prenup contracts on “equitable” (fairness) grounds. The author debates whether she wants a prenup, but financials aside, she concludes: “The conversation alone is a personality test. Are you about to marry a person who wants to be reimbursed for the wallpaper you put in the nursery? [Or] does this person look at you and think: I want to give [you] everything?” Also, a potential quote of the year: “When Taylor Swift and Travis Kelce announced their engagement, one X user joked, “that prenup is about to be longer than any book travis kelce has ever read.”

(8) The Rise and Fall of the “Bathtub Trust” - A Deep Dive into the Sanitary Ware Cartel

The Bathtub Trust was the nickname for a short-lived Gilded Age industrial cartel of enameled iron bathroom fixtures (bathtubs, sinks, toilets, etc.). While less well-known than the great trusts like Standard Oil or US Steel, the Bathtub Trust is notable for a few reasons: (1) the invention of the Arrott Process for mechanized enameling illustrates how inventions that convert variable labor into fixed capital, with economies of scale, drive industrial consolidation; (2) the device of “neutral” third-party patent licensors was a novel workaround for antitrust law (although it didn’t work!), coordinating the behaviors of competitors using licensing terms instead of explicitly anti-competitive contracts; and (3) the use of “royalty and rebate” terms in the patent licenses, to be administered by the neutral intermediary, resulted in a self-enforcing contract that overcame the compliance issues of past trusts. Despite their creativity, the organizers forgot the government doesn’t appreciate creative workarounds of its policy priorities… “it looked and smelled like an old-fashioned price-fixing trust, and the Taft administration was in no mood to tolerate trusts.”

(9) Excel Turns 40: Why We Can’t Quit Microsoft’s Spreadsheet App

“Is Excel secretly the greatest piece of software ever written?” It has been 40 years since the debut (on Mac!) of Microsoft’s version of VisiCalc and Lotus 123. Spreadsheets were the original killer app for the personal computer: ”It transformed the Apple II almost instantly from a niche tool for gamers and teachers into the first mass-market PC.” While the narrative today is that Excel won via bundling with Microsoft Office, an underrated part of its early popularity was its calculation speed: “Excel first distinguished itself with speed[.] Intelligent recalculation […] used an algorithm to update only the cells affected by a given change.” There are 500 million paying users for Excel today, and its dominance is unchanged—the two main serious threats it has faced, cloud-based collaboration in Google Sheets and AI vibe-modeling, have had no visible impact on growth. The Office 365 bundle is still under $200 per year, so pricing has a long runway ahead.

(10) How WhatsApp Took Over the Global Conversation

“WhatsApp delivers a hundred billion messages a day, around the same number as there are stars in the Milky Way. The platform has more than three billion active users each month—pretty much the same number as Facebook, more than YouTube, and roughly twice as many as Apple’s iMessage. If you exclude China, where WhatsApp is banned, the app serves approximately half the human population older than fourteen. Three-quarters of WhatsApp users are on Android devices, with the rest on iPhones—which more or less reflects the global smartphone market. One reason that WhatsApp has traditionally lagged within the U.S. is because of the unusual dominance of iPhones. But that is changing. Last year, the U.S. was one of the app’s fastest-growing markets, with the number of monthly users exceeding 100 million for the first time.”

Miscellaneous

100x return investing in paint rollers. Thoughts on healthcare costs. Introducing agent engineering. Criteria for acquiring mobile apps. Good life advice. Capital markets in Uzbekistan. Reflections after 8 years as CTO. Issues at Novo Nordisk. Bankrupt rollup of port-o-potties. A history of NYC cemeteries. Oil bargain with Venezuela. Dutch ownership of Duivelsberg. Bull case for Plaid in payments. The future of consumer health apps. Profile of the founder of UNIQLO. Clever way to increase arable farmland. The origins of Coca-Cola bottling franchises.

Miracles, matriarchies, AI agents, Alexander the Great, Hermès, Epic Systems, cocaine kingpins, medieval latrines, and much more. Happy holidays everyone!

(1) The Fatima Sun Miracle: Much More Than You Wanted To Know

The “Miracle of the Sun” at Fátima is one of the most famous and inexplicable religious miracles of all time. In 1917, at the predicted time and location (based on three shepherd children’s visions of the Virgin Mary), over 50,000 people in Fátima (Portugal) witnessed the sun “dance” and exhibit other strange movements for 10 minutes. The miracle is unusual because (1) it was predicted in advance, (2) it was witnessed by thousands, including many religious skeptics, (3) there was no obvious incentive for fabrication, (4) there is still no good explanation for what they witnessed, (5) it occurred recently (in 1917) and was well-documented by contemporary sources, and (6) it has survived all attempts at debunking (including this article). “The Miracle has a reputation as the final boss of paranormal experiences, the ultimate challenge for would-be religious skeptics and debunkers.” Can you solve the 100+ year old mystery of what they saw in the Cova da Iria?

(2) Where are the Matriarchies?

Nearly all human and primate societies are patriarchal. Why? The consensus among sociologists is role specialization—males specialize in warfare and females specialize in caregiving. But why do gender specializations consistently produce male-dominated hierarchies? It is much deeper than physical strength. Dictators don’t gain power by physically overpowering their enemies in hand-to-hand combat, they gain power with superior coordination of resources, and the most important resource to manage is other people (armies, logistics, political parties, etc.). So the theorized causal chain is strength → warfare → management → control.

(3) Welcome to the Era of Experience

This AI paper is a few months old (ancient history for AI papers), but it makes a compelling case for the fact that the next phase of AI development will require agents that can interact with, and learn from, the real world as a means of collecting more data. The core limitation that constrains further AI development will soon become a lack of novel real-world data: “valuable new insights [like] new theorems, technologies, or scientific breakthroughs, lie beyond the current boundaries of human understanding and cannot be captured by existing human data.” New data will come from “allowing agents to learn continually from their own experience.” This paper covers some ways that agents will be designed to do this more effectively: (1) data streams that carry over from one conversation or experience to the next, so that the agents can continually learn and adapt over time; (2) action and observation in digital environments (MCP is already moving in this direction, although the streaming-focused learning is nascent); (3) integrations with more diverse “reward functions” like health metrics, profits, likes, energy consumption, etc.; and (4) more computationally efficient mechanisms for “thinking” than human language.

(4) Nobody Has A Personality Anymore

The downsides of therapy-speak, self-reflection, and psychoanalysis: “Therapy-speak has taken over our language. […] There are no feelings at all anymore; only [dys-regulated] nervous systems. […] This is the “healthy” way to think that previous generations were so cruelly deprived of. […] I find it strange that we think this is freeing. This brutal knowing. That this self-surveillance is the liberated way to live. [We] underestimate it, this miserable business of understanding ourselves. […] Explain anything long enough and you will find a pathology. We’re thinking about ourselves enough. After a lifetime spent trying to make sense of every experience, [this] generation might realize that the only problem they had, all along, was being human.”

(5) Book Review: Alexander to Actium

What happened to the Greek world in the 300 years between the conquests of Alexander and the fall of Ptolemaic Egypt to Octavian? This book review covers some of the history: “Golden age Greece was a Greece [of] individual city-states constantly in competition: Athens, Sparta, Thebes, Corinth, Argos, and so on. […] It was this quarrelsome, decentralized patchwork that produced the Greek miracle. […] With Alexander’s godlike achievements, though, that old Greek world crumbled. Suddenly, Greek civilization was dominated by massive kingdoms, ruled by kings whose only claim to legitimacy was that they or their fathers had won their lands at the tip of a spear. […] The kingdoms were essentially for-profit estates run for the benefit of mighty god-kings. Alexandria, the crown jewel of the Hellenistic world, home to the Great Lighthouse and Great Library, was effectively a giant consumer good for the Ptolemaic royal family, where all the surplus of Egypt’s countryside had to be shipped, on pain of death.”

(6) The Slow Death of Epic Systems

The headline is overblown, but indirectly explains how Epic could become healthcare’s IBM (in a bad way). Epic Health is the dominant operating system in EHR for the healthcare industry, but it exists more for maximizing billing than patient outcomes. As a result, they don’t use (or save) the torrent of data generated by healthcare (patient vitals, time-series data, etc). Because of AI, this “data exhaust” is now extremely valuable; it is arguably the last untapped frontier for real-world AI training data. The companies that excel at collecting and learning from that data to improve patient outcomes will supplant Epic as the critical layer of healthcare software. “Clinicians will live inside those tools, not inside an interface built for billing. […] There is precedent for this in every major industry: the core orchestration/data layer eventually recedes into the background while workflow and data intelligence move up the stack [while] the EHR becomes middleware.”

(7) Accommodation Nation: Elite Colleges Have an “Extra-Time-On-Tests” Problem

Students at elite schools abuse lax policies for “learning disability” accommodations. The main benefit is extra time for exams. The issue is most prevalent at elite colleges: “At Brown and Harvard, more than 20% of undergrads are registered as disabled. At Amherst, the figure is 34%. […] 38% of Stanford undergrads are registered as having a disability.” One enabling factor is that the definition of disability has widened to include basically anything: “Students with disabilities [aren’t] kids in wheelchairs, [they’re] rich kids getting extra time on tests.” The ADA hates this argument… “Having ADHD or anxiety might make it difficult to focus. But focusing is a skill that the educational system is designed to test.”

(8) The Cocaine Kingpin Living Large in Dubai

Daniel Kinahan, an Irish drug dealer, runs a billion-dollar cocaine empire from the UAE. This story covers how he entered the trade, scaling his operations through a money laundering partnership with Hezbollah, reinvested the proceeds into a boxing empire, and avoided arrest along the way. Dubai sounds like a comfortable place for criminals… “The family has made extensive use of the Dubai hawala network—an untraceable money-transfer system based on trust among individual brokers. […] Daniel and his brother have also bought various expensive properties in the U.A.E., and have incorporated several companies in Dubai. Financial oversight by Emirati authorities was evidently cursory: the brothers were deemed “low risk.”

(9) The Hermès Heist: How An Heir to the Luxury Dynasty was Swindled Out of $15B of Shares

Nicolas Puech inherited a 6% stake in Hermès, currently worth more than €13 billion. His stake was quietly liquidated and stolen over the course of two decades. His trusted Swiss banker, Eric Freymond put Puech’s bearer shares (!) in Swiss bank accounts and secretly sold millions of shares to LVMH for an attempted stealth takeover of Hermès a decade ago, while the remainder was sold over time and the proceeds diverted into accounts controlled by Freymond. Puech alleges fraud, is apparently penniless, and has sued Arnault, LVMH, and others. Freymond took the truth to the grave, and the location of Puech’s stock—impossible to track due to being bear shares—appears to be lost forever.

(10) What Actually Makes You Senior

“There is one core skill that separates senior engineers from everyone else: reducing ambiguity. […] The moment you hand them something fuzzy [is] when you see the difference. When senior engineers do it well, it looks easy. Like nothing was even done. The project just goes smoothly. Fewer surprises, production fires, or emergency meetings. But what actually happened was that someone did a lot of invisible work upfront. […] If you’re wondering whether you’re operating at a senior level, here’s a test: what happens when someone hands you something abstract/fuzzy/complex? Do you wait for someone else to clarify it for you? Do you start coding immediately and hope for the best? Or do you spend time up front making it concrete enough that you and your team can actually execute with confidence?”

(0) Miscellaneous

Winners of the robotaxi race. SaaS concept of Max MRR. Rage bait for the Pura Vida fans. Patrick Drahi takes off the gloves with creditors. Ode to the heart. The reasons that hangovers get worse with age. Economics of the mobile airport lounge. The pop fiction of Oliver Sacks. The history of Domino’s Pizza. Lessons from building Manus. Update on North Korea. The family of Florida sugar barons. A detailed history of the Delaware Chancery Court. Investing in niche sports. Medieval Erfurt latrine disaster.

Data moats, AI scaling laws, Bending Spoons, Berkshire Hathaway, Minnesota, ringtones, venture-backed bagels, alkaline hydrothermal vents, and more. Enjoy!

(1) Data and Defensibility

The classic half-baked software business model is “data → ?? → moat!” Data is neither necessary nor sufficient for a moat (a barrier to competition), but some data moats are real. The two broad groups are (1) data control and (2) data loops. Data control can be unique/proprietary (Nielsen); fragmented (CoStar); movement/networks (Visa); usage (Salesforce); IP/patents (S&P); and regulatory (FICO). Data loops can come from UGC (Facebook); SEO (Booking); gravity/stickiness (Toast); value exchange (Verisk); and trust/standards (Moody’s). There are more. But not all data moats are equal: the best are usage and trust, and the worst (which the author dismisses entirely) are learning loops: “Product optimization loop [moats]: lol, no. A/B testing is not even a shallow ditch.”

(2) The Bitter Lesson is Misunderstood

The “bitter lesson” refers to the thesis in AI scaling that model architecture doesn’t matter as much as throwing more compute and data at the problem. The massive AI capex boom in the past couple years is based on the first half of this lesson (scaling compute), but “the god of Compute we worship is actually yoked to an even greater one — the god of Data.” But getting new data may become prohibitively expensive: “we have basically ingested the entire Internet, and there is no second Internet. For language models, there’s no obvious way to scale the data.” Two possible workarounds: (1) architects who design clever ways to squeeze more intelligence out of each existing unit of data, and (2) alchemists who can conjure up new data. But data will soon be the hard constraint: “The AI companies that survive 2026 will be the ones who figured out data scarcity in 2025.” I would put it like this… the Copernican Revolution didn’t happen by thinking harder.

(3) Building Bending Spoons

Another recent interview with Luca Ferrari, the CEO of Italian app acquirer Bending Spoons. I was most surprised how, even more than acquiring companies, he focuses on acquiring talent: “for a long time on Slack, [his] label was recruiter.” The interview has a lot more on the hiring, compensation, and culture that seems to be his central focus. Luca also throws the gauntlet to would-be competitors: “[Even] trying to build Bending Spoons—if you understand what you’re doing, which is a prerequisite to even have a chance—[is] dauntingly painful. Many years from the ground up, cultivating the little garden. […] There’s no shortcut. We’re just not a model that, when people see it, they want to replicate. A lot of people have known about [us] for years. I’ve talked to investors, and I’ve seen nobody try, because they understand it’s just too painful.” :)

(4) Berkshire Hathaway: Letter from Warren Buffett (November 2025)

After more than 60 years, this could be Buffett’s last letter to shareholders. He’s 95 years old and reflecting on his long and lucky life: “Father Time […] finds me more interesting as I age. And he is undefeated; for him, everyone ends up on his score card as “wins.” [To] my surprise, I generally feel good. Though I move slowly and read with increasing difficulty, I am at the office five days a week where I work with wonderful people.” It’s a touching letter, filled with nostalgia over the many great business leaders and friends that he has worked with over the years, the charitable pursuits of his children, and his love of America (especially the Midwest!): “Remember to thank America for maximizing your opportunities.” The end of an era, the daedalic hall has fallen.

(5) “The Largest Funder of Al-Shabaab is the Minnesota Taxpayer”

Brutal headline. Minnesota is currently awash in welfare fraud with housing (HSS), school meals for children (Feeding Our Future), Medicaid-funded autism programs, and more. Much of the fraud is from groups in Minnesota’s Somali community (with evidence that at least some of that money has flowed back to the terrorist group Al-Shabaab). Of course, this has naturally turned into an issue of being “racist versus not racist,” but immigration politics aside, it does seem like a lot of welfare fraud is going on: “Autism claims to Medicaid in Minnesota skyrocketed from $3M in 2018 [to] $399M in 2023.” The Donald also weighed in… going to be a rough election cycle for the DFL.

(6) How the Custom Ringtone Industry Paved the Way for the App Store—Then Vanished

Remember ringtones? They used to be huge: “In 2002, 30% of all SMS traffic were requests for downloadable ringtones.” But the billion-dollar industry was silenced overnight: “From the 2007 peak of $1.1 billion in global ringtone sales, the industry shrunk by 97% over the [next] decade.” Some explanations for what happened: (1) the shift to texting; (2) fading novelty (caused by the Crazy Frog ringtones?); and (3) changing social norms with silenced phones (at least outside of the NYC subway). My own pet theory is more obscure: Apple made it difficult to install custom ringtones on iPhones. It’s much easier on Android, and Android is dominant with ringtones.

(7) Inside Cursor

Inside the corporate culture of AI startup Cursor: “My best evidence that the prize is the mission is that during my fall at Cursor, I overheard zero chatter from employees about getting rich. At Stripe and Figma (and most other startups), this was a favorite lunch table topic among the first few hundred employees at a deca-corn. Yet at Cursor, as the valuation goes up and up, I haven’t heard a peep about the second homes people will buy, the great-great-grandchildren that will be put through college, or the time they’ll take off traversing the world. If people have dollar signs in their eyes, they’re not talking about it much. And I think it’s because the thing most of them would do if they could retire tomorrow would be whatever they’re doing now at Cursor.”

(8) Ford and the Birth of the Model T

The early success of Ford was built on more than just the assembly line: (1) the use of very tight tolerance for machined parts (1/64 of an inch); (2) interchangeable parts; (3) casting the engine block as a single piece; (4) shipping cars as parts to assembly plants to reduce shipping costs; (5) removing color/style options to increase standardization (the Model T originally did have some options other than black); (6) creating special purpose machining tools for each part of the line; and (7) only a decade later, in 1913, the creation of the worker-level assembly line itself. Economies of scale led to lower prices, which led to more volume, which led to more economies of scale with process improvements: “By 1913, comparable cars to the Model T cost nearly twice as much as the Model T did. And the price continued to fall. By 1916, the cost had dropped to just $360—a two-thirds reduction in just six years.”

(9) PopUp Is Wall Street’s Bet a National Chain Can Actually Make Good Bagels

Seems like VC funds are getting into brick-and-mortar retail. Blank Street, 7th Street Burger… and now PopUp Bagels. PopUp Bagels is a bagel chain with 16 locations, mostly in NYC, that launched during COVID. Unlike mass-market models with burgers, chicken, pizza, and cookies, bagels are still mostly hyper-local (with the exception of a few chains like Einstein Bros). Their model seems to be keeping a light footprint to improve payback and make buildouts affordable for franchisees: “PopUps are barebones, averaging 700-1,200 square feet with limited seating, making [for] speedy takeout operations requiring only a few workers. Each [store] costs roughly $300,000 to $900,000 to open.” But the franchising route raises some questions about the store economics (why franchise unless the store-level ROIC is low)? The biggest question is probably the same one as everyone asked about Shake Shack years ago: will this work outside of NYC?

(10) On The Vital Question by Nick Lane

A book review of Nick Lane’s The Vital Question on the origins of life. Lane’s theory proposes two energy breakthroughs. First, life emerged from alkaline hydrothermal vents where alkaline fluid meeting acidic ocean created natural proton gradients. The mineral wall of the sea vents acted as the first cell membrane, and the natural proton gradient drove the chemical reactions. Since every cell today powers itself using this exact gradient process (chemiosmosis), life likely emerged where this energy could be efficiently taken from the environment. Second, eukaryotic complexity in particular arose when early cells swallowed a symbiotic bacteria known today as… mitochondria. Mitochondria are famously the powerhouse of the cell: they gave the parent cells hundreds of internal powerhouses instead of one external membrane-bound energy system like prokaryotes, dramatically expanding their energy budget, and this energy efficiency is why only eukaryotes evolved complexity. It’s definitely easier to read this review while using ChatGPT.

(0) Miscellaneous

Zohran reintroduces himself. Somalia’s Pirate Stock Exchange. Best spots to swim in the Bosphorus. Colombia’s golden age of cocaine. The bear case on the AI boom. Dan Gilbert’s revival of Detroit. Profile of Fabrice Grinda. Who is OpenAI’s auditor? Yet another deep dive on Bending Spoons. Taking the slow train across Mongolia. Using airport slots as loan collateral. Does the code matter? The company as a machine for doing stuff. Jeff Yass on prediction markets. Fighting hospital over-billing using AI. Teamshares is going public via SPAC. How Airbus took off.