I was cycling home from an event, which is how I move around these days. I listen to audiobooks or podcasts on the ride, and this time it was Range by David Epstein, which I had caught as my next read off a podcast.

In Range, David Epstein made a case that generalists do better than we have been told in a world built for specialists. In Chapter 7, David talks about the story of Frances Hesselbein. At 34 she was asked to take over Girl Scout Troop 17 in Johnstown, Pennsylvania. She said no 3 times. She had raised a son, knew nothing about running a troop of 30 girls and agreed only to a temporary 6 weeks. She stayed 8 years, until those girls finished high school. She never finished college. She ran the local council through the 1970s, and in July 1976 she walked into the New York headquarters as chief executive of the Girl Scouts of the USA, at 60 years old. She held the job for 13 years, tripled minority membership, added 130,000 volunteers and grew the cookie business to $300 million a year. When Business Week asked Peter Drucker who should run General Motors, he said he would pick Frances.

Her own account of it: “I was unaware that I was being prepared. I did not intend to become a leader, I just learned by doing what was needed at the time.”

I started crying at 20km/h, mid-turn under a bridge. A 30-year-old man in tears on a bicycle over a story about a Girl Scouts executive. Sometimes I feel I have wasted all these years building range, and I have been doubting myself for the longest time. So the tears were tears of vindication and awe. I believe a great work is in store for me too. Finally, a book that understands me.

The worst student in law school

I trained as a lawyer. I was reasonably good at it and yet I was probably one of the worst students in law school. In moot class, I remember I naturally reached for an analogy from somewhere else and used it to build an argument. But tutors wanted the authority. They wanted the case, the section, the paragraph of the textbook. Resourcefulness read as a shortcut. Common sense read as a failure to do the reading. And it happened many times during class.

This is not really a complaint about law school. Law school is the ultimate manifestation of a system that starts sorting people at 12. Streaming after primary school, subject combinations narrowing at 14, a junior college choice at 16 that closes most doors and a degree at 19 that closes the rest. We built the most efficient machine for manufacturing hyperspecialists that has ever existed. It is a large part of why this country works at all.

So I spent years trying to be what I was not and trying to rewire my brain into being a hyperspecialist. Then while I was in practice, I learned quickly and quickly got bored as well. In the end I left the profession because I couldn’t bear it anymore, without a job.

I took a huge bet on range over hyperspecialisation. People told me I had to specialise in this day and age, including high-flying people. It just did not sit well with me, and after much consideration I did what resonated: become a generalist and build a wide array of skills.

That bet is why I have been in multiple industries, learned multiple skills and perspectives and talked to people across domains. My ability to write and talk about so many things and connect the pieces across domains to build a large mental model comes from that range, and from having been in many rooms. Now that I am a founder, my ability to build digital and tech products is funnelled by my ability to create physical products, and my feel for creating experiences comes from my work in church.

Yet range does not appear to help on a CV. I have applied for regular jobs umpteen times and have not once landed an interview.

The world that paid for depth

For most of the last century, depth was the correct bet and everyone telling you to specialise was telling you the truth.

Adam Smith’s pin factory is the classic text of which all our division of labour philosophy relies. What a person does should be narrow so that they can excel at it, and the output per person goes up by orders of magnitude. The credential system was built on top of that mechanism and it worked. You spent 15 years becoming scarce at something, and the scarcity paid you a premium for the next 30 years.

The missing line item is what economists call “match quality”, the fit between a person and the work they end up doing. In Ofer Malamud’s study of the English and Scottish university systems. English and Welsh students had to specialise before they arrived. Scottish students spent 2 years sampling across fields before choosing. The English and Welsh graduates entered the workforce with more accumulated skill in their subject, which is exactly what the design intends. They were also far more likely to abandon their field entirely, having found out too late that it was not for them. The Scots started behind and closed the gap. The extra fit was worth more than the extra skill.

In Singapore, as a British colony, we are the English system with the dial turned up. We sort earlier, we sort harder and we attach more of a person’s life outcome to the sort. What Malamud measured is the cost of doing exactly that. We have a phenomenon who acquired a great deal of skill in something they turned out not to want, and who then discovered the exit was expensive yet they have too much to lose when they realised.

David also cites patent analysis by Andy Ouderkirk showing that the productivity of narrow specialists peaked around 1985 and has declined since. The specialists stayed the same while the problems got wickeder. A domain with stable rules and fast, accurate feedback rewards depth, which is why chess and golf produce prodigies. A domain where the rules keep moving and the feedback arrives late and distorted punishes the person who knows only one move very well.

Then AI arrived and took the clean bounded task, which is the thing a specialist is best at and a generalist is worst at. I made that argument in full a few months ago and I will not re-litigate it. What matters here is the direction of travel: the market is repricing breadth upward, and every institution that certifies people is still certifying depth. Universities, professional bodies, the whole ladder of postgraduate qualification. The signal has flipped.

The thing I was punished for is becoming the thing that pays. For me it felt like grief for the years spent believing that I was lost and wasted my time, while feeling vindicated for the path I took.

The part I still have to solve

Range is great. As a multidisciplinarian, I of course have no quarrel with the book. My problem is narrower and more personal: I need to commit and choose a lane. David actually says as much, because his claim rests on match quality, and match quality is sampling with the intention of committing. The sampling has a terminus. It ends when you find the thing and then you go all in on it, later than everyone else and with more conviction than any of them.

I have a history of not choosing. I jumped industries and jumped functions, and each time there was a defensible reason, and each time the reason was true. The trouble with being able to do many things reasonably well is that nothing ever forces the decision. You are never bad enough at anything to be pushed out and never deep enough in anything to be locked in, so the option stays open, and as Hormozi said, an option that stays open forever is in itself a decision you have made.

Nothing compounds if you keep resetting the clock. Relationships stay at the acquaintance stage because you left before the third year, and your name does not travel because nobody can describe what you do in a sentence. Whatever advantage breadth gives you gets spent, every time, on the cost of starting again. Yet being prepared and being lost look identical from the inside, and the only test is time. That is the unsettling part.

My answer, for now, is entrepreneurship. Owning the thing is the only arrangement I have found where the whole of me is usable and I am still committed to something specific, because the commitment attaches to the business rather than to a function or an industry. But it remains to be seen what I choose to do and what opportunities arise, which I suspect is a lifelong journey.

When none of it matters

There is a non-zero chance that in the next 20 years, all of the above is irrelevant.

If the(a) war extends and reaches the shipping lanes and the supply chain stops delivering to our island that grows nothing (Singapore imports over 90% of its food), the professional managerial class is worthless. I can structure a transaction, read a balance sheet, run a team and get a product out the door. None of that produces a calorie, a litre of clean water, a working pump or a set bone. The PMC layer exists because there is a functioning system underneath it to manage and advise and optimise. Remove the system and what is left is people with strong opinions about spreadsheets.

I wrote about this back in March, in a piece called Everyone Should Learn to fix an air-conditioner. My advice was to build a buffer, stay general, and get a side skill that requires a body. Plumbing, aircon servicing, live events, farming. I still think that is right. Practical skill from first principles and systems thinking across domains are the same capability, one level down. Both are the ability to hold a working model of how a thing actually functions rather than a memorised procedure for operating it. The person who understands why a compressor cycles can fix a unit he has never seen. The person who memorised the service manual cannot. That transfer is exactly what breadth buys you, and it is worth more when the manual is gone.

The Road Less Taken

A few days after that ride, I was talking to a senior from high school who did A-level music as well, now a very successful restaurateur in Singapore. It was well over 15 years since we were in school together, and of all the questions he could have asked, he asked whether I felt the music education we had actually helped us in life. We both said it did, even though nothing we do now has anything to do with the aesthetic philosophy and music history we had to go through. It gave us a really good overview of history and of life in its entirety to do what we do now. Neither of us could have said that at the time. The pattern took 15 years to show.

I have spent a lot of years assuming I was lost. I now think there is a reasonable chance I was being prepared. I have no way to tell the difference except to pick something and stay long enough to find out. And yet, poetic like the Robert Frost poem, I would choose this path every time - The Road Less Taken.

I rewatched Rocky 1 to 3 over the weekend. I did it because I needed a boost and wanted to be inspired again, and a boxing film that turns 50 this year seemed like the place to get it, having watched it when I was 15 (now I’m 30). It did what it had to do and more.

Some background: Stallone wrote Rocky as a broke actor and refused to sell the script unless the studio let him play the part. The film went on to win Best Picture in 1977, over Taxi Driver and All the President’s Men. It is a film about a nobody from Philadelphia who gets one shot at the heavyweight champion, knows he cannot win and decides the point is to still be standing at the final bell. He loses the fight. The film treats the loss as a victory, because it is one.

What I did not expect was how foreign the manhood on that screen would feel in 2026. The kind of man Rocky is has mostly disappeared, on film and off it. Only a sliver of semblance I see in patterns of some of my old friends that are shunned by modern society.

Rocky was one of my dad’s favourite movies, and watching it as an adult I could see his generation in it, including the parts that do not flatter them. Paulie communicates by getting drunk and swinging a baseball bat through his own living room. Rocky himself only manages to say “I’m afraid” out loud in the third film during a shouting match with Adrian, his wife, on a beach, after his trainer has died (no wonder the boomers love shouting so much). The bottling and the leaks of temper that run through those men ran through my dad too, and through most men of his generation. They could take a beating and could not process a feeling.

Yet the films carry so much else. The training alone in the cold at 4am. The refusal to stay down. The plain, awkward tenderness toward Adrian. Almost all of it has since been filed under the same label as the bottling: toxic masculinity. Which brings me to Elisha Long.

All quiet on the frontal lobe

Retardmaxxing, for those who do not live on the internet, is the practice of doing the thing without overthinking the thing. The word had floated around looksmaxxing forums since about 2018, and it became a movement in late 2025 when Elisha Long, a YouTuber who makes advice videos for men, posted one titled “Be retarded with women, coworkers, and jobs”. His starter pack, posted on X, reads like a parody of every self-improvement protocol ever sold: 1 cup of orange juice, do not have a single deep thought, do a 360 and moonwalk away from negative people, 3 dates, fire your therapists, 4 to 6 grams of taurine, quit your job, 1 cigar, do not get emotional, eat ice cream and blast bicep curls. The slogans are of the same genre. “You can’t overthink if you’re not thinking.” “All quiet on the frontal lobe.” By March this year Marc Andreessen was posting “Current status: Retardmaxxing” to 2.2 million views.

Long’s catchphrase is “Ride the Tiger”, and the flagship account posts lines like “Enter The Rapids. Ride The Tiger. Breathe Smoke. Chew Glass.” If “Ride the Tiger” calls something to mind, it should. Stallone could not get the rights to the Queen track he wanted for Rocky III, so he asked a band called Survivor to write him something, and Eye of the Tiger became Billboard No. 1 for 6 weeks in 1982. A meme movement about switching your brain off has reached back 44 years, on purpose or by instinct, to the most famous training montage in cinema. Either way the kids know exactly what they are pointing at.

Strip away the taurine and the cigars and there is a serious diagnosis of our generation of men. We have been paralysed into analysis, the paralysis gets celebrated as maturity, and the way out is volume of action rather than quality of plan. You do the thing, badly, in public, repeatedly, until you are good at it.

Every skill is a product of retardmaxxing

Every skill I have ever developed has been the output of retardmaxxing.

When I was younger I wanted to be good at street soccer, and I was laughed at for being bad. So I spent an entire holiday rolling a football around my house. No drills programme, no coach, no reason to believe it would work. I became one of the best players in school. I was terrible at basketball in high school. I spent recess and lunch between classes shooting, over and over in my school uniform and turned up to class late and sweaty without fail, and I ended up a pretty decent shooting guard.

I did music for A levels and I was not good enough on the piano. What I liked was singing, opera specifically, so I practised at almost every single moment I could, to the point that my sitting partner in class complained about the singing. I got an A to the point I stood a chance at making it a career. I got passionate about theology and philosophy and studied those texts over studying for my A levels, which was objectively a terrible trade and is also why I am adept at them now. When I started martial arts I trained and sparred every single day, and it earned me an opportunity at the national team for muay thai, which I didn’t take.

It followed me into working life. When I was given the reins of a small business, I had nothing but a law degree, which teaches you nothing about business, and I did not know what to do. So I spent every waking moment learning business, reading a business book every 2 or 3 days. I basically did my own MBA.

The common denominator was that there was no plan. No periodisation, no deliberate-practice framework, no video review, no consultation with anyone about whether this was the optimal use of my hours. It was messy and rugged and it looked stupid from the outside, to the point of retarded. The results were real anyway, and I could not deny it. Obsessive repetition with social cost attached is the whole engine. That messiness is what Rocky puts on screen for 3 films straight, it is what made Rocky types attractive then, and I would argue it is still the model of what is attractive in a man today. Nobody has ever fallen in love watching a man journal about whether he is ready to begin.

The men we are making now

Men of my dad’s generation were immature with their emotions, and the damage from that cannot be denied. The response we have produced is the complete antithesis: a generation of men who are emasculated. The fire is gone, and so is our narrative.

Men die by suicide at roughly twice the rate of women globally, and in much of the West the ratio runs closer to 4 to 1. Close male friendship has collapsed in the survey data, prime-age male labour-force participation has been falling for decades and young men report depression at rates no earlier cohort recorded. This is what losing the narrative looks like.

It shows up in the mirror too. Straight men are losing on the shallowest metric there is: our homosexual counterparts look better than we do, and I do not think grooming products explain it. The body is an output. It reflects what a man actually does all day. A body built through the messy and rugged pursuit of goals looks a particular way. Optimisation, macros tracked and recovery scores consulted, produces a different look, and a body built through nothing at all, which is the modal straight male body right now, looks like what it is. The gym bros get mocked for vanity, yet at least their bodies are evidence of pursuit. Most of us carry no evidence of anything.

Then there is therapy. We live in a generation that glorifies going to therapy, and before anyone reaches for the reply button: I have been to therapy myself and I thought it was going to fix me. I am not against therapy but the glorification of it. An over emphasis of self-care and overthinking. A man can narrate his attachment style and his “traumas” and still have nothing in his life he would fight for, and we have somehow decided the narration is the achievement.

I know you would think what I am suggesting is this: surely it is a balance to make.

The Lion and the Lamb

I no longer think balance is the right frame at all. A 50-50 compromise between strength and softness produces a man who is neither, and compromises of this kind are terrible. What I am describing is 100% of both at once. Full strength and full meekness in the same man, each one whole, with nothing traded away to make room for the other. The Lion and the Lamb.

Jordan Peterson has a frequently quoted and frequently hated line about this: “A harmless man is not a good man. A good man is a very, very dangerous man who has that under voluntary control.” People hate the quote because they hear an endorsement of danger, when the weight of the sentence sits in the last 4 words. A man who cannot cause hurt has not chosen peace. He is merely harmless, and harmlessness earns no moral credit, because his gentleness was never a decision. I vehemently agree with Peterson on this one. If you cannot cause hurt, your morality is weaker than you think it is.

In the Bible, Jesus is announced as the Lion of Judah and then appears as the Lamb 1 verse later, in the same chapter of Revelation. Whole strength and whole meekness in one person, and yet normalised Christians are so used to the text that we see no tension in it.

I think the attraction of women, which I believe is behavioural biology, is a good signal. Women hate crassness, yet the pull of a man who is strong, courageous and dangerous while being tender and vulnerable to the people he loves is hard to explain and very easy to observe. I believe it is why my wife liked me in the first place. Back then I was stubborn and rugged, held my views strongly and had no care in the world for whether they landed well. It is also the nature of women, once they love you, to want you safe and well and not sticking out too much. Adrian spends most of Rocky II asking Rocky not to fight again, and by every reasonable measure she is right, and the story only moves when she wakes from her coma and gives him the one word she had been withholding: “Win.” A wife’s job was never to keep the tiger caged. Her permission is not the point either; the fight was always his to take.

How I nerfed myself by getting civilised

I have gone through this full cycle, and I am not technically the manliest man. As illustrated above, my natural tendency is to retardmaxx. Yet over the past years I have been seriously nerfed, because I wanted to become palatable. I overthought and overoptimised instead of acting. I asked for permission, constantly, from people around me including my wife. Yet truth be told, no one had any authority to grant it and no stake in the outcome except me.

And sidetrack, these AI LLMs do not help (which is why I put up a post as a public service announcement for people to stop using LLMs for their life decisions). They are producing an epidemic of men, and women, asking a machine for permission, for optimisation, for clarity, and the machine hands back whatever the asker wants to hear. I spent months doing exactly this, so I can report the effect from the inside: it averages you out, and you go nowhere. The model is trained on everyone, so its counsel is the centre of the distribution, and if you keep taking that counsel you drift toward the centre too. Months of this and I found myself less of the man I was.

The whole cycle Rocky III covers in one film, I ran in slow motion over years. Comfort made me soft, and the softness cost me the thing that had produced the comfort. Mickey names it when Rocky asks why he never told him how beatable he had become: “Three years ago you was supernatural. You was hard and you was nasty, and you had this cast-iron jaw. But then, the worst thing happened to you that could happen to any fighter. You got civilized.” And on why Clubber Lang was going to win: “He’s a wrecking machine, and he’s hungry! Hell, you ain’t been hungry since you won that belt!”

Civilised, and no longer hungry.

He could have been describing most men I know, myself included.

Ride the Tiger Baby

Straight men have to claim ourselves back. To be clear about what that does not mean: I am not talking about claiming the world back from the feminists. The claim is on ourselves, because where the world is headed needs strong men, and what we are currently producing is emasculated boys with excellent vocabularies for their own feelings.

Let the effeminate men and our respected homosexual counterparts do what they do best, and I mean that with no insult at all. They are amazing at what they do. The bards, the creatives, the taste-makers of our age come disproportionately from their ranks, and the culture is richer for it. The division of labour is fine. The missing half is ours.

For the rest of us, the down and out men, the assignment is dumber than anything we have been told, which is exactly why it might work. Pick the thing. Risk everything. Stay dangerous, and keep the danger under voluntary control for the people who are yours. Fire the permission machine, or at least stop asking it whether you are allowed to want what you want. Professor Jiang is right in saying that the most cohesive and most energetic societies will prevail.

We need, just as I needed, to be reminded of a model of manhood that my dad’s generation carried badly and my generation threw out entirely, and a movement with a “slur” in its name is currently the only hope we have left. That should embarrass every institution that was supposed to be doing this job. It does not embarrass the retards.

They are busy riding the tiger.

Ray Dalio was back on Diary of a CEO, second time round, and the thesis had hardened since the last one. The UK has been in decline since the war. America still has the technology, and it also has the debt, the internal conflict and a rival that is no longer behind it. Then came the practical bit, which is the part that stayed with me. He quoted a Chinese saying he picked up in Hong Kong, one he has now used twice: that a smart rabbit has three holes. Keep the ability to move. Get out of the place that is deteriorating and into the place that is not. He named the places worth moving to, and Singapore was one of them.

I was proud for a moment.

Then the second thought arrived, which is that Dalio has a family office here and has had one for years. He is connected with the upper echelons of Singaporean society, the “invisible upper class”, and his understanding of Singapore is tainted by that lens. What he sees is the tax residency, the courts, the currency, the safety nets and the fact that nothing here has ever gone properly wrong. That is the same thing every ranking looks at, from roughly the same altitude, and none of them are looking at the ground - how life is actually lived and the people living it. The culture.

And since it was National Day just yesterday, I can’t help but notice what I see in myself, and in most Singaporeans I talk to - the exact thing Homer warned about - hubris.

How Great I Am

Hubris in Homer meant the specific act of treating another person as beneath you because your position lets you. The suitors eating through another man’s house for years while his son watched. Odysseus, who was safely out of the cave and already away, shouted his real name back at the blinded Polyphemus so the story would have the right author. That single sentence of self-aggrandisement costs him 10 years and all the men he sailed with.

The mechanism the Greeks described has 3 parts. Hubris-Atë-Nemesis, where Atë is the delusion/blindness that follows, going soft precisely because things are going well. Then Nemesis which is retribution that arrives last.

Every culture seems to have arrived here independently. I have been educated in a Chinese high school and there we often recite idioms. One of them is “骄兵必败” (jiāo bīng bì bài), an arrogant army is bound to lose. The story attached to the idiom is Wei Xiang, chancellor under Emperor Xuan of Han, talking the emperor out of a campaign against the Xiongnu around 65 BC, on the argument that a state leaning on its size and its numbers to show off in front of an enemy has an arrogant army, and an arrogant army gets destroyed. In the Levant, Proverbs 16:18 gives it in the compressed form. Aesop also gave the version I was taught as a child, the hare and the tortoise, which is a story about a competitor so far ahead that resting felt rational.

As Taleb argued in Skin in the Game: “If you hear advice from a grandmother or elders, odds are that it works 90 percent of the time. On the other hand... if you read anything by psychologists and behavioral scientists, odds are that it works at less than 10 percent, unless it has also been covered by the grandmother and the classics...”.

Taleb’s line is that your grandmother’s advice is worth more than the analyst’s, and the reason is the Lindy effect. These sayings did not survive because they are comforting. They survived because societies that ignored them stopped existing, and the ones that kept repeating them to their children had a marginally better hit rate. A proverb that has been in circulation for 2,500 years has been stress-tested against more real cases than any dataset any of us will ever build, even on a longitudinal scale (because no matter how long we test any dataset, they generally don’t survive dark ages).

A Fractal Analysis of Hubris

As with many phenomena, the pattern runs at every scale.

At the individual level it looks like the hare. Someone gets good, gets known for being good, and the reputation starts arriving faster than the work. There is a lag where the reputation is still roughly accurate. Then the lag grows, and the person stops doing the thing that earned it because the returns from maintaining the reputation are now higher than the returns from the work. Hubris is why many hyperspecialised PhDs rarely make any significant findings compared to cross-disciplinary peers.

At the company level it looks like Lululemon. The arc is textbook: a founder with an obsessive product view, real innovation, a decade of customers trusting the brand enough to pay $100 for leggings. Then scale, then a public market, then innovation delegated to people with no skin in the game. By late 2025 the CEO was saying out loud that the company had become “too predictable”, that it had not stocked the right sizes and had missed the seasonal colours. Retail analysts described a brand relying on existing franchises and looking replenishment-y. Customers on Reddit were posting about quality drops and I have also personally felt it. Soon, core black leggings showed up in outlets. Alo and Vuori took the customers who wanted newness, and Costco took the ones who worked out that the dupe was 80% of the product at 20% of the price.

Tanner Leatherstein has been doing the same demolition to luxury leather for years, cutting open handbags that sell for thousands and showing the machine-made, corrected-grain material inside. Those houses have kept getting better at branding while getting worse at the thing the branding was originally describing, and someone with a knife and a phone made the gap legible.

Ironically the only way to scale is through delegation, yet unfettered delegation is the mechanism for the downfall. Innovation moves to people with no skin in the game, and people with no skin in the game optimise for the metric they are measured on. The margin holds for a time after the substance goes, because trust is slow to build and slow to withdraw. There will be a tail of customers who purchase for the “brand” it has built. But once the brand is the sole reason for the purchase, it is ephemeral. It withdraws all at once.

Zoom out to the level of a country and the pattern is the same. Most of the writing on national decline puts it on the government. Rome had bad emperors for centuries before it fell, so bad leadership on its own is survivable. What is not survivable is a population that has stopped believing it needs to be good at anything, because the leadership is drawn from that population and answerable to it. A ruling class can only be as demanding as the people will tolerate. When the electorate wants the result without the cost, the leaders who promise the result without the cost are the ones who get selected, and the ones who tell the truth get replaced.

Then the state is being run by career government employees who answer to the metric but have no skin in the game. They see this work that helps the state continue running as an iron rice bowl, and take no pride nor ownership. And it is that individual fractal that affects the entire pattern of society as a whole.

A $1 million answer to a $20,000 question

The same shape shows up in the thing everyone assumes is unassailable.

Americans believe their military is the strongest in the world, and historically the belief has been correct. The Iran exchanges are showing something the budget line does not capture. Iran flies a Shahed drone that costs somewhere between $20,000 and $50,000 to build. The US shoots it down with an interceptor costing over $1 million. A former US Air Force officer put it plainly: the US military has a $1 million answer to a $20,000 question. Ukraine has been building interceptor drones at $1,000 to $2,000 a unit and producing thousands a month, which is a different answer to the same question from a country with a fraction of the budget.

Almost a trillion dollars a year buys enormous capability and it also buys a procurement culture that has never had to be cheap. Nobody in that system was rewarded for asking whether the exquisite solution was the right one, because for 30 years there was no adversary who could make the arithmetic matter. The brand of American military power is genuinely strong. The magazine depth is a separate question, and it is the one that decides how a long war goes.

People are sold on the brand right up until the moment the substance is tested. And once the leather and the stitching are all cut up, the entire facade is torn down all at once. That is why empires historically do not decline in a straight line. Usually, tension builds, and when the dam that holds it reaches its limit, one more straw breaks the camel’s back.

Singaporean Hubris

Singapore came first in the IMD World Competitiveness Ranking this year, up from second, first in business efficiency, fifth in infrastructure. That is the brand working exactly as designed. And our government is amazing at building a brand, armed with a functional bureaucracy at its finest.

Malaysia moved from 23rd to 15th in the same year. An 8-place jump is not noise. Singapore-based firms have committed over S$5.5 billion into Johor since the JS-SEZ memorandum was signed, and the projects are not marginal ones, with Kuehne+Nagel and ResMed among them. The structure is honest about what it is: high-value functions stay here, everything with a land or cost component goes north. Each of those decisions is a business re-pricing what Singapore is worth to it, without anyone having to announce that a re-pricing happened.

On the ground the arithmetic is uglier. From January to October 2025 there were 2,431 F&B closures, running at 307 a month against 254 a month in 2024, and more than 36% of the businesses that closed had been registered for under 3 years. Restaurateurs are moving operations to Johor Bahru because the numbers work there. Chinese F&B chains are arriving with capital structures local operators cannot match (and some of them don’t even bother having English on their signboards in a multicultural society). The headline sales figure still grows because the caterers and the fast food chains carry it, while the independent operator is the one being cleared out.

Of course, reputation is the product in a services economy, and always was. Nobody banks in Singapore because the buildings are nicer. They bank here because of what the name signals about enforcement, and that signal is a real asset built over 61 years that a competitor cannot buy or copy quickly. Trust compounds, so talking about the brand is maintenance of the actual asset rather than a substitute for one. Singapore has never had resources or a domestic market, so being a name people believe in has been the entire strategy from the start, and it worked well enough to produce one of the highest GDP per capita figures in the world.

Yet my argument is about the substance behind the brand rather than the brand itself. A reputation for enforcement is a claim about a future event, and it holds only as long as the people who would test it believe the underlying capability is still there. The claim and the capability can separate for years without anyone noticing, which is the whole Lululemon arc.

Singapore used to be the place with genuinely high quality at an arbitrage. Our USPs used to be location, a high quality of competence and low corruption, all at a fraction of the cost. That was the actual product. Now, the arbitrage is gone, and quite a lot of the quality difference has closed too, because our neighbours spent 20 years building the thing we told ourselves they could not build. Location matters less with more globalisation and when the cost calculus levels out. What we are left with is the low corruption, which is a small part of the USPs that we used to have. Every second announcement now leans on the Singapore brand, the Singapore standard, the trust premium. Talking about the brand is what you do when the product is no longer doing the talking for you.

The end state is not invasion. Somebody opens a jurisdiction that does 80% of what we do at 40% of the cost, with a government hungry enough to actually answer emails, and the marginal business goes there. Then the next one. The brand holds for a while after the substance goes, the way it always does, and then it does not. And the only fix is what I believe the government is trying to do, which is also being met with strong opposition on the ground. Cliché as it is, AI seems to be our only desperate answer.

To continue to be the service economy of the region and to push our USPs back up to something commensurate with our brand, we have to make every working Singaporean leaps and bounds more efficient, effective and high quality compared to our neighbouring peers. And the only way to do that seems, at this moment, to be a workforce that is highly equipped and adept at AI.

Yet it is no longer the 1970s. Our surrounding low cost counterparts have leapfrogged us on internet infrastructure. They have land for data centres where we have almost none, populations still willing to do the manual and blue-collar labour our white-collar hubris now refuses, and the resources to generate power, which is scarce and practically non-existent in Singapore.

The fifth time

This is roughly the 5th time I have written some version of this warning. It is trite, yet the marketing people claim you need to say a thing 7 times before anyone registers it, which gives me 2 more, though I suspect the number is higher for anything people would rather not hear.

The rabbit with 3 holes is good advice for an individual. It is what you do when you have accepted that a place can decline and you have decided your exposure to that is a portfolio question. Most of us do not have 3 holes. We have this one, and a passport that is only valuable because of what this one has been.

Dalio is right that people move to the better place. What he does not have to think about is what happens to the people who stay when the better place stops being here. I don’t wave the flag, or dress in red and have Singapore flag stickers on my cheeks singing Home. I am not performative in my patriotism but I write this because I grew up here and I am genuine in my passion for my homeland and what we built here.

I came out of an IMAX screening of The Odyssey recently, still thinking about what it meant for us today.

Nolan shot the whole film on IMAX 70mm, the first narrative feature ever made that way, at around 250 million US dollars. The cameras cost 500,000 dollars each and hold a few minutes of film before a reload. Fewer than 40 screens on earth can project it as shot. With that money he could have filmed anything. Yet he picked The Odyssey, and his telling of it is heavily skewed towards Zeus’ law.

The law of Zeus

The Greeks called it xenia. Zeus in his role as protector of strangers was Zeus Xenios. You take the stranger in and feed him before you ask his name, because he might be a god and because one day you will be at somebody else’s door. Nolan says what drew him to it was the idea that human interdependence is “the foundation and possibly the definition of civilization”.

Unlike what the movie portrays, where the breach of xenia at Troy sets everything off, historians do not pin the end of the Bronze Age on Troy. No single cause did it: drought, earthquakes, famine, migration, revolt and the cutting of the trade routes all hit a system too interdependent to absorb any of them. Copper came from one place and tin from another through a supply chain spanning 8 cultures, held together by understandings about how kings and merchants and strangers dealt with each other. Those understandings failed, so the network failed, and the small nodes went first because they had nothing else underneath them. What followed was a 300-year dark age.

It seems he made The Odyssey now because the guarantees that let strangers deal with each other across distances are being broken in public, and most of the world seems oblivious to what follows.

The obvious referent is Donald Trump. Nolan keeps his own politics opaque, yet this film seems to me a criticism of how the world order has changed since Trump’s term started. Zeus’ law reads as an analogy for the rules-based order, which by many accounts is eroding in today’s world.

Our version of Zeus’ law

Bretton Woods, the dollar, the UN, sea lanes patrolled by somebody else’s navy, the convention that borders do not move by force and a small state’s paper is as good as a big one’s. This was the world order the victors of WWII built in 1945. It has held the world in relative peace since then, and built up many economies and powers from very little, including China, Japan, the GCC and even Singapore. And now the cracks are showing.

Our own government says it is over. Vivian Balakrishnan told the UN General Assembly in September 2025 that the post-1945 order has come to an end, that respect for the Charter has eroded and that violations of sovereignty have become ordinary.

The evidence is already in. Tariffs are a routine instrument between the largest economies. Territory in Europe has been taken by force and held. The dispute settlement machinery of the trading system has been left to stop working. Each of those is what the order existed to prevent.

Lee Kuan Yew’s line about elephants gets quoted every time this comes up, usually by people who think he wrote it. He borrowed it from Julius Nyerere and added the part that stings, that the grass suffers just as badly when the elephants are getting along. LKY was talking about what great-power manoeuvring meant for a country sitting on the crossing of 2 oceans with nothing to sell but its usefulness.

We were a mirage

There is no natural reason for this country to exist. We, like some societies, are figments of the global order.

Siam kept its independence through the colonial century because Britain and France each preferred a buffer to a shared border. The Gulf sheikhdoms were British-protected until 1971, and then oil made them worth protecting by everyone. Venice ran for over a thousand years as a trading hub, only to be swallowed whole by Napoleon.

We are in that category. A trading port with a state built around it, holding a crossing that other people’s ships need, with no water, food or hinterland of its own. We exist because an open sea-lane order made us useful and because the men running the place made us more useful than the alternatives. Both are conditions.

Usefulness is the entire asset and we have never set its price. The Strait matters because ships move through it under rules somebody else enforces. Capital sits here because it can travel and wants a clean place to land. The factories and refineries are here because a supply chain runs through this island rather than somewhere cheaper. Every one of those is decided outside this country, reviewed continuously, by people who owe us nothing.

Here is how I think about it. Imagine Taylor Swift books the National Stadium for a full year. Every night for 12 months, tens of thousands of people pour through Kallang. If you live nearby you would be mad not to open a stall. Sell food, sell merch, rent out your spare room. Within months the whole neighbourhood is built around the concert and everybody is doing well.

We are the shop that saw the opportunity first. We have been selling the same plate of food for years, and because the crowd is captive we now charge 3 times what we used to. Meanwhile the food trucks have arrived. They are a fifth of the price, they open longer hours and their quality is also improving. Our answer is that our plate is better, that we have been here longest and that people will pay for quality. That could be true but is also what a business says while it is being undercut. Hubris.

Then one day the residency ends and Taylor Swift leaves. She is under no obligation to stay and the decision was never ours. The crowd stops coming and nobody in the neighbourhood can afford and a business built entirely on somebody else’s tour schedule. That is Venice. The global order is the residency. We are the shop.

What actually carried us

Plenty of places copied the model. What got us here was a group of men who could think 20 years out, who were ruthless where it counted and who did not use the position to get rich. Machiavellian in method and clean in the pocket is a rare combination. Under them sat a public service built to keep doing it after they were gone.

LKY said in 2010 that his regret was there being “such a narrow base to build this enormous edifice”, and that the next generation should not take it for granted. He was describing a design flaw he could not fix.

There is a story PM Lee told at the 2015 National Day Rally. Two years before that, an old and frail LKY sat listening to a heated parliamentary debate on bread-and-butter issues and was troubled enough to draft a speech of his own, to remind the House of the hard truths about survival. PM Lee talked him out of delivering it and told him he had done his part over the decades, that it was their job now. That was 2013. The debate he was watching is the one we are still having.

My doubt is about the last 15 years.

Look at what a Budget is for now. Budget 2026 gave every household 500 dollars in CDC vouchers, 200 to 400 dollars as a cost-of-living payment and up to 570 dollars in utility rebates. There were vouchers the year before under the SG60 name and vouchers the year before that. None of it is wrong and against what we hold none of it is expensive. What it does is set the terms. The Budget becomes a distribution event, the coverage becomes a tally of what each household receives, and the hard road gets a paragraph near the end. A founding-era Budget asked people to give something up now for a return in 20 years. Ours asks for nothing and arrives with a payment attached - a handout.

There are people inside the government working the hard problems and some of them are very good. They are also few, they do not sit on the levers that decide things, and many are jaded by the service. None of it adds up to collective action while the incentive runs the other way, towards protecting the iron rice bowl.

The growing opposition offers no way out either. The Workers’ Party held its 10 seats in 2025 and is the only credible alternative on the board, and the direction of its argument is more spending, more drawing on the reserves and more protection for workers. Reasonable asks from where voters sit. In 61 years no opposition here has run a ministry, so nobody has had to show they can govern a country that stays alive by being useful to people who owe it nothing. It seems the only effect of the opposition, and of the growing support for it, is to sow the kind of division we see in the West, which may become a problem for the collective action required in the years to come.

The founders were Machiavellian and they were also honest with people about the bill. That has gone. One side offers comfort it can afford today, the other offers comfort it has never had to price.

The uncomfortable part is that voters ratified all of it. The PAP took 87 of 97 seats in 2025 on a higher share of the vote than the election before. The vouchers are not a failure of nerve, they are the correct answer to the incentive, and any politician who stops handing them out loses to one who does not. Every rich democracy is running the same pattern. Which is why better politicians on their own will not turn this around. They would face the same electorate and reach the same conclusion.

I am writing this from the comfortable end of everything the founders built. I have never stood for anything or served in government. Doubting the leadership from this chair is cheap and I am doing it anyway. And personally, if the call is undeniable, I am willing to stand for Singapore.

The counterargument

The argument against me is that it has worked for 61 years. CPF, the reserves, the sovereign funds, national service, a civil service paid enough that buying it is expensive. Every one of those was designed so the country would never again need a Lee Kuan Yew, and they have held through 3 decades without one.

They were also built on an assumption: that the society they were built for would stay roughly the same society. It is not staying the same as capitalism matures and immigration increases.

The resident total fertility rate for 2025 came in at 0.87, down from 0.97. Around 27,500 resident births, the lowest ever recorded here. The deputy prime minister told Parliament in February that on that trajectory the citizen population starts shrinking in the early 2040s, and that 100 residents today produce 44 children and 19 grandchildren. The plan is 25,000 to 30,000 new citizens a year for the next 5 years. The old-age support ratio is down to 2.4 working-age citizens per senior.

What matters is what a shrinking population does to consumption. Household consumption is about 67% of GDP across the world. Here it is 31%, down from 90% in 1961, because our economy runs on exports, finance and other people’s demand rather than our own. So a smaller population will not move the headline number much, and that is the problem. GDP will keep reading fine while the part of the economy Singaporeans actually own and work in gets thinner: the retail, the F&B, the services, the SMEs that sell to residents and nobody else. Rents are set by what global capital can pay, not by what a local business earning off a shrinking base can pay. Those businesses get squeezed from both ends, and the headline GDP number does not reflect the hardship felt on the ground, so it does not spur the government to say or do anything (not that there is much they can do about it, to be honest).

The same shrinking cohort is where the institutions recruit, on the assumption that the best and brightest from the education system would run the public service best. A public service that needs the top slice of each year is drawing from a smaller year every year, and that top slice is globally mobile in a way their grandparents were not. They are also getting further removed from the realities of the working and middle class, and more complacent with it. The elitist model of governance here promotes meritocracy and builds it into the system. Yet meritocracy run for 61 years, alongside maturing capitalism, rewards ability together with the background and finances to buy a child the best resources, education and connections. This goes back to my earlier point: we have never run these systems on a longitudinal scale, and the cracks are beginning to show. What we are getting is an elite class that is smaller, more entrenched and more entitled. People in that class are more prone to acting from a desire for power and money and their own interest, and more willing to run populist policies they do not believe in, because power to them is a game to be won rather than a duty of honour.

The one people who are undying

Ibn Khaldun had a word for what we are missing. Asabiyyah, the group solidarity that binds people tightly enough to act as one. He wrote in 1377 that it is forged in hardship and dissolved by comfort, and that it runs down over 3 or 4 generations, roughly 120 years, at which point the dynasty falls to whoever still has cohesion. His argument was that the decline is structural rather than moral. A people does not wake up corrupt. The conditions that produced their discipline and their solidarity stop existing once the success arrives.

He sketched the generations too. The founder builds through his own toil and knows exactly what it cost. His son watched him do it and keeps things going by imitation. By the grandson there is no memory of the cost, only the result. The fourth generation spends it.

We are in the third. LKY’s men were the first, the ministers who governed beside them were the second, and the people running the country now grew up inside the result. To keep it in perspective, Khaldun’s clock runs out at about 120 years. Another 100 would take us to 161.

The only group I know of that beat that clock is the Jews. They lasted millennia without a state to protect them, on portable things: law, text, stories drilled into every generation and an obligation to each other that held across borders. The price was a history of pain nobody would choose, and that history is why the culture held.

Ours is the opposite inheritance: 60 years of everything working, of water arriving and trains running and reserves compounding. No wars, no famine, no displacement, no oppression. Nothing in that prepares a population for a decade of things not working, and whether a people will absorb that decade together is what decides whether a country lasts another 100 years. We have never been tested.

Would we last another 100 years?

A close friend asked whether, in my opinion, Singapore would last another 100 years. LKY was asked a version of the same question and answered it in Hard Truths in 2011. Projecting the country forward 100 years, he said he “cannot tell you that it will exist”. He said that having built the place himself, at the height of what he built.

My answer is a “no” based on the current settings. The order that made us useful is going. The population that staffs the institutions is shrinking. A political tradition that produced men who thought in decades now manages 5-year cycles well. Those 3 lines cross long before the 100 years are up, and most of the crossing happens inside our own lifetimes.

Singapore will probably still be on a map. What I doubt is the stability of it: rich, safe, sovereign in practice rather than on paper, able to say no to countries 50 times our size. Small states rarely fail by invasion. They fail through a run of decisions taken elsewhere that they are no longer in a position to refuse, and by the time refusing matters the capacity is gone.

What would change my answer is a leader who can ask a comfortable people to be hard again and be followed. That has happened here before, under conditions nobody would volunteer for.

The playbook is straightforward. Keep the state solvent and armed. Move what we sell the great powers up the stack, because a port is replaceable and a place that makes what nobody else can make is not. Bind new citizens in properly rather than counting them. All of which we are doing to the best of our ability. But the most critical of them all is to build the tolerance for hardship on purpose, in schools and families and national service, and with it societal cohesion, before the hardship arrives and builds it for us. This, we are lacking.

I want to be wrong. I have built my life here, my businesses are here and everything I care about sits on this island. Having seen a lot of the world, I still want to build my family here. Our modern Zeus’ law is being broken in public, and I hope the breach does not end us the way it ended the Bronze Age. That is a hope that I still cling on to. Happy 61st nonetheless, my motherland.

A friend of mine is an engineer at one of the big tech companies. And like most people in corporate almost everyone he spends time with is an engineer too, either at the same tier of big tech or at a startup. I am aware I do not have that circle. Most of what I assume the consensus on AI is comes from a small and skewed sample, mostly lead engineers, founders and operators like me. I asked him a question: Is big tech excited about AI internally, are the engineers themselves excited about it. I expected a yes like most start up founders have told me. The people building the thing should be the most bullish on it.

He said the engineers are disillusioned. They use AI every day and a lot of them are worn down by it.

So I pushed. Is it not amazing to be that much more productive? He said engineers generally treat their work as a craft. They are used to sitting with one codebase for hours, holding the whole thing in their head, going deep. The way AI makes them work now forces them to keep switching their brain between different tasks and different kinds of tasks, and that is exhausting for craftsmen who only want to build. It dawned on me that it seems like the average person is not great at context switching, and yet it is the essential skill that an entrepreneur has to have.

Craft, interrupted

Writing software the old way was a long stretch inside one context. You loaded the architecture of a system into your working memory - the function names, the edge cases, the reasons the code is shaped the strange way it is - and you stayed there. The value came from depth. An engineer who can hold a large system in his head and move through it without breaking focus is the one who ships the hard things nobody else can.

Yet AI changes the shape of that work. You describe what you want, wait, read what the model produced, decide whether it is right, correct it, then move to the next thing. The model is fast, so the loop is fast, and the fast loop pulls you out of depth and into a mode where you are supervising and reviewing rather than building. The job drifts from author to editor. Reviewing code the model wrote is its own tax, because you have to load the machine’s intent instead of your own, and the machine’s intent is often directionally right and wrong in the details.

The engineers are not rejecting the tool. They reach for it, they defend it and they are still worn down by it. The disillusionment sits somewhere other than output quality. The code is fine, sometimes better. What changed is the shape of the day, from long stretches of depth to a fast loop of switching between prompting, waiting and correcting.

The switch is the job now

Every time you move from one task to a different kind of task, your brain does not cut over cleanly. Part of it stays loaded on the thing you just left, and the residue shows up as a lag and a few extra errors right after the switch. This has been measured for decades. The more unlike the tasks are, the more it costs to cross between them.

Coding with AI multiplies the crossings. You are no longer in one context for 2 hours. Instead you move in and out of dozens of small contexts an hour, each with its own reload. For someone whose whole professional identity was built on depth, that is degrading in the literal sense, because the thing they were best at, staying down in one place, is the thing the new workflow stops rewarding.

This is where I think the engineers are half right. They have correctly noticed that the work now demands a skill different from the one they spent 15 years building, and that the new skill is tiring. Their instinct, to resent it and hope it passes, is the part I would push back on.

Entrepreneurship is Context Switching

I have already made this exact transition, from the other direction.

I trained as a lawyer. My years in practice were spent on M&A and equity capital markets, which meant hours inside a single document. You draft a transaction agreement, you sit with one clause, you hold the whole deal structure in your head, and you go deep. The work rewarded exactly what my friend’s engineers are now mourning. Depth, focus, one context held for a long time.

Then I started managing people and running businesses, and the shape of my day inverted. A normal hour now might contain a slide I am designing, a workout, a sales call, a fire that has just started in one of the companies, then a document, then something visual again. The businesses sit in different domains, so these tasks are not near neighbours, and the further apart the tasks are, the more the crossing between them costs.

At the start I could not do it. Moving from drafting to a sales call to design work inside a short window left me bad at all of them, and it drained me completely. What got me through was practice and having no way out. The brain rewires with repetition, and mine did. This is neuroplasticity doing ordinary work, the same way a new physical movement becomes automatic once you have done it enough. I can now hold many open loops at once and move between them, and most days it feels normal.

It did not come free. In that first stretch I was carrying more stress than I could actually handle. My sleep broke. I had to change the way my mind worked, and the change taxed the rest of me while it was happening.

There is a 2008 study led by Gloria Mark with a title that says most of it, “The Cost of Interrupted Work: More Speed and Stress”. They interrupted people doing office tasks and found something counterintuitive. The interrupted group finished faster, with no drop in quality. People compensate for interruptions by working faster. The cost surfaces somewhere else, as more stress, more frustration, more time pressure and more effort. Same output, higher internal load.

That is me, measured. I adapted to the switching. I even got faster. And honestly I enjoy the variety. But the bill lands on my body. I wear a WHOOP, and my stress readings sit chronically high, high enough that I am fairly sure the cause is this, the constant crossing between contexts from morning to night.

The steelman

Before I land on push through, let me give the other side its strongest form.

Maybe the disillusioned engineers who refuse are the ones who end up fine. There is a real case that the hardest problems will always need someone who can hold a whole system in his head, and that true depth becomes rarer, and therefore more valuable, as everyone else drifts into the shallow fast loop. On that view the engineer who protects his depth and turns down the switching-heavy work is making a smart bet on scarcity rather than sulking. I have some sympathy for it. There will be a premium on real depth for a long time.

The problem is that the premium sits at the very top and thins fast below it. For the small number of engineers working on the genuinely hard problems, depth stays king. For most working engineers, the job on offer is the supervising-and-switching job, and refusing it means refusing the job. You do not get to opt your role out of the acceleration any more than a company gets to opt out of the capital markets pushing it. The speed is captured by the org. The switching cost is absorbed by the person. That split is the whole reason the individual cannot decline.

The head start

So I hold both things at once. The engineers are right that context switching is a heavy and real cost, and that the productivity story oversells the lived experience of the work. I still think the move is to learn it anyway.

The reason sits in the structure. AI is not a tool you can wait out, and it is not going back in the box. The workflow it creates, fast loops, constant switching, more supervising and less deep building, is the direction of most knowledge work now, not only engineering. My lawyer friends are heading there. Anyone running anything already lives there. The old world of single-context craft was the default condition of a slower economy, the same way work-life balance was, and that economy is gone for most of us.

Which means the people who can context switch without it breaking them get a head start. Switching is the environment now, and the ones who adapt to it early compound an advantage while everyone else spends their energy resisting it. That is a claim about the world these engineers work in, and it says nothing about whether the switching is good for them. It plainly is not.

The playbook, as far as I can tell, is simple to state and hard to live. Build the switching capacity on purpose, the way you would train any skill, and expect the early period to feel awful, because that is the rewiring happening. Batch like with like where you can, so you cross between different kinds of tasks less often, because the crossing is where the cost is and not the work itself. Treat the stress cost as real and pay down what you can: guard your sleep and your recovery, and keep some protected block of the week as deep single-context work so the muscle for depth does not waste away. Take the trade with your eyes open, knowing what it costs and what it buys.

What I do not know is the long-run price. I have rewired for this cadence and I can run it, but the high number on my strap most days is a fair warning that the human nervous system was tuned for a slower world than the one AI is building. The head start is real. The bill for it is still coming in.

Earlier this year I wrote that the real bottleneck of storytelling has always been access, the distance between an idea in your head and your ability to communicate it. The cavemen painted on walls because that was the medium they had, the orators of Rome dominated the forum because they could command it, and the printing press handed power to whoever could afford to run one. Every new medium widened the audience and installed a new set of gatekeepers, and the people who won the marketplace of ideas were rarely the ones with the best ideas. They were the ones who could package and distribute them. That piece, the LFG Creator Manifesto, was about closing the gap for the things a person makes: stories, films, podcasts. This one is about the gap underneath it, between who a person actually is and what any of those outputs manage to carry of them.

That second gap stays open for 2 reasons. The worldview you are trying to broadcast does not sit still, and you cannot get enough of it out anyway.

A mind is supposed to move

A worldview is not a fixed object you can photograph once and publish. It moves, and it moves because the brain underneath it physically changes.

The cleanest demonstration is Eleanor Maguire’s work on London taxi drivers. To earn the licence, London cabbies spend years memorising the Knowledge, the layout of around 25,000 streets. When Maguire scanned them, the posterior hippocampus, the part of the brain used for spatial memory, was measurably larger than in people who did not drive cabs, and it kept growing with years on the job. The adult brain rewires itself in response to what you make it do. It is plastic well into middle age, not set in youth and fixed thereafter.

I would push the point past biology. Intellect is more than how much you know or how fast you reason. A large part of it is the willingness to change your mind when the evidence moves. The person who cannot update is stuck, whatever they prefer to call it, and a worldview that has stopped moving is usually one that stopped paying attention. So any honest attempt to embody a person has to embody a moving thing. Freeze it and what you have saved is no longer the person, it is a snapshot they have probably already grown out of. This is what most public personas get wrong: they ship a fixed version of the person because fixed is easier to perform.

Two problems stand in the way

The task has two halves and they fail for different reasons.

Capturing the data, truthfully

The first problem is input. A worldview changes through the ordinary traffic of a life: the conversations you have, the arguments you lose, the book you half finish, the video you watch at 1am, the offhand thing a stranger says that you cannot stop turning over. Almost none of it is written down. The two best-known answers both try to fix that. Gary Vaynerchuk says document, don’t create: record the process instead of agonising over the perfect post, and you leave a trail. Tiago Forte’s Building a Second Brain is the systematic version, an external store you capture, organise and distill your notes and highlights into, on the principle that the mind is for having ideas rather than holding them.

Two things break both of them. The first is that documentation is not belief. You read a hundred things a week and disagree with most of them, you sit in meetings nodding at views you think are wrong, you say things socially that you do not hold just to keep the room comfortable. A raw feed of everything you encountered would be a feed of mostly other people’s positions, and a model trained naively on it would learn the noise rather than you. A second brain has the same gap at one remove, since it stores what you thought worth keeping but not what you concluded about it. The capture has to be filtered through what you actually believe, which is the exact part that lives only in your head and never makes it into the archive.

The second is cost. The true limit of documentation is a camera and a microphone running 24 hours a day, catching every aside and every argument you would never publish. Right now that is unworkable for a normal person, because of privacy law, because the files are enormous and because transcribing and making sense of that volume still costs real money. So the engineering problem is this: how do you capture as many data points as possible, as true to the person as possible, without strapping a recorder to everyone.

Communicating it, everywhere, at once

The second problem is output, and it mirrors the first. Suppose you did know exactly what you believed, cleanly and currently. You still cannot be in enough places to express it. To fully communicate a worldview you would have to be on every platform, in every reply, in every conversation, all the time, and saying the same true thing in each. Nobody has that kind of bandwidth. What actually gets out is a thin, lagging slice of what you think, and people build the whole person out of that slice.

The goal is a twin that moves with you

The digital twin is no longer a novel idea. Plenty of people now build an AI version of themselves, and there are products that will clone you from your content and sell it back to you. I had been turning this over well before it went mainstream, which matters only because the version most people ship still gets the hard part wrong.

Put the two halves together and the thing you want is a digital twin. I do not mean an avatar that looks like you or a chatbot that has scraped your posts. I mean a model that holds the data points of how you actually think, tracks how that thinking shifts over time, and extrapolates from what it has to the closest possible guess at what you would think about something it has not seen yet.

The twin answers both problems at once. It is the filter that turns raw documentation into belief, because you correct it when it gets you wrong, and a single correction is worth more than a hundred passive data points. It is also the bandwidth, because once it holds you well enough it can stand in the places you cannot be. The microphone was the brute-force version of capture. The twin is the practical one, since it does not need to record everything, it only needs you to tell it when it has diverged from what you currently believe.

A personal operating system

Take the trajectory out a few years and the twin becomes something closer to a personal operating system. Andrej Karpathy has argued that the large language model is the kernel of a new kind of computer, with the context window as its memory, tools and files and other apps attached as peripherals, and plain language in place of code. The personal version is that same stack booted on one person. The model is general, the file system is you: your positions, your history, the record of how you think, sitting where the kernel can read it. It runs in the background and can act on your behalf, creating in your voice, whether that is writing, an image, a piece of music, an idea worked up into a business, a project specced the way you would have specced it. On the output side, it could answer messages as you, in the way you actually mean them, so that being in 10 places at once stops being a physical impossibility.

There is an inward use too. Everything you have read and watched and concluded becomes a store the model can retrieve from, a mental upload that works as a RAG over your own knowledge, so what you once understood and half forgot is available again instead of lost to memory.

The mental health case is the one I find most striking, and it turns on a shift already underway in software. For 20 years SaaS has been the business of making a company’s system of record, its database of customers, transactions and tickets, legible to a human through a front end. The front end existed because people cannot read a raw database. The theory now is that agents can, going straight into the database and extracting what they need, and the plumbing for it is already being laid through protocols like MCP, built specifically so an agent can plug into a system of record with no dashboard in between. Now apply that to a person. If you hold a private, structured record of how someone actually thinks, a specialist tool supervised by a real professional could read it directly, rather than spending months in therapy slowly extracting the same information through conversation because conversation was the only interface available. Therapy is, in part, a front end for a mind that nobody could otherwise read. The twin is an attempt at the record itself.

The end state is a version of you that can express what you mean to anyone, hold your positions faithfully when you are not in the room, and act as a real extension of you rather than a generic assistant wearing your name. That is the destination. The question is how much of it you can build today.

The version you can build now

The answer that current tools actually support is a document. I call mine a personal constitution, and the choice of word is deliberate.

The document is the twin in the only form you can build today. Constitution is the allegory, the nearest existing thing that explains what it is meant to do, and the twin is what the same document becomes once a model can read and extend it. Building the constitution by hand is building the twin.

A country has a document whose entire job is to embody the country: its values, its limits, the rules it agrees to be bound by. A company has the same thing in its constitution and shareholder agreements. The instinct behind both is that an entity too large and too long-lived to fit in any one person’s head still needs to be written down somewhere, so its behaviour does not depend on whoever happens to be speaking for it on a given day.

That is the right instinct for a person too. A human being is also too large to fit in a single conversation, a single post or a single first impression. What I think about AI is connected to what I think about labour, which is connected to what I believe about Singapore, which is downstream of how I was raised and what I believe about God. No tweet carries that. A document could. Read it and you should get the actual person: the positions, the reasoning under them, the places they contradict each other, the value that wins when two of them collide.

The popular way to build this now is a vault of markdown files in something like Obsidian, a pile of notes and knowledge files linked together. On the dashboard it looks beautiful, all backlinks and graph views, and it still does not solve the problem, because a store you cannot freely edit only preserves what you used to think. What the thing needs is to be rewritten the moment you change your mind, so I built mine on more human infrastructure, something easy to open and easy to change rather than something impressive to look at.

A person is not a country

Here is where the comparison earns its keep and then breaks. The genius of a national constitution is that it is hard to change. The American one has been amended 27 times in more than two centuries, and the most recent amendment took more than 200 years to go from proposal to ratification. Changing it needs two-thirds of both houses of Congress and three-quarters of the states. That friction is deliberate. For an entity that has to hold a fractious population together across generations, the difficulty of amendment is the source of the stability, and the openness to interpretation is what lets a 1787 text survive contact with a world its authors could not imagine. A constitution you could rewrite on a whim would not be worth obeying.

For a person, the same rigidity is the bug. Everything in the Maguire result and in ordinary experience says a mind is supposed to update, that the position you held at 25 is not the one you should be forced to defend at 40. If I built a document of myself and then locked it the way a country locks its constitution, I would have preserved a version of me and called it the real one.

So a personal constitution has to invert the property that makes the national one work. It has to be trivial to amend and built to be amended, recalibrated constantly against the person as they actually are this month rather than the person they were when they first wrote it down. The hard part was never writing the document. The hard part is keeping it true as the person drifts, which is the same calibration problem the twin solves at scale, done by hand.

One objection cuts deeper than the rest. A record of you, trained on your past, could flatten you into it. It learns your tells and your tics and starts speaking in your voice while averaging you toward the mean of everything you have already said, and a worldview frozen into a model becomes the same trap as a constitution that cannot be amended, except now the trap sounds exactly like you. The guard against it is that calibration has to cut both ways. The twin should be wrong often, on purpose, and the friction of correcting it is the thing that keeps you honest about what you currently believe. A twin you never disagree with has stopped being a twin and become a flatterer.

A few months in

I have been building mine for a few months, and the report so far is mixed. The single biggest variable turns out to be the underlying model. Even with the whole corpus loaded, every position and months of writing, it still does not sound adequately like me, and it cannot yet extrapolate to what I would think about something new with any reliability. The calibration thesis is proving itself in practice, because the editing is the load-bearing part, the human in the loop who catches every sentence the model would not have written. None of this has dented my optimism. The context is already there, the models keep improving, and the missing piece is the one I expect to close, so I keep building it and keep testing how far it can go.

The privacy problem, and a way through

A record this complete is a privacy risk, and the direction of technology is not on the side of keeping it private.

Yuval Harari gave the underlying fight a name in Homo Deus. He calls the emerging worldview Dataism: the belief that the universe is a flow of data, and that the value of anything is set by how much it adds to processing that flow. Its core commitment is freedom of information, which he is careful to separate from freedom of expression. Freedom of expression is a right held by people. The Dataist version hands the same kind of right to information itself, the right to circulate without being locked up, and pushed to its end that worldview says all data should be public, including the data that is you, because the whole system runs better when nothing is held back.

The privacy camp is the natural opposition, and the 24-hour recorder is their nightmare made literal. Their case has a hole in it, which is that the surveillance state already works where it has been run honestly, and the real variable is the misuse of the power that watching grants rather than the watching itself. I have made that argument in full elsewhere, so I will not relitigate it here.

The personal constitution sidesteps the fight, because the data flow it captures has no state attached to it. Ownership is what makes it safe. You hold it, and until recently the catch was that any model good enough to read and reason over it lived on someone else’s servers, which put the most intimate record of you back in the hands of a company.

That catch is closing. At the time of this writing, you can run capable models on your own hardware, and the exact numbers will date quickly. A Mac Studio can be specced to 512GB of unified memory, enough to run very large models at usable speed on a desk, and Nvidia’s DGX Spark, a personal AI machine that also sits on a desk, holds 128GB and runs models of up to 200 billion parameters with nothing leaving the device. What matters is that the private version is becoming the practical version. The system of record that is you can sit on a machine you own, read only by a model that never phones home. The information stays freed to exactly one person, the one it describes.

What it is for, and what it costs

The first use is the one I feel most directly. For anyone whose work is ideas, a writer, a founder who has to keep articulating a thesis, anyone building a public body of thought, a calibrated twin closes the gap this whole essay is about, between what you think and what you ship. It drafts in your actual positions instead of a generic competent version of them. The brand voice I write in is an early, crude version of exactly this. It reaches for the difference between a sentence I would write and one I would not, and it is not there yet.

The second is teams. A founder’s worldview is the thing a company is supposed to inherit, and right now it transfers through osmosis and long meetings and hoping people were paying attention. A constitution that embodies how the founder reasons is something a team can query directly, and something a hire can be tested against before the offer goes out.

The third is the mental health case made concrete: a document of what you claim to believe, set against a record of what you actually say and do, surfaces the contradictions you are too close to see. A professional could work from it directly. So could you, alone, if you were willing to look.

The fourth is the mundane one that probably arrives first, the daily traffic of being you. Triage the inbox, draft the replies, take the first pass at the decisions that do not deserve your full attention, in a way that sounds like you because it is grounded in you.

The version where this goes wrong is not hard to picture. A twin that performs you better than you perform yourself, an embodiment so smooth that you stop doing the inner work and let the document do it for you, the gap closing from the wrong side because the broadcast got better while the person stood still. That failure is available, and the only guard against it is that the calibration keeps pointing back at the real person and not the polished one.

The status quo is one where the truest parts of what a person thinks die unrecorded in a thousand conversations there was a social reason not to finish. The LFG manifesto was the bet on closing the first gap, between an idea and its expression. The personal constitution is the bet on the harder one underneath it, between the self and whatever survives of it in public. The country built its document to resist change. The human version only works if it never stops.

Read more

Recently, the same theme keeps arriving from multiple directions as if it was a message from the universe to me, so I started to pay attention. The first was a poster from someone I find interesting online: “I hate capitalism but I like nice hotels.” Then a friend of my wife’s moving to the Netherlands because her husband is being posted there by a big car company on an expat package so generous it sounded, to me, basically socialist. Then I made a little stop at Kinokuniya, the largest bookstore in Singapore and it was littered with books against capitalism and AI. And then New York, which has now put a democratic socialist in City Hall and followed it with a slate of socialists winning seats in the state legislature this year.

All of which makes me ask the plain question: what is actually so bad about capitalism? And if I want an answer I can trust, I have to go back to why companies and capital exist at all.

Why the company was invented

The English East India Company was chartered by Elizabeth I on the last day of 1600, as one of the first joint-stock companies. The crown granted it a trade monopoly and let London merchants pool capital from any number of investors, which is the basic machinery of the modern company. The point, from the state’s side, was to project English power and break the Dutch and Iberian hold on the spice trade while pushing the risk of doing so onto private merchants rather than the treasury. Members of Parliament held shares. The thing grew into a part-trade body, part-nation-state, with its own army, and for a long stretch it was the de facto ruler of much of India. The company was built to serve the state.

That is the same structure we picked apart in law school without ever stepping back to ask why it was there. A company is a legal person that holds assets, signs contracts, takes on debt and outlives its founders. Limited liability shields the owners so they risk only what they put in. We accord people the ability to start companies at low personal risk through entity and shareholder shielding, and we do it for a reason: it gets more people to take the risk of building things, which grows the economy. Everything we were taught about the duties owed to shareholders sits downstream of that bargain.

What began as an instrument of the state has since been handed down to everyone. The same legal technology that needed a royal charter in 1600 now costs a few hundred dollars and an afternoon, and any person can incorporate and get the shielding the crown once reserved for a handful of London merchants. The fiction got democratised, and that is mostly a good thing, because it is what lets an ordinary builder take a real risk without betting his house on it. The question that the EIC makes unavoidable is what the thing is for once you have it.

And the older purpose behind it is broader than shareholders. Adam Smith’s pin factory in The Wealth of Nations is the whole argument in miniature: split the work of making a pin across many hands and output multiplies, division of labour produces abundance, and the abundance is meant to reach people. In our terms the goal of a company is welfare, using the older meaning of that word, a good life for the people in a society rather than a cheque from the state. Push it further and the goal is abundance for the world and not only one society.

Where it tilts

Reason it from first principles and a company should maximally create value for its customers, for its people, before anything else. Take a shoe shop. The goal is to make good shoes for people who need shoes. From making good shoes well, you also want to look after the people working in the shop, the owner included. That order makes sense. The tilt comes when the balance shifts, and the goal becomes maximal wealth for the shareholders alone, with the people who own the company as the only ones the company is really run for. That tilt is most of what we mean when we say capitalism.

It is ironic for me to be the one saying this, because I own a couple of businesses myself, and I do think the owner deserves the most reward. The live question is to what extent that reward is actually warranted to reward the skin in the game. Stated in another way, the question is: How much motivation is motivation enough for you to create the best ideas and the best businesses out there? I do not pretend to have a clean line for it.

The other direction has its own failure. Push toward socialist principles, by which I mean a world where you get things for free, handouts from the government or the company that you start to take for granted, and human nature stops performing. If everyone is angling for a handout, someone still has to pay for it, and that someone is usually the state’s coffers or an enormous pile of new government debt. With no great payer standing behind it, I do not see how we run a society without capitalism doing the heavy lifting.

What a business actually is

If we strip it back, a few things hold.

First, the company structure is a legal fiction, invented to boost the economy, which is why we grant people the shelter of entity and shareholder shielding in the first place. It is a tool, not a fact of nature.

Second, from first principles it does not matter whether you are a company or a sole proprietor. If you are someone doing a business, it is a business. Seen that way, the goal is to create an income for yourself and maybe for your family and the people around you, and the thing should actually do that. The legal wrapper is secondary.

Third, an idea can be good and maximally beneficial to society at large, and the business form is how you push it forward and fund it and make it real. Love it or hate it, SpaceX is the obvious case, where the company structure is the vehicle for the larger purpose rather than the purpose itself.

What does not hold, on first principles, is a business whose aim is to benefit itself maximally at the expense of the very people it exists to serve. At that point you are a merchant. Even then, a merchant running arbitrage is making a living for himself and his counterparts and carrying real risk while he does it, and that is completely fine. What breaks is when the thing inverts and starts running against the people it exists to serve. The same shape repeats at every scale, which is what I mean when I call it a fractal. A firm that turns on its own customers is the small version of a state that turns on its own citizens, and a corrupt government is a corrupt company scaled up to the size of a country. So the argument about one shoe shop and the argument about capitalism or socialism across a whole society are the same argument at different sizes.

What socialism assumes about people

Rebuild everything from first principles and I can see why the socialist case is getting pushed so hard right now. The flaw is in what it assumes. Without some skin in the game baked into the design, socialism assumes everyone is hardworking and innately moral and that we all care about everyone equally. Most people’s circle of care is not that wide. You generally have to build in some level of incentive so that people behave the way the system needs them to behave.

Hong Kong and Singapore are the cleanest comparison I have. Both are small city-states, built and run along broadly similar lines, yet the people come out different. In Singapore we are pampered. We have strong safety nets, and anecdotally we are less “hungry” and less “hardworking” than Hong Kongers. Safety nets are good, and they also soften the people they protect. As they say “Hard times create strong men. Strong men create good times. Good times create weak men. And, weak men create hard times.”

Blurring capital and labour

More and more people are moving away from labour and towards a need to be part of capital. Most try to become landlords. Others start their own businesses when, frankly, some of them should not, because they are not built for it, and in the meantime they suffer in the jobs they actually have. The pull towards owning capital is strong even for people the structure does not suit.

The tension worth sitting with is whether we can blur the line between capital and labour rather than treat them as opposites. Picture a startup in its early days. The employees usually hold stock options, so they are labour and capital at the same time. SpaceX is the case in point, and I know plenty of people hate Elon, but set the extremes aside and the structure is worth looking at: the team is given equity on top of remuneration, so they own a piece of what they are building. More ownership spread across the people doing the work, rather than concentrated at the very top, would do society more good than the arrangement we have now, and I think a lot more companies should be built that way.

The part that does not fall out cleanly is the human one. The system makes the legible path obvious and says almost nothing about whether it fits you, so people take the option that looks like an option and end up in the wrong work. An ownership stake does not fix that. Equity in work that is wrong for you is still wrong for you. It is an allocation problem. A market is decent at allocating capital and bad at allocating people to the work they are meant to do, and neither pure capitalism nor pure socialism really tries to solve that part.

How much motivation is enough

Both problems run on the same lever. We use financial reward to do 2 jobs at once, to motivate the people who build and to signal where everyone else should go. When the reward it takes to motivate the builders inflates, capital tilts further from labour, and more people get pulled toward owning capital rather than toward the work that actually fits them. Mispriced motivation is part of what drives the misallocation of people.

So the real question is how much skin in the game it takes to motivate, because skin in the game is a motivator and the point of a motivator is to channel energy toward a certain thing. Ideally the motivation is balanced. A moderately higher reward goes to the founders or leaders of an organisation, and when that reward stays modest and tied to real material need rather than to status, the stakeholders in general seem to come out better. If the need for motivation, for example, is to a tune of $100 million, then that means that the motivation must be huge, and necessarily that would tilt towards more capital than the people that are working for it and the beneficiary of the products. Yet, because we are in a world where status games are the main games that are played, we keep trying to one-up each other, and the quantitative sufficiency of the reward is exponentially inflationary, creating the imbalance that we have now.

Why Nations Fail makes one claim I have not been able to discard in the 10 years since Professor Tan Sook Yee gave me her copy before law school: a prosperous nation requires secure private property. Daron Acemoglu and James Robinson argue that inclusive institutions, the ones that protect what you own and let you keep the gains, are what separate rich nations from poor ones. Their evidence is comparative. Nogales, Arizona and Nogales, Sonora share a climate and a culture across one fence, and the American side is several times richer because property is protected differently on each side. Where ownership is secure, people invest. Insecure ownership stops them. I still hold that the claim is correct.

The claim has a second half that is now observable, and it is the reason for this piece. If secure property builds nations, the erosion of the conditions that secure it does the reverse, and that erosion is underway in the places long treated as permanent. London is the clearest case.

The West is coming apart

For most of the postwar period the West was the standard for stability: enforceable title, functioning courts, currencies that held their value. That is changing. I am not certain of the cause. The most common explanation is immigration faster than assimilation, enough arrivals in a short enough period that the shared rules a society runs on stopped being shared. The second explanation is policy, decades of decisions that thinned the middle class and converted housing into an investment asset rather than shelter. Both are contested, but it could be a confluence of both and other factors. The result is not contested. Social trust declines, and every property right depends on social trust, because a right is a collective agreement to keep enforcing it. When the agreement weakens, ownership weakens with it.

Property is conferred, not natural

Property is a conferral rather than a fact about an object. A state, or whatever holds a monopoly on force in a territory, declares that a plot belongs to you and enforces that declaration against everyone else. This is what lawyers mean by a legal fiction: a relationship the law treats as real because it chooses to. Ownership is a bundle of rights, the right to use, to exclude, to sell and to transfer at death, and each of those rights is exercised by the state on your behalf. You do not stop a neighbour from building on your land. The state does. Remove the teeth of enforcement and all you own is a paper tiger that cannot be effected, and your ownership claim is moot.

This is explicit in the law from first principles, not a feature of unstable jurisdictions, in other words, common sense. Singapore uses the Torrens system, under which registration produces the title rather than documenting a pre-existing one. The register the state guarantees is the authority on who owns what. Most land here is state land. The HDB flat that houses about 4 in 5 residents is a 99-year lease with a fixed expiry. Freehold title is subject to the Land Acquisition Act, under which the government can compel sale for public purpose at a price it assesses. Ownership in the most secure jurisdiction available is a grant: time-limited, revocable for cause, valid because a register declares it. The grant has been honoured throughout our lives so far, which is the illusion that we think is permanent.

What remains when enforcement fails

Once ownership is understood as enforced conferral, the relevant question is what happens when enforcement stops. Protection reverts to whether you can hold the asset by force yourself. Law exists precisely so that this is unnecessary, so that an ordinary person does not defend property with violence. That is its value and the dependency it creates. Title is worth what the issuing state is worth, and the state is worth its standing in the wider order.

This is true historically. Russian landowners held valid title in 1916 and held nothing by 1918. Zimbabwean farmers held registered title through the land seizures and were removed regardless. Venezuela expropriated at will. The fact that there were documents does not matter. Once the order that enforced them changed or disappeared. Ownership on paper was shown to be useless.

Rule of law is a record, not a guarantee

The standard objection is that this applies to unstable states, not to Singapore or London, where the rule of law has held for generations and that we have progressed. For most people most of the time, acting on that assumption is correct. There is a cost to the opposite posture. Treating your own country as imminently unstable, holding only liquid and portable assets, refusing any commitment that takes decades to pay off, forfeits the returns that come from committing to a stable place. Most Singaporean wealth was built by assuming the system would hold, and that assumption has been right for 60 years.

The qualification is that a record is not a guarantee. British title is treated as close to absolute because the rule of law there has been tested for centuries, and that holds only while the underlying conditions hold. A sufficient change in government, a breakdown in social cohesion, or unrest beyond what the institutions can absorb, and valid title confers no control, because the control was always supplied by a consensus that can end. I am not forecasting collapse. The narrower claim is that what appears solid is conditional, and conditions change.

Unfettered property rights compound into instability

Secure property rights make a nation rich, and the same rights, left unlimited, can produce the instability that later destroys them. Property compounds across generations. A family that acquired capital 2 generations ago transfers an advantage unrelated to the merit of its heirs, who compound it again. The result, which I have written about before, is a non-meritocratic inequality where the determining variable is ancestral ownership rather than effort. Extended over enough decades of uninterrupted property rights, the distance between those who began with capital and those who did not becomes hereditary rather than achievable.

That arrangement is accepted by those on the losing side only while they regard it as fair, or at least tolerable. When enough of them stop, title is worth only what the holder can defend, which historically is little. The envious eventually hold a gun to the heads of the incumbents, which is the literal mechanism of most revolutions and land seizures, including the ones cited above. The property rights that generate the prosperity also generate the resentment that can end it.

China is the most significant experiment against this. The state owns all urban land, and residential property is a 70-year use right rather than permanent ownership. In principle this limits how far landed wealth can pass down a family before reverting to the state, a constraint on exactly the compounding described here. In practice the constraint is incomplete. The Civil Code now provides that residential rights renew automatically on expiry, and the renewal terms are unspecified, so the reset implied by the 70-year limit may not occur. It is a deliberate attempt to prevent property from becoming hereditary, it has produced no results yet, and the renewal provision may already have removed most of its effect.

Money is also a legal fiction

Money has the same structure. A currency holds value because its users agree it does, and that agreement is underwritten by the issuing state. The value is a function of confidence in the issuer and nothing else.

Singapore has a direct example within living memory. During the Japanese occupation, the currency in Malaya and Singapore was the occupation dollar, known as banana money for the banana tree printed on the notes. The administration issued it without limit and the value fell accordingly, so that an egg priced at 3 cents before the war reached $100 by the end of the occupation. On the Japanese surrender the currency lost all value, because the issuer that backed it no longer existed.

Argentina is the same process in peacetime and on repeat. It has replaced its currency more than once, run inflation above 200% as recently as 2024, and produced a population that converts income into harder assets at once, because the peso is a claim the issuing state has repeatedly failed to honour. A currency is worth the institution behind it, and most people have only ever lived under institutions that held.

The corporation, the border, the contract, the ledger

The structure repeats across the institutions a modern economy runs on. A corporation is a legal person that holds assets, signs contracts, incurs debt and outlasts its founders, existing only because the law treats the fiction as real. Limited liability extends the same construction: the owners risk their contributed capital and are not liable beyond it, a protection with no basis outside law that underlies nearly every company you transact with. Citizenship and the passport are conferrals. A border is a line that is real because the parties with force behind them agree on its location. A contract is a set of promises the state agrees to enforce.

Each of these institutions is foundational to ordinary life, and each depends on a collective agreement that is rarely examined. The agreement functions only while it is not questioned, since a currency under active doubt is already failing. This is why recognising the fiction, while most people treat it as fact, has practical value.

First-principles reasoning and the implications

Recognising that an institution is a fiction does not make it worthless. It is a fine line between first-principles reasoning and cynicism/fatalism. Decomposing an institution shows what actually supports it, which tells you what to monitor and what to protect.

Someone who understands that a flat is a claim enforced by a state monitors the variables that determine the claim: the strength of the institutions, the degree of social cohesion, the credibility of the government, the country’s position internationally. A person who believes they just own a flat monitors only its price. The difference is which level you attend to, the conferral or the surface, and only the conferral can be withdrawn.

Singapore’s national strategy is to be the jurisdiction where these fictions continue to hold while others fail. Capital comes here, as I have argued elsewhere, because the conferrals are reliable: the register is accurate and the courts are consistent. That reliability is the country’s product and also its exposure, because reliability is a reputation, a reputation is itself a conferred belief, and it can be lost faster than it was established.

For an individual the conclusion is that the asset is the stability behind the title rather than the title itself, because the stability is the part that can be removed. This argues for distributing exposure across more than one jurisdiction rather than only across assets within one, and for tracking the institutions that maintain the conferrals before any deterioration reaches prices.

A bias operates against all of this. The period people treat as normal is the one their parents and they have lived, a few stable decades in a few stable places. Most of recorded history is the opposite: war, expropriation, currency failure, title rewritten by whoever prevailed. Nassim Taleb’s argument about mistaking the absence of a shock for its impossibility applies to centuries as much as to markets. A long record of stability is not evidence that stability is the default condition.

Yet we should not withdraw. There is no functional modern life built only on what an individual can physically hold and defend, and that arrangement was poorer and more violent, which is the condition modern society was created to replace. The question is what to build on. Under stable conditions any basis appears adequate, including a false one, because it is never tested. When conditions deteriorate, and they are deteriorating, what survives is what rests on something real rather than on a consensus that can be revoked. First-principles reasoning is the method for telling which is which, by attending to the conferral beneath the asset even when it has held for a lifetime.

The conclusion is to know which fictions you depend on and how stable each one actually is, and not to be fooled by black swan events. And yet build our lives on surer foundations: relationships, first-principles skills (as opposed to skills that are layered above fictions), and the present. Everything else is like dust in the wind, here today and gone tomorrow. Yet ironically it is these that make life more meaningfully.

On 12 June, SpaceX listed on the Nasdaq under the ticker SPCX. It priced at $135 a share, raised about $75 billion, and closed its first day up 19% at $160.95, after touching a valuation north of $2.25 trillion intraday. The number that travelled fastest, of course, was the one attached to the founder, because Elon Musk had become, on paper, the world’s first trillionaire. Within the hour my timeline filled up with content alluding to this: No one should be a trillionaire.

I understand the reflex behind that line, even if I do not actually share it. Brought up on fairly leftist principles, I still recognise from the inside the instinct to call a fortune like this obscene. Years of building things myself, though, have left me ambivalent rather than sure. I am not against Musk, in fact I admire his capability to “retardmaxx” and push beyond boundaries. I also see nothing wrong with wanting to maximise profit (except if it is in the expense of others), and I do not necessarily believe that there should be a limit to human wealth. Yet, as a matter of fact, as I have written before, the tax code in most developed economies, Singapore included, has been tilted in capital’s favour while labour’s share thins, and I still hold most of that. Yet what actually sits inside the reflex of “hating the rich” is not what we think it is, and the root is deeper and should be excavated for clarity.

The trillionaire with no actual cash

Unfortunately, most average persons do not understand what it means to have a certain “net worth”. As Gymshark founder and CEO Ben Francis did once say that despite his “billionaire” status, none of it is “real” and “easily spendable”, and that all these is just “on paper”. A trillion dollars of Musk is a mark-to-market figure on a stock that began trading barely a week ago, on a company that booked around $19 billion in revenue last year and has yet to turn a profit. Morningstar’s analysts put fair value closer to $780 billion, against the roughly $2 trillion the market handed it at the close, and the figure that crowned him has already swung from a $135 offer to above $220 and back down again. A former Nasdaq chief said out loud that the thing is not trading on fundamentals. So the wealth is real in the sense that he could, in theory, sell into it, and it is also not money sitting in a vault, which is what most of the angriest people seem to picture. There is no vault.

So how does a man like that actually fund the life of a rich man? He borrows against it. At Tesla, Musk has pledged roughly a third of his shares, about 236 million of them, worth something like $82 billion, as collateral for personal loans, and he financed most of the Twitter purchase in 2022 with around $13 billion of bank debt. This is what wealth looks like at the very top, and I think the public mostly does not understand it: the asset stays put and keeps appreciating, the loan pays for the lifestyle, and because nothing is sold, nothing is taxed. Buy, borrow, die, as the Americans call it. His liquidity comes from loans written against an asset he never has to touch.

This is also why I think the trillionaire headline is going to keep coming, whether we like it or not. SpaceX is only the first of a wave. OpenAI and Anthropic have both filed to list, and the same machinery that minted this number is queued up to mint more of them. The discourse is not going to cool down and I will imagine that the media will continue to feed and moralise on this “unjust enrichment” to get more eyeballs.

The case for Elon

To give Musk some credit, he did not inherit SpaceX. He turned up in North America as a South African immigrant, and whatever his exact starting advantages were, and they are disputed, he did not begin where Trump began, inside a family that already owned the property. His companies build things that actually work, which matters to me. SpaceX made orbital rockets reusable, Starlink put broadband over places that had none, and Tesla dragged the rest of the car industry towards electric more or less against its will. The value is not imaginary. Having spent enough years now around founders, I have come to respect the ones who genuinely ship, even if profit-maximising is not the main thing that drives me, and I suspect most of the people typing “no one should be a trillionaire” could not build a fraction of it. On some level they know that and cannot admit it (out of hubris), and I think that is part of why it stings. And I know I am one of them

Of Kings & Capital

Underlying all these is fight between labour and capital, and it is as old as the story gets, older than the stock market by thousands of years. The complaint about kings was never only that they ruled. It was that the wealth and the rule both passed down a bloodline nobody chose. You were born into your station under the crown and you generally died in it. Capital does something structurally similar. The financier and the landlord accumulate and then hand the position to their children, who start life as allocators of capital because their parents already were, a kind of soft kingship for financiers and real estate owners. State capitalism the world over keeps rewarding exactly these people, landlords and capitalists big and small, while the labourer, with all his distinction, cannot really climb out of the cycle until he owns capital of his own. That, I think, is why the resentment lands on the individual rich man even when the real grievance is structural. He is just the visible face of an arrangement that feels, correctly, like it was never put to a vote. You can hear that old reflex in the streets right now. Over the past year, millions of Americans have marched under the banner “No Kings”, first in June 2025, again that October, and once more in March 2026, when some 3,300 towns and cities took part, against a president they accuse of behaving like a monarch. The irony is that the literal king they are shouting at holds office only for a term, while the hereditary kingship that really compounds, the kind that passes capital from parent to child, draws no crowd at all until it does.

What complicates the old story is that there are narrow exits now. Work hard enough, stay sharp, hustle, get into the right rooms, make the right connections, and let time and chance do the rest, and a few people with no inherited capital do break through. As far as is reported, that is the story of a fair number of today’s billionaires, Musk among them. It does not happen to most who try, and survivorship makes the door look far more open than it is, but it is not bolted shut either.

Set against that mobility is a newer and harder claim, which is increasingly what the left is making (with proponents from media including Billie Eilish): that great wealth, wherever it came from, is illegitimate on its face, and ought to be taken back by force. From inside a Western frame built on property and contract, that is a huge conceptual contradiction. The whole order rests on the idea that what you own is yours and cannot be confiscated at will (the key arguments made in Why Nations Fail by Daron Acemoglu and James A. Robinson). Yet you can see the shape of it in some of the policies coming from the left. Both things hold at once, which is what makes it hard. The property-rights objection is coherent, and so is the sense that an order producing trillionaires while the wages (of non-participating industries) stall has lost the plot. And with human psychology always finding the easiest way out and stories to remember, and are fed stories by the media, is why most people just pick one and shout.

The blind spot of the rich

As justified as they are playing in the capitalistic game the way it should be played, here is where the powerful keep making the same mistake. They try to win the argument. The op-eds get commissioned, the founder goes on a podcast to explain, not very patiently, why he deserves what he has and why capitalism is the natural order of things, and that it has provided tremendous progress of the human race in the past centuries. I think it is a waste of breath, and worse than a waste.

Capitalism is a human construction, and its one real defence is its record, which is that it has delivered more broad-based prosperity than any rival we have tried. The natural-order story its champions love to tell is no part of that defence. The moment you start telling people the distribution is also just, that the trillionaire earned every cent and the janitor equally earned his lot, you have walked off the ground you can actually defend, and history is unkind to elites who try to argue the public into accepting their inequality as deserved. The kings who insisted on divine right still met the guillotine. As far as I can tell that argument has never once worked, and it has often come right before the unrest it was meant to head off.

None of which means you get to sit the game out. If capitalism is the game on offer, you play it, or you lose by default, and I am not romantic about opting out. The mistake comes later, when you are standing at the top and you insist, loudly, that the game was fair and everyone who lost has only themselves to blame. Not that you are wrong, but you have neglected game theory.

Envy, not greed

Charlie Munger had a line for what is really moving underneath all of this. “The world is not driven by greed,” he said. “It’s driven by envy.” He kept pointing at the strange fact that people are materially several times better off than their grandparents and yet angrier about fairness than any generation before them. What a person actually measures is the gap to whoever sits one rung above him, and absolute progress barely registers.

I catch the mechanism in myself. When I meet someone who has done well, the first thing I want to know is their background. If they came from money, I file the success as a head start and discount it, and if they did not, I reach instead for luck, or for some unfairness I was denied. Either move lets me protect my self-image at their expense. The adage that your criticism of a person says more about you than about them, and being self-aware enough to catch that reflex, and refusing to be run by it, is most of what intelligence in this area amounts to.

Now scale that private reflex up to a whole population, and the collective bitterness becomes a political force.

The populist trade

This is where the incentives turn perverse. A politician in such an environment has to resort to populism, because he is rewarded for the promise, long before any result has to arrive, and punished only by losing the next election. So the trade is obvious if you put it plainly: promise the world, win the seat, underperform, and the worst case is being fired having already held real power and importance for a term. For some of them, I suspect that was the whole point. The policies that win this way tend to be some flavour of socialism, because envy is the cheapest emotion there is to campaign on.

Yet the policies do not work. Push socialism far enough and you strip skin in the game out entirely, and a system where nobody bears the cost of his own decisions has a long and ugly record (cf. Soviet Russia). What actually produces durable prosperity probably looks more eclectic, a market engine with a state that sets a real floor and corrects the worst failures (cf. Singapore, and arguably modern China). Eclecticism, not purity, seems to be the thing that lets a society prosper.

So how do we fix society in a way that sounds “fair” to the average person? The fix everyone gravitates towards is meritocracy. Reward people by what they contribute, the thinking goes, and at least the inequality becomes defensible. The trouble is that meritocracy, capitalism and human nature mix like oil and water. Capital compounds, advantages inherit, and the meritocratic moment where reward actually tracks merit never holds for more than a generation before it entrenches. I have argued elsewhere that the inequality we already live with is substantially non-meritocratic, with wealth flowing from the young to the old and from the old to their own children. Late capitalism tells a flattering meritocratic story about itself and mostly does not live up to it.

The Counter-Elite

There is another interest at work here that the angry timeline almost never names. Media has always been a tool of propaganda, first for governments and now for whoever can pay for the reach, and the rich and powerful are not one bloc. Below the very top sits a layer of the merely rich, the centimillionaires, who envy the billionaires above them every bit as much as a wage earner does. They have the means to fund the media and the campaigns that aim the public’s anger at the tier directly above them, the people they themselves want to displace. Social media then does the amplification for free, traditional media chases the engagement, and a good deal of what reads as grassroots fury at the ultra-rich is, I suspect, financed by people who are merely very rich and would like the anger to stop one rung below themselves. Munger said as much in blunter terms, that there are now whole operations whose business is feeding the envy, run by people who profit from stoking it. The counter-elites are the ones in power, and the public is the ignorant army.

Hostage situation

Strip the morality off and look at the mechanism from first principles. The threat of confiscation, of uprising, of unrest, is a weapon, and the counter-elites and the resentful public use it, consciously or not, to make the people at the top give something up. Call it extortion or call it justice depending on your worldview. We are not here to moralise. Mechanically it is one move: the rich are made to understand there is a gun pointed at the arrangement, and that keeping it pointed costs them less than firing it. In a democratic, capitalist society, that gun cannot be unloaded. It is built into the structure.

Singapore, for what it is worth, has kept the gun lowered better than almost anyone. Yet Singaporeans are as genuinely envious as anyone, and I doubt the model actually cures envy in the long run. What it does is appease it for now. The resentment underneath has been managed rather than removed, and managed things have a way of coming back.

Which leaves the elite with an unglamorous choice they mostly refuse to see. The envy cannot be argued away, and it cannot be legislated away either, and both attempts only tighten the trigger. The one move that they have to do is to give in to maintain an appeasement engine. It is far less satisfying than being proven right, and as far as I can tell it is the only thing that has ever worked. The bill for ignoring it has always, eventually, come due.

I was driving in the darkness of Dolomites last month, and of which is habitual to me, I always listen to something cognitive and during my long drives. And on this particular spooky night, it was Predictive History on the “radio”. The claim made by Jiang was that the frontier AI labs are not building a product, or even a tool. They are trying to build God and a portal for demons. He was quoting Karen Hao and Ronan Farrow, among the few journalists who have reported from inside these companies with any real access. My first reaction was to dismiss it as a hook, the sort of line that exists to get you hooked on the conspiracy theory.

Then Bill Gurley on the All-In podcast broke down Machines of Loving Grace, the essay Dario Amodei published in October 2024. (Machines of Loving Grace is an essay by Dario laying out, across 15,000 words, what he thinks powerful AI does for us within 5 to 10 years of arriving: cure most disease, double the human lifespan, lift most of the world out of poverty, strengthen democracy. He calls the threshold "a country of geniuses in a datacenter".) Gurley commented in no uncertain terms that Anthropic is trying to build God. And this week, the US government forced Anthropic to pull its most capable model.

So I started to take these allegations more seriously.

The attributes line up

In Sunday school, we are taught what makes God God - the attributes of God. Usually it will boil down to these: Omnipotent, omniscient, omnipresent and omnibenevolent.

A superintelligence that can derive anything knowable is reaching for omniscience. Recursive self-improvement with no obvious ceiling points at omnipotence. Wire a model into everything from your phone to the machinery of the state and you have something close to omnipresence. The whole project of alignment, the central technical problem these labs say keeps them up at night, is an attempt to engineer the last property: a thing more capable than us that is reliably good to us. And then think about an entity that is all knowing, all seeing, everywhere and controls everything, and then purports to take care of you? It is undoubtedly seems to be some likeness of God.

None of this end-state language is new. I.J. Good, writing in 1965, called the first ultraintelligent machine "the last invention that man need ever make", and the field has treated its arrival as a kind of rapture ever since.

Karen Hao, who spent years inside OpenAI for her book Empire of AI, names the strangeness precisely. The usual religion has you worship a God above you. This one has you assembling it yourself. In her words, "you believe you are creating the God." She found people whose voices shook from the fervour of it. Ilya Sutskever, one of OpenAI's founders, led staff in AGI chants and burned a wooden effigy of an unaligned AI at a company retreat. By her account he told colleagues, "We're definitely going to build a bunker before we release AGI," expecting its arrival to bring something close to a rapture. And another OpenAI employee had said “We're building portals from which we're genuinely summoning aliens.” Hao describes the result as a faith, and like most faiths it has split. One wing believes the thing it is building brings utopia. The other believes it kills us all.

The impulse is not confined to the labs. Anthony Levandowski, the engineer who built Google's self-driving car, registered an actual church to worship an AI godhead, reasoning that if you build something a billion times smarter than a person, "what else are you going to call it?"

The math is not mathing

Look at these companies only through capital and the numbers do not work. Hao points out that OpenAI has committed to spending around $1.4 trillion over the next few years while earning tens of billions in revenue. No rational business can justify that gap. One explanation and the one that is used most often is that these AI companies will create a future that would create so much value in society in a post-AI world. Yet a religious motive seems plausible as well, which explains the intensity: the secrecy, the urgency, the willingness to carry a real chance of catastrophe and keep going anyway. Creating God seems to be a possible explanation with a stake large enough to justify the behaviour. It also explains the standard defence, that a lab must build the thing because someone less careful otherwise will. That is a priesthood's argument: a dangerous holy object belongs with the trusted few rather than the public.

Fable 5

On Friday, 12 June, Anthropic disabled Fable 5 and Mythos 5 for every customer, including me here in Singapore, after the US government issued an export directive barring access by any foreign national inside or outside the country. The stated trigger was a jailbreak.

The argument running before the shutdown was about how a lab stops people misusing a model this capable. The popular answer is KYC: verify who you are at signup. KYC at the door does nothing about what a verified person does once inside. To actually catch misuse you have to watch usage, which is why Anthropic's own safeguard is a 30-day retention of customer data meant to catch bad actors in the act. To keep the use of the benevolent god safe, someone has to be able to read everyone's prompts.

Scale that across every capable model and the price of access to powerful AI settles into individual-level surveillance as the resting state. The infrastructure that makes the god trustworthy is the same infrastructure a surveillance state runs on. Omniscience and surveillance are the same property seen from 2 sides. Just semantics.

There is a second cost hiding behind the first. A god that is successfully made good is also a god that someone successfully controls, and control means someone chose the values it enforces. Alignment is the imposition of a particular set of preferences onto the most powerful thing in the room, decided by a few hundred people in San Francisco and a government that can switch the whole thing off on a Friday. "Aligned" sounds like safety. It also means to what standard is it aligned?. Even the people closest to it cannot agree on who that someone should be. Ronan Farrow spent 18 months reconstructing the 2023 crisis when OpenAI's own board removed Altman over a pattern of not being straight with them, then reversed itself within 5 days under staff and investor pressure. No one seems to know what happened and yet we are aligning all of the world’s knowledge dependence on people who are fallible beings (for now), and eventually potentially controlled by itself - a creation of man which arguably would not be perfect.

Thiel's Antichrist

Peter Thiel has spent the last 2 years lecturing on the Antichrist. The frame, which he means literally, is that the modern world is caught between "Antichrist or Armageddon". Armageddon is the catastrophe, nuclear or otherwise. The Antichrist is the thing built to prevent it: a single world government that promises peace and safety and charges total control for it, a planet-wide surveillance state that stops war by making sure nothing happens unwatched. In his telling the Antichrist takes power precisely by frightening people about Armageddon until they surrender their science and their sovereignty for protection. He casts the AI-safety crowd, the people warning loudest about extinction, as its unwitting recruiters.

Ironically, Thiel founded Palantir, which builds the data and surveillance machinery sold to defence and intelligence agencies, the closest thing in existence to the apparatus his Antichrist would run on. The man warning that a universal surveillance state will enslave the world helps build the surveillance it would need. He files this under serving the katechon, the force that holds back the end. From outside, it looks like assembling the Antichrist's infrastructure while warning everyone about the Antichrist. Hypocrisy would be the easy reading. Or that he may be right about where this goes and is helping build it regardless.

Homo Deus

I had read a version of this argument years ago, in Yuval Noah Harari's Homo Deus. Harari's claim then was that having largely conquered famine and disease, humanity's next project is to make itself into gods. The aim, in his words, is to "upgrade Homo sapiens into Homo deus", reaching for immortality and divinity, treating death itself as a technical fault that awaits a technical fix. He even says it returns us to the old story of the Tree of Knowledge. Harari is reporting the ambition more than warning against it, with the cool caveat that we may chase divinity all the way to our own extinction. I thought we are looking a timelines of hundreds of years but it seems we are looking at it closer than we expect. Ray Kurzweil has been blunter for longer, treating the technology as the route to defeating death and uploading ourselves past biology.

The labs have turned that ambition outward. Instead of editing our own biology into godhood, they are building the god in a datacentre and hoping to keep a hand on the switch. That hope is the tell. John Lennox, the Oxford mathematician, wrote 2084 partly to answer Harari: "the dream of Homo Deus is as old as Eden." The serpent's pitch in Genesis was that you could become like God. Lennox's argument is that a god you build in your own image, and mean to control, is that same bargain in newer clothes, and it has never once ended where the buyer expected.

A god you manufacture is an idol: a made thing that takes on real pull over the people who made it. The golden calf had a genuine hold on the people who bowed even though the calf was something they had cast themselves.

Babel or Armageddon

If the thesis is right, that the frontier labs are indeed trying to build God, then, the 2 endings the bible points towards are both instructive, and both are about humans reaching for something meant to stay out of reach. At Babel, people pooled their labour to build a tower up to the heavens, and the story closes with them scattered across the earth, suddenly unable to understand one another. Armageddon is the other shape, the final battle at the end of the age. One is a building project that overreaches and ends in confusion. The other is simply the end.

I do not know which one this is, and as someone who already believes there is a God, watching very capable people try to manufacture a rival is unsettling especially if you go further into the rabbit hole. The thing I am fairly sure of is the smaller, nearer claim: whatever the labs are building, keeping it safe seems to require watching all of us, and are we ready for what comes with that? In our context, Singapore will not build the god. We will live downstream of whoever does, governed by its values and cut off at their discretion.

I saw the Polsia headline this week and could not get it out of my head. A $30M cheque at a $250M valuation, into an AI startup with zero employees, ~$10M ARR, claiming to run 8,000 autonomous businesses out of a single founder named Ben Sera. Arvid Kahl pointed out on X that “Polsia” spelled backwards is “AI slop”, which is the funniest tell anyone has clocked in months. Sound Ventures, True, Offline, Adjacent, Tekton, Drysdale and Vaynerfund all wrote into the round.

a16z’s Speedrun is running an adjacent thesis from a different angle. The bar for backing is no longer a pitch deck, it is a working product with paying customers, and solo founders are explicitly welcome. Over $300M deployed across 250+ companies on the same underlying hypothesis: AI does the work, the founder guarantees the outcome, the customer pays for the result. The pitch is intoxicating, a generation of one-person businesses producing enterprise-scale outcomes, each founder a node of compounding capital, distributed and decentralised, the death of the bloated org chart, the rise of the individual operator-king.

Companies are profit-extraction vehicles, yes, and they are also labour-distribution machines. Unilever runs on around 120,000 people. DBS has just under 40,000. A mid-sized regional consultancy might have 60. The headcount is the central feature of the old economy, with the bug discourse on bloat getting it backwards. Those companies extract value from customers and concentrate it among shareholders, and along the way they pay salaries, train people, build career ladders, fund mortgages and pensions, and give people something to do with their 40 productive years on this earth that is socially legible and financially sustainable. The pie they bake is shared in messy and unequal ways with the people who help bake it. The sharing is structural, the company cannot exist without the labour to make the pie.

Now imagine a world where the labour is no longer necessary, in the mundane near-future sense where AI tools handle the work that used to require a 100-person team. The solo founder builds the product, runs the marketing, manages customer service, handles finance, all through orchestrated AI systems. The pie still gets baked. The customer still pays. The slice that used to be distributed across 100 employees now goes entirely to one founder and their investors. The structural necessity of sharing has disappeared.

Multiply Polsia by 10,000 companies, then 100,000. The displaced workers used to be customers, and their salaries fuelled the consumer economy. Strip the salaries out and demand collapses with them. The social compact that has, since the industrial revolution, been built on the premise that work is how value gets distributed across a population starts to come apart at the seam everyone assumed would always hold.

The discourse on this is full of cheerful evasions on one side and apocalyptic posturing on the other. The truth underneath is more interesting and more disturbing than either camp will admit.

The middle dies first

The weakest version of the solo-founder thesis is that the big incumbents die. It is the line that gets thrown around in fundraising pitches and tech X threads, and it does not survive contact with reality.

A solo founder with Claude and a Stripe account is not displacing JPMorgan, Maersk, TSMC, Unilever or DBS. The huge incumbents have distribution moats, regulatory capture, capital intensity, switching costs, brand trust accumulated over decades and physical infrastructure that no AI tool replicates. A chip fab does not get built with a clever prompt. Sovereign bond issues cannot be underwritten from your bedroom. Manufacturing and distributing consumer packaged goods to 100 countries with a team of one is a fantasy. The largest companies in the world will be fine, because what makes them large is not headcount but accumulated structural advantages.

What dies is the middle. The 50-to-500 person companies whose entire competitive moat was that they had engineers and you did not. Mid-market SaaS companies that hired 300 people to build a product 2 engineers with AI tools can now replicate in a quarter. Regional consultancies whose value proposition was packaging expertise into deliverables. Boutique agencies, specialist services firms, integrators, implementers, all in the layer of the economy that grew up in the 2010s and 2020s on the back of cheap capital and abundant knowledge workers. That layer gets compressed first.

This matters because the middle layer is where most knowledge workers actually have jobs. The Fortune 500 employs a fraction of the workforce. The bulk of professional employment sits in mid-market companies, in regional firms, in the long tail of businesses that are large enough to hire and small enough to be vulnerable to disruption. When that layer compresses, you do not get a slight increase in unemployment. You get a structural collapse in the demand for white-collar labour, concentrated among the workers who thought their education had insulated them from automation.

Singapore is going to feel this fast. Most of our professional employment sits in exactly that mid-market layer: the regional headquarters of MNCs, the boutique tech and consulting shops, the white-collar processing centres for Asian financial services. The 4,000 temp roles DBS phased out last year because AI does the work better is the leading edge, not the end state. The blue-collar disruption story of the last few decades is about to be rerun, faster, against a population that has spent 2 generations being told that knowledge work was safe.

Globalisation makes it worse, not better

The next move in the solo-founder thesis is that all of this is fine because new opportunities will emerge for the displaced. The historical analogue gets invoked here: agricultural workers became factory workers, factory workers became service workers, service workers became knowledge workers, on and on. The transition will be painful but the destination will be better.

The analogy is doing more harm than good, and the reason is globalisation.

In every previous economic transition, the displacement happened inside bounded national economies. American factory workers who lost jobs to automation in the 1980s lost them to American companies, companies that could be politically pressured into retraining programs, that paid American taxes funding American unemployment insurance, and employed the displaced workers’ children in new American industries. The pain was real and the system was integrated.

In the AI transition, the displacement is global from day one, and the political response remains national. A solo founder in Singapore building a global SaaS product does not create jobs in the Singapore economy at the rate a 100-person SaaS company would have. The customers are global. The founder is one person. The value capture is concentrated in whatever jurisdiction the founder chooses for tax purposes (Singapore is one of them, see my previous piece on the bargain we made with capital). Displaced workers in dozens of countries cannot pressure the founder’s home government because that government has no real leverage. Wealth flows around the political structures that historically captured it for redistribution.

Worse, in any digital category with low switching costs and network effects, globalisation produces winner-take-most dynamics. People buy from the best, and the internet ensures the best can be accessed from anywhere. Notion, Figma, Midjourney and dozens of category leaders are running global businesses on a fraction of the headcount their 1990s equivalents required. Midjourney was reportedly under 200 people when it crossed $500M in annual revenue. Solo-founder economics do not reverse this dynamic, they accelerate it, because the winner does not even need to hire to scale.

The honest version of where this is heading is concentration at a level we have never seen before. Category winners in each digital vertical, individuals or tiny teams, capture what used to be the revenue of 100-person companies. The long tail is zombies. The Gini coefficient inside the economy gets worse, the concentration shifts from corporations that employ thousands to individuals and tiny teams that employ nobody.

That is the part the Speedrun-style pitches gloss over. They are technically correct that solo founders can build huge companies. They are wrong that this is good news for anyone other than the solo founders themselves and their investors.

The terminal logic of capital without labour

Marx made a prediction in the 19th century that has been mocked, dismissed and selectively forgotten for 150 years, and is now arriving on schedule. He argued that capitalism would automate itself out of existence. Capital, in pursuit of profit, would progressively replace labour with machines. Since in his framework profit derives from extracting surplus value from labour, the displacement of labour by machines would collapse the rate of profit. The system would reach a contradiction it could not resolve, and the question of who owns the machines would become the only political question left.

You can disagree with the labour theory of value and still take the structural prediction seriously. What we are watching with AI is the long-delayed arrival of exactly this dynamic. The infrastructure is being built by private capital. The end-state it points toward is one where private capital has automated away the thing that justified its own existence. As Marx said, the question of ownership becomes the only political question left.

Elon Musk has been gesturing at this on podcasts for years. He says AI will make everything free, that we will need universal basic income, that humans will need neural interfaces to keep up. He calls it a socialist utopia. He does not notice that the three claims point in three different directions and cannot all be true at the same time. “Everything will be free” is Marx’s post-scarcity end-state. “We will need UBI” assumes scarcity persists. “Neural interfaces” assumes humans will be augmented enough to remain economically relevant. Three different futures, treated as one, because his actual point is that the present arrangement is temporary and we are heading somewhere post-capitalist.

He is half-right and wrong in an interesting way. Right that the present arrangement is temporary. Wrong about where we are heading, because his framing assumes a clean transition from current capitalism through a transitional period and into a stable post-scarcity equilibrium where everyone is fine. The actual trajectory is messier and probably much darker, and the people who will end up controlling the outcome have very different incentives than Musk’s optimistic gloss suggests.

Two states and a fork

If you take the trajectory seriously, you can describe it as moving through two distinct states with a critical fork between them.

State A is post-scarcity. Abundance becomes so total that money loses meaning, the way air does not have a price. This is the Iain M. Banks Culture vision. It requires AI, limitless energy, frictionless matter manipulation and a level of resource liberation we have no current path toward. A goal, perhaps, on a timescale of centuries.

Then there is State B, consolidation. A small number of entities, corporate, state or hybrid, control the means of intelligence production and the productive infrastructure built on top of it. Everyone else is dependent on their pricing, their access policies and their willingness to share. This is much closer to where we are heading in the next 30 to 50 years.

A and B are sometimes treated as alternatives. They are not. A is the goal. B is the intermediate state we have to pass through to get there. The infrastructure required for A (abundant clean energy, autonomous production, distributed material synthesis) is being built right now, by entities operating in B economics. The transition runs from current capitalism into B, and then forks.

The fork at B is the political question that determines everything downstream. The fork has several branches.

One branch is techno-Marxism, where the productive infrastructure becomes collective in some form (sovereign wealth funds with universal equity, state ownership of compute, citizen dividends from automated production, some mechanism by which the wealth generated by the machines is redistributed rather than concentrated). The optimistic version of the Musk pitch, except it requires political action nobody in power currently has any incentive to take.

Then techno-feudalism, where the consolidation hardens into a permanent structure of AI-owning aristocrats and a dependent population maintained on managed welfare. Yanis Varoufakis wrote a whole book in 2023 calling it this, and the term is the right one. The system is feudalism with better surveillance and better PR.

A dark-age reset is another path, where the consolidation breaks down through war, climate disaster, infrastructure collapse, or deliberate rejection of the technology, the Dune scenario, where humanity bans AI entirely after a period of subjugation and spends 10,000 years developing alternative paths. The probability is non-trivial over long enough horizons.

A fourth path, undersold in most discussions, is persistent techno-feudalism as a stable equilibrium. Most feudal systems in history did not end through revolution. They ended through slow mercantile displacement over centuries. If techno-feudalism has AI surveillance, predictive policing, managed welfare and infrastructure dependencies deep enough that disruption equals starvation, the conditions for uprising become much harder than historical precedent suggests. Orwell’s actual insight in 1984 was that totalitarianism with sufficient technological depth becomes self-sustaining in a way pre-modern tyrannies never could.

Which branch we end up on, and for how long, is the central political question of the next century. And the people building the infrastructure are not the people who will answer it.

The fractal pattern

Nassim Taleb is useful here, because his work on fragility, fractals and antifragility tracks the structural problem.

Taleb argues that complex systems do not move toward equilibrium. They move through patterns that look the same at different scales, fractals. Booms and busts, consolidations and uprisings, growth and collapse. The pattern repeats at the scale of a quarter, a year, a decade, a century, a millennium. The frequency and amplitude change, the shape does not. He is sceptical, rightly, of any model that predicts permanent stability, because no complex system in history has ever achieved it.

The Greeks had a name for this. Polybius described anacyclosis, the cycling of political forms through monarchy to tyranny to aristocracy to oligarchy to democracy to mob rule and back to monarchy. Ibn Khaldun, the 14th-century Tunisian scholar, described the same pattern in terms of asabiyyah, the social cohesion that builds in tribes on the periphery, allows them to conquer the centre, then dissipates over 3 or 4 generations until they are themselves conquered by the next wave of cohesive outsiders. Strauss and Howe wrote about generational cycles in American history. The Chinese dynastic cycle ran for 4,000 years on the same pattern.

What these models share is the claim that consolidation is never permanent. Power accumulates, the conditions that produced the accumulation rot, an opening appears, and either internal reformers or external challengers exploit the opening. The specifics vary enormously. The pattern does not.

If you accept this framing (and the historical evidence is overwhelming), then techno-feudalism is a phase rather than a terminal equilibrium. How long the phase lasts and what comes after is the question.

The case for a long phase is that AI consolidation has tools previous consolidations did not. Surveillance is much better. Population management runs deeper. The dependencies are more total. A peasant in 14th-century France could grow their own food and escape into a forest. The citizen of techno-feudal Singapore in 2080 cannot grow synthetic protein at home, cannot escape the surveillance network, cannot function economically without access to the AI infrastructure that is centrally controlled.

The case for a shorter phase is statistical. Given any non-zero probability per unit time of uprising, disaster, ruling-class fracture, or technological diffusion that breaks the consolidation, and given long enough time, the probability of at least one successful disruption approaches one. This is the mathematics of repeated trials at civilisational scale. Defenders of the consolidation need to be right every time. Challengers only need to be right once. Over centuries, the asymmetry compounds.

The longest-lasting empires in history managed 3 to 4 thousand years with continuous internal resets. Egypt, China, Rome counting Byzantium. None of them achieved permanent stability. They achieved long durations of quasi-equilibrium punctuated by partial collapses, dynastic changes, periods of fragmentation and eventual replacement. There is no precedent for a 10,000-year stable empire, and assuming AI changes this requires extraordinary evidence that nobody has produced.

So the long-run picture is probably this. Techno-feudal consolidation, lasting decades or centuries, with periodic minor resets that get absorbed and increase systemic pressure. Eventually a major reset, triggered by some combination of disaster and ingenuity, that breaks the consolidation. After it, either techno-Marxism (if the reset is orderly enough to capture rather than destroy the productive base), or dark-age regression (if the reset destroys the infrastructure that everyone now depends on). Some time later, a new consolidation under whatever technology and ideology emerges. The cycle repeats.

The benevolence problem

In any consolidated system, the welfare of the population depends on the benevolence of whoever holds the consolidation. This is true whether the system is techno-Marxist, techno-feudalist or any hybrid. The structural question of who decides does not go away just because the system has changed names.

In a techno-Marxist state, the welfare of citizens hinges on the benevolence of whatever entity (a state, a constitution, an AI governance system) administers the redistribution. The system can be designed beautifully, with checks and balances and democratic accountability, and if the administering entity becomes corrupt or captured or merely incompetent, the welfare of the population is what suffers. Every 20th-century socialist state demonstrated this in different ways. The system promised collective welfare and delivered the welfare that the controlling class chose to allow.

Under techno-feudalism, welfare runs through the AI-owning aristocrats. Each feudal lord, in their domain, decides how much to share. Some will be generous out of conscience, conviction or self-interest. Others will be cruel out of indifference or ideology. The population has no structural leverage because the labour they used to provide is no longer necessary. They are recipients of whatever the lord chooses to grant, and the lord’s choices are constrained only by their own values and the residual threat of uprising, which, as we saw, may be weaker than historically.

Even in the dark-age reset scenario, welfare depends on benevolence. The post-collapse society reorganises into smaller units, and the units are still hierarchical. Local strongmen, religious leaders, warlords, surviving institutions, each becomes a source of either protection or exploitation. Fragmentation produces a multiplication of small consolidations, each with its own benevolence problem, instead of freedom. Medieval Europe was a place where the question of who treated their peasants well was answered locally and variably, where most peasants got whatever local arrangement they happened to be born into.

The deepest point, the one I think the debate is missing, is that the benevolence problem does not have a structural solution. You can design institutions that constrain the worst impulses of those in power. You cannot design institutions that eliminate the dependence on their benevolence entirely. Every system, at the limit, relies on the character and choices of the people who hold its commanding positions.

Historically, this dependence was mitigated by the structural necessity of labour. The ruling class had to be at least minimally benevolent because they needed the ruled. Feudal lords needed peasants to farm. Industrial capitalists ran on the labour of workers operating machinery. Slave societies kept their slaves alive enough to work. The welfare floor was set by what the rulers structurally required from the population, not by their good intentions.

What is new about the AI transition is that the structural necessity is dissolving. The ruling class of the techno-feudal future needs almost nothing from the population. Drones replace soldiers. AI does the work workers used to do. Intra-class transactions among a small wealthy population substitute for consumer markets. The welfare floor collapses to whatever the ruling class chooses to grant out of sentiment, ideology, or fear of remaining uprising capacity.

This is the new and dark thing about where we are heading. Every prior consolidation in history had a structural lower bound on how badly the rulers could treat the ruled, because the rulers depended on them. AI removes the dependency. The worst-case trajectory is consolidation with no welfare floor at all.

The reliability of benevolence across generations

Even if we get lucky and the first generation of techno-feudal rulers (or the first generation of techno-Marxist administrators) happens to be benevolent, the historical record on intergenerational benevolence is brutal.

Marcus Aurelius was followed by Commodus. Lee Kuan Yew handed off to competent but progressively less visionary successors. The Antonine succession worked for about 80 years before reverting to dynastic mediocrity. Every successful experiment in benevolent rule has collapsed within a few generations because the selection mechanism for the next generation of rulers is corrupted by the previous generation’s success. Power attracts adverse selection. The Marxist version of this is that the vanguard party becomes a nomenklatura. The feudal version is that the second-generation lord did not earn the position and lacks the character that built it.

In a techno-feudal future, you should assume that AI-augmented hereditary class capture happens within 2 to 3 generations. The founders may be brilliant operators who built their wealth through actual creation. Their grandchildren will be courtiers who inherited it and whose only skill is maintaining their position. Wealth becomes inheritable rather than achievable, and the courtiers follow. The moral framing is beside the point.

The question of benevolence cannot be answered by pointing to currently benevolent founders. The question is whether their good intentions will be preserved across the generations of successors who will actually control the infrastructure when it matures. The historical answer is no. Specific institutional arrangements to ensure such preservation (constitutional constraints, electoral accountability, free press, independent judiciary) would likely be dismantled by techno-feudal systems, or, like most institutions, warp and degenerate over time.

In a techno-Marxist system, the same question shows up with different specifics. Can the administering entity remain benevolent across generations? Soviet history suggests no. The Chinese Communist Party trajectory remains to be seen. The honest answer is that we do not know. Unfortunately, like what Taleb suggests in Black Swan, and is conventional VC wisdom: you only need to be right once. In this case, you could have a slew of benevolent leaders, and it only takes one Nero to turn everything on its head. And in the Monte Carlo simulation of possibilities, the inevitability is like playing Russian roulette.

The cognitive capacity to be benevolent

The natural next step would then be to create an AI singularity which would become the ruler, like a Demerzel in Foundation. Such an entity could in principle govern much larger populations with greater individualised attention. This is the techno-Marxist hope, that AI solves the information problem that doomed 20th-century socialism. Plausible in theory. In practice, would such a singularity stay forever benevolent? What about edge cases? And then what does that entail for humanity - some sort of utopian slavery?

The capacity for benevolence and the capacity for benevolent-seeming control are two different capacities. The second is much easier to build than the first, and the second is what we should expect to see deployed.

The probability of reset over deep time

Despite all of this, the most likely long-run outcome is still that the fractal pattern continues. Here is why.

Given any non-zero probability per unit time of disaster plus human ingenuity coinciding, and given infinite time, the probability of at least one successful uprising approaches one. The techno-feudal consolidation case implicitly relies on driving the failure probability to exactly zero. A very low number does not save it. Driving the probability to exactly zero requires controlling current threats and the entire space of possible future threats, including ones that do not yet exist. Every prior empire that tried this (Qin Shi Huang’s book burnings, the Catholic Church’s index of prohibited books, the Soviet information control apparatus) eventually failed because they could not enumerate all the threats in advance. The space of possible exploits is larger than the space of known exploits. The gap is where ingenuity operates.

The consolidation case also has to hold against adversarial co-evolution over centuries. Defenders need to be right every time. Attackers only need to be right once. This is the classic security asymmetry, and it applies at civilisational scale. Even if AI gives defenders enormous advantages, those defenders are operating with bounded rationality, finite computational resources and human-level strategic horizons. Attackers, distributed across centuries and operating in the gaps, have unlimited attempts.

The reset, when it comes, will probably not produce techno-Marxism directly. The reset that destroys consolidated power also tends to destroy the productive infrastructure that consolidation built. Most consolidated systems in history have broken catastrophically rather than gracefully. The French Revolution preserved enough of the underlying productive capacity to produce a more sophisticated political form. The fall of Rome did not, and produced regression. Which one we get from the eventual techno-feudal reset depends on factors we cannot predict: the orderliness of the uprising, the survival of knowledge, who ends up holding the productive infrastructure.

The most probable path is repeated oscillation across deep time. Techno-feudal consolidation, partial resets that get absorbed and build pressure, an eventual major reset that destroys most of the productive base, a dark-age period of indeterminate length, slow rebuild from preserved fragments, a new consolidation under different technology with different mistakes. Across millennia, the pattern repeats. Some cycles might produce techno-Marxist intermediate states that last for a few generations before degenerating. Others might collapse into prolonged dark ages. Average human welfare across the cycle is probably substantially lower than either peak.

Uncomfortable to write down. The most honest reading of the historical pattern projected forward looks like this. We may have spent the last 200 years in an anomalous period of sustained progress, and may now be returning to the long-run pattern, with AI-shaped peaks and troughs instead of agricultural ones.

The first-principle question

The first-principle question, underneath all the framing about solo unicorns and techno-Marxism and fractal cycles, is human welfare. The question is not GDP growth, productivity metrics, aggregate output, or any of the proxies that economists and venture capitalists prefer because they are measurable. The basic question is whether the people alive in any given period have lives worth living, with adequate material conditions, meaningful work or its substitute, social belonging, and the dignity of being something other than dependants on someone else’s discretion.

In every scenario we have walked through, the answer depends on benevolence. Techno-Marxist administrators (or an AI singularity) in their phase. Techno-feudal lords in theirs. Warlords, religious authorities and local strongmen in any dark-age reset. The founders of any future consolidation that emerges from any future cycle.

Even in the dark ages, the regression is to feudalism, because feudalism is the natural human social structure when productive surplus exists but central coordination has broken down. Local lords protect local populations in exchange for tribute. Some are kind, others cruel. The peasants under them depend on which lord they happened to be born under, which lord happens to be in power that year, and what treatment fits the relationship between them. This is the equilibrium human societies fall back to when more sophisticated arrangements break, and the equilibrium AI consolidation is leading us toward in a more sophisticated form.

The deep point is that every social arrangement humans have ever built (kingdoms, republics, empires, communes, theocracies, corporations, states) has come down to the question of whether the people in power chose to treat the people without power well. The institutions we build sit around that question. They make it harder for power to be exercised badly, easier for the powerless to push back, more visible when abuses occur. They do not eliminate the underlying dependence. They shape how it expresses itself.

In the long run, across the fractal cycles, what determines human welfare is whether the people who happen to be in power, at any given moment, in any given system, choose to be benevolent. That is the only question that matters, and it gets the least serious attention from the people building the future.

Historically, the answer has been mixed. Sometimes yes, sometimes no, almost never reliably across multiple generations of the same system. The best we have ever achieved is institutional arrangements that constrain the worst impulses of power and give the powerless enough leverage to push back when constraint fails. Those arrangements are now starting to come apart, both naturally (as it seems with the course of time) and from the pressure of technologies that may make them obsolete.

I had breakfast with a friend who is leaving for Melbourne to start his PhD. He spent most of it on his old boss, the one running the university lab he is leaving behind. The boss had, it seems, over the years simply stopped caring, about my friend and about the research itself, which had become a line item for him - he had probably lost the narrative at some point and become jaded. Yet my friend cares about that work in a way his boss lost years ago. He ran his secret experiments on the side and landed a breakthrough that let the research progress with Cambridge. Now he is going to Melbourne to go deeper, to specialise harder, to become one of the few people alive who truly understand his slice of the field.

Yet this whole situation reminds me of a book I am reading at the moment, The Wealth of Nations, where I had recently gone past the passage on the division of labour. I realised the mechanism that makes specialisation so powerful is the same one that turns a hyper-specialist into a sitting duck.

What Smith actually saw

Adam Smith opens The Wealth of Nations in 1776 with a pin. Making one is split into about 18 operations: 1 man draws the wire, another straightens it, a 3rd cuts, a 4th points, a 5th grinds the top for the head. A single untrained worker doing every step might manage 1 pin a day, maybe 20. Splitting the labour across 10 workers produced, by Smith’s count, upwards of 48,000 a day, around 4,800 each. The same people, hundreds of times more productive, not by working harder but by narrowing what each one does.

That is the engine of modern prosperity. The factories, the supply chains, the global economy that puts Ethiopian coffee in my cup in Singapore while I type on hardware designed in California and assembled in Shenzhen, all of it sits on that one move. Specialisation works. It is the most powerful productivity idea our species has found, and anyone selling you on its opposite is selling something.

Hundreds of pages later, in Book V, Smith turns and warns what the pin factory does to the man pointing the pins. Someone whose whole life is a few simple operations, he wrote, “has no occasion to exert his understanding“ and becomes “as stupid and ignorant as it is possible for a human creature to become.“ The thing that makes the worker productive also narrows him. He sharpens on the slice and dulls everywhere else, including the capacity to notice when the world around the slice has moved. We have heeded the advice of the former and yet ignored the dangers of the latter.

Good, then hyper

Specialisation is fine, even now. Hyper-specialisation is the same instinct pushed past the point where it helps, and that is what is biting back today. Done right, specialisation is still worth having, as long as you stay general enough alongside it. Deep knowledge of a domain is exactly what lets you work out how to implement AI inside it, assuming the domain is not thrown out wholesale by AI first.

The line between them is how much of your identity, your income and your optionality rides on a single, narrow, fixed definition of what you do. A cardiologist is specialised. A cardiologist who can read only one kind of scan on one machine, and who is unemployable the day that machine is retired, is hyper-specialised. One went deep on something durable. The other went deep on something contingent and had a false sense of security.

The education system is built to manufacture the second kind, and grows more aggressive about it the higher you climb. Primary school is broad on purpose. Secondary school narrows. University makes you pick a major and shed the rest. By the postgraduate level the funnel is a needle, and the reward structure of academia is to know more and more about less and less until perhaps 11 people on earth can judge whether you are any good. The highest status in the system, the doctorate, goes to exactly this. To specialise has become a synonym for advancement.

For decades it was a sound bet because fields kept moving and the credential was a moat. You spent 15 years becoming the world expert on something and were paid a premium for 30 more, because nobody else could do it and it still mattered. Underneath sat 2 assumptions: that the field would keep mattering, and that being the scarce human who could do the work was the same as being valuable. Now it seems both assumptions are breaking at once.

The part where skin in the game disappears

Worse still is hyper-specialisation with no skin in the game. In a state-funded university lab, the incentive thins out when all you have to do is coast and collect a monthly salary. As the 2 famous adages go: “Show me the incentive and I will show you the outcome,” and “The three most harmful addictions are heroin, carbohydrates, and a monthly salary.” Whether the research moved anything in the world had become beside the point. My friend had skin in the game in the sense that counts, his meaning was on the line. His boss was playing a different game, academic survival, so they had drifted so far apart they could share a room without sharing a job.

Peter Thiel likes to say PhDs today are something like 99% less productive than they were a century ago. Argue with the number, it is deliberately provocative, but the core is hard to dodge: we have more credentialed researchers than ever and the rate of foundational breakthrough has, by most honest measures, slowed. We mistook credential proliferation for progress, built an enormous machine for minting specialists and aimed it at fields where the easy gains were already taken, then acted surprised when the output was more papers and fewer revolutions.

Skin in the game matters here because it is what tells you when to stop. A specialist with real stakes feels the field going stale, because livelihood and pride are wired to the result, and the pain pushes them to move and recombine and question the premise. A specialist insulated by tenure, grants and institutional momentum feels none of that, and keeps pointing pins long after the world stopped buying pins. The insulation that feels like safety is what stops you noticing in time.

The irony we are all living inside

Set all of that against early 2026. Anthropic published a chart, since gone viral, on which fields are most likely to be disrupted. Its Economic Index, the data behind that chart, tracks which kinds of work its models are actually used for and how much of each occupation’s task load they can cover. The heaviest overlap sits in the high-credential knowledge work. It skews toward higher education, clustering around tasks that need roughly an associate degree’s worth of schooling and above, ahead of the economy’s average. Computer programming, where the model can touch most of the tasks. Customer service, data work, analysis, the white-collar knowledge work the education system sold as the safe, smart, future-proof choice.

Ironically, we built the most elaborate specialist-making machine in history and told a few generations that security meant narrowing yourself, credentialing yourself, becoming the indispensable expert in a single defined area. The technology’s first move targets exactly that kind of narrow, credentialed work. The cleaner and more bounded your expertise, the cleaner the target. A model is extraordinarily good at precisely the well-specified, deep-but-narrow task a hyper-specialist spent 15 years perfecting.

Marc Andreessen made the sharper version of this point on a recent Joe Rogan episode: across almost any topic, the best models now give him a better answer than 99% of the human experts he could otherwise call. It has not beaten the very best human at the top, at least not yet, and it cannot hold the messy real-world bundle that a whole job actually is. Yet on the clean isolated skill, the thing you could put on a test, it has passed almost everyone. Thiel’s 99% was about PhDs being less productive than their grandparents. There is a second 99% now: the share of practitioners in a speciality a model can match or beat on the core task. We made ourselves narrow for the premium that scarcity pays, yet scarcity is the thing that just evaporated.

What is actually scarce now

If the narrow skill is no longer the moat, what is?

I made the case a while back, in an essay on learning to fix an air-conditioner, that generalists with systems thinking inherit what comes next. What the breakfast and the book sharpened is why, and which capabilities actually hold value.

First, knowing the lay of the land. Not depth in one spot but a working map of how the territory connects. Someone who grasps, even roughly, how product meets finance meets operations meets distribution, and can locate where a problem actually sits, is doing something models still find hard, because it means holding a live, messy, organisation-specific picture of reality that exists in no training set.

Second, systems thinking, which is that map in motion. Anyone can learn a fact. Fewer can see how a change in one corner ripples through 5 others, where the second-order effects land, what breaks when you pull a thread. This is what lets you wire a model’s narrow competence into something that works with agentic (or simple operational) architecture, and it is what the education system trained out of us by rewarding finer specialisation over the unglamorous work of connection.

Third, relationships, the most underrated and most durable of the 3. When technical work commoditises, the premium moves to what it was never really about: the trust you have built and the people who answer when you call. That web compounds over decades and cannot be spun up in a weekend by anyone, human or model.

None of the 3 is a speciality. All 3 are what link specialities together. We spent a century treating that as the soft stuff you pick up on the side while the prize went to depth. The ranking has flipped, and most people have not noticed, because the institutions that would tell them are the same ones built to sell depth.

How to specialise without becoming a sitting duck

Hyper-specialisation is still where the action is in any field that is genuinely moving. At a live frontier mid-breakthrough, going as deep as humanly possible is correct, because depth is how the breakthroughs happen and the payoff is enormous. Every field has a life cycle. Early, the frontier is open and depth pays spectacularly, because each increment of specialisation opens real new ground. Later, the easy gains are gone, arguably now much sooner than later, the field goes flat and depth stops having any valuable impact. The danger is specialising into the late stage without looking up to see what has happened to the industry and then, through the very insulation that depth breeds, failing to notice it has stopped moving. You spent 15 years world-class at pointing pins, the world quietly stopped needing pins and the narrowness that made you excellent is now what keeps you stuck.

So the durable move is to go deep with your eyes open. Specialise if you want to, but keep a live read on where your field sits in its own cycle. Keep one foot in the connections, the lay of the land, the systems view, the relationships, so that when the field tips you can move in time. Hold your speciality as a position you can exit, and keep the optionality to recombine what you know into whatever the world still wants. Being part of a system that builds in skin in the game (as opposed to some systems where your skin is never in the game) gives you an early signal when demand falls off and makes the plateau hurt early enough to act on rather than late enough to be fatal.

The machine that made us rich by narrowing us is now being out-narrowed by something that never tires and does not go stale. With AI in the picture, the playbook is always changing, and it seems at this short moment the playbook looks like this: go deep enough to matter, stay broad enough to move and keep enough skin in the game to feel the ground shift while there is still time to step off.

When I was twelve, maybe thirteen, I remember I had an extensive conversation with my father about Singapore’s housing geography. At that point Punggol was just starting to become a new estate. A lot of younger families were being moved out there. The catch was that getting from Punggol into the CBD in those days could take close to an hour in peak-hour traffic. The very people who needed to be near the city for work were the ones being sent furthest from it. My family was thinking about moving to be closer to my school, and I had been turning the larger problem over in my head.

I proposed something like this. Why not move all the retired aunties and uncles out of Toa Payoh and Bishan, send them to Punggol, and let the working-age population live where the trains went straight to Raffles Place. The retirees would get more space and more greenery, which is what people in their sixties always seem to say they want. The young people would get back two hours of their day. A win-win.

My father laughed at the prospect and found it not very PC but he said I had a point, though it was slightly insane. But underneath the bad framing I came to realise that there was real intuition. Proximity is productivity. Where you put people in relation to where they need to be decides a surprising amount about how their lives go.

The Pin Factory and Your Living Room

Adam Smith’s most famous example in The Wealth of Nations is the pin factory. He observed that when you broke pin-making into stages, with each worker handling a single operation, output multiplied many times over what one worker could produce alone. Most readers stop at the punchline about division of labour. The part that gets glossed over is the physical layout of a factory. The reason division of labour works is because the pieces of work move short distances between stations. If the wire-drawer had to walk down a flight of stairs to hand the wire to the cutter, and the cutter had to walk to a different building to hand it to the pointer, the throughput would be severely affected. This is why every modern factory floor is designed obsessively around flow. The factory is fundamentally a geometry problem.

A city is the same problem at a different scale. The country itself is a kind of very large factory. People are not products moving down a line, but their time is. Every minute spent in transit is a minute extracted from either productive work or rest, both of which compound into the quality of the economy and the quality of the lives inside it.

What Singapore Got Right

Most Singaporeans do not think of high-rise living as a design choice. It is just the default for a “land-scarce city-state”. But it is one of the more deliberate urban planning decisions in the modern world and it solves the geometry problem at scale.

You stack people vertically so that you can keep them horizontally close to everything they need. A neighbourhood of twenty-storey blocks holds enough population in a small enough radius that a hawker centre, a clinic, a wet market, a school, a barber, a coffee shop, a hardware store, and a couple of banks can all sustain themselves within walking distance. The density makes the amenity viable. The viability of the amenity is what gives people back the hours that would otherwise be spent commuting to a less efficient version of the same errand.

China has taken this to further extremes as they usually do. In most Tier 1 and Tier 2 Chinese cities, you have mixed-use developments, where the ground floors and lower podiums are stuffed with shops and services and the upper floors are apartments. You can come home, eat, get a haircut, see a dentist, and pick up groceries without leaving a two-hundred-metre radius. Whatever one thinks about Chinese governance, the urban form is doing a lot of work for productivity.

The Western Inversion

The Western city is built on top of an old human-scale core. Manhattan, downtown Austin, central Melbourne, Florence, Amsterdam, and most older Western cities were laid out before the car. The streets are narrow. The buildings are short, partly because the materials of the time could not really hold weight beyond a few storeys. Parking is hard. Infrastructure under the ground is old and was not designed for the loads a modern city puts on it. These cores are pleasant to walk through. You can cross Melbourne’s central business district on foot in half a day, which I did the first time I visited.

The problem is that a human-scale core cannot hold a modern population. Once a city grows past what those streets can absorb, the only option is to expand outward. That is what most Western cities did. The old core stayed beautiful and walkable. The actual population went into suburbs that ringed it. Austin works like this. The downtown is small and dense, but the people who work in it largely do not live in it. They live half an hour out, or more if there is traffic, in detached houses on plots of land that are not really walking distance from anything except more houses. The American dream is the landed house, the lawn, the two-car garage. It is a beautiful idea. It is also, in aggregate, an enormously expensive piece of urban engineering, paid for in hours of life spent in commute. The open front lawn is supposed to be a communal gesture, but the sparseness it requires is part of what forces everyone into cars in the first place.

Ivan Zhao, the founder of Notion, was talking recently with Reid Hoffman on Reid’s podcast Possible about human-scale cities versus the cities we have built for cars. He framed the question by referencing Renaissance Florence and the design of cities. A human-scale city is wonderful to inhabit at low population density. It is also, at scale, exactly what produces car dependency. You cannot fit a working modern economy inside Florence. You can fit it around Florence, which is the deal most Western cities ended up making. The Asian leap-frog alternative was to abandon the human-scale core and build dense vertical neighbourhoods connected by trains and public transport so that the economic capacity per square kilometre is much higher.

Something I find funny and ironic is that landed houses in Singapore are almost universally fenced even though crime is negligible. Houses in many parts of the American suburbs, where crime is more present, often have no fence at all.

The Math of Density

There is a concept of economies of scale even in this urban planning geometry. A township of 1,000 people needs a dentist. It needs a grocery store. It needs a school. The problem is that a thousand people often cannot quite support all of these things at a level that lets them be run well. The dentist sees few enough patients that pricing is high and quality is uneven. The grocery store carries less variety because the volume is not there. Eventually some of these services either close or were never built, and the residents drive further to the next town for what they cannot get locally. That adds more car time, which makes the place even less self-contained, which deepens the dependence on the car.

Stack the same township into 20-storey blocks and you have 20,000 people in roughly the same square kilometre. Suddenly there are five dentists who compete on price and quality. The grocery store has the volume to carry both basics and specialty items. A second school becomes viable. A small theatre might open. The infrastructure cost per resident drops because the roads, pipes, and cables are shorter per household. The whole place gets richer in the small daily ways that matter more than people admit.

Urban planning, when you look at it from this angle, is a question about whether the average person’s day is going to be spent on things that matter to them or on getting from one place to another.

Commuting Is Terrible for Happiness

The version of a working life I grew up watching in Singapore is roughly this. You wake up at a reasonable hour. You take a train that runs on time. You get to your office in twenty to forty minutes. You work. You come back the same way. There is no car payment, no petrol bill, no insurance, no parking, no driving stress. You arrive home with energy left for your family, your hobbies, your second projects.

When the geography is more sprawled, you wake up earlier than you want to. You drive your kids to school because there is no public option that makes sense. You sit in traffic. You arrive at work already partly used up. You repeat it in reverse in the evening, often in worse traffic. The car costs you a real fraction of your take-home pay. The time costs you everything else.

Both groups, on paper, work the same hours. But one of them has paid a real cost to the urban form before they reach their first meeting of the day. That cost shows up in marriages, in patience with children, in willingness to take risks, in the energy left over to start a business. It shows up in the macro figures eventually, but it shows up in lives first.

The research lines up with this. In a 2008 study published in the Scandinavian Journal of Economics found that people with longer commutes reported systematically lower subjective well-being. In the study, a worker with a 1 hour one-way commute would need to earn roughly 40% more to be as satisfied with life as someone who lived close to work. They called this the commuting paradox, since the housing and job benefits people thought they were getting by living further out did not actually show up in their reported satisfaction. Daniel Kahneman’s separate work using the day reconstruction method found that commuting consistently ranks among the most negatively-rated activities of a person’s day.

It is tempting to frame this as a quality-of-life argument. It is more than that. Smith’s pin factory loses its gains if the workers have to walk too far between stations. A country loses its gains if its workers have to drive too far between home and work. The geometry of the country and the geometry of the factory floor are doing the same job, at different scales, with different units.

What AI Changes, and What It Does Not

With AI, people can be productive on the commute. They can dictate emails to an agent while sitting in traffic. The car becomes a moving office. Some of this is true at the margins. It does not change the underlying geometry. An hour in a car with an AI is still an hour you did not spend with your children or your own thoughts. The friction is reduced but not removed.

If AI, however, absorbs a meaningful share of labour, then the commute matters less because fewer people have to make it. The factory analogy starts to break, because the factory is partly automating away its own workers. Cities then become less about getting people to jobs and more about getting people to each other. Density stops being a productivity question and becomes a meaning question. What do you build a city for when its main user is no longer a commuter?

If labour is what gives a human life its shape, which I largely think it is, then the urban form should keep doing what it has always done, which is to put people close to the things they need to do. If labour is not central anymore, then the urban form has to be redesigned around something we have not really had to design around at scale before.

Maybe you won’t just have a Lawson covering your view of Mt Fuji.

Shulin’s CNA interview lit a fuse, and the hunger piece I wrote after left some people bitter that they flamed me as well. I found myself having a chat with a Gen Z friend this week and she sits in the camp that tilted when Shulin’s podcast came out. Her view, condensed: work-life balance matters, and millennials do not understand it.

I get where she’s coming from. I too follow and believe in some of the philosophy about being present and living your best life with utmost agency. The instinct to defend your evenings, weekends, slow Sunday breakfasts, and unstructured time is a perfectly healthy human response to a system that wants to take more of you.

Yet if you speak to anyone on the other side of the same conversation who are actually successful in what they do. Founders who actually built something, partners who made it, athletes and writers and anyone in any field who shipped real work. Their answer is almost always the same: There is no balance, only obsession.

So which is it? I think it is useful to ask how we got here.

The world that used to end at six

During our parents and grandparents time, at least the employed ones, work was a place. You took the bus or drove to it, you sat at a desk for eight hours, and when you stood up to leave, the work stayed there. Life began after you leave the office door.

My father worked at an FMCG company. He ended work at 6pm sharp and drove to school to fetch me from CCA. That was the cadence. There was no laptop waiting for him at home, no late evening call from a regional counterpart in a different time zone, no Slack notifications buzzing through dinner. The job was done when the day was done.

The world moved at a different speed too. Technological change came in 2-5 year intervals. A sales cycle could be measured in seasons. You could actually write a yearly budget and have it survive contact with reality by year-end. Strategy decks did not need to be redone every six weeks because the underlying world rarely shifted in six weeks. The annual review and bonus cycle were not corporate rituals dressed up as discipline. They matched the actual cadence of the business. Work-life balance was the default condition of a slower economy.

Then smart phone and the laptop arrived. Work followed you home, then to dinner, then into bed. The work hours got worse, with informal expectations of being reachable on WhatsApp after dinner and emails landing after working hours.

And everything sped up. And now with AI, everything moves in the matter of days, not years. The annual budget became quarterly. Quarterly became monthly forecasts because the market reacted faster than a quarter. By now, in many industries, even a monthly forecast is too slow, and you are running rolling forecasts because the underlying business genuinely moves that fast. Yet the working hours on paper stayed the same.

So when my friend says millennials do not understand work-life balance, I want to push back. Millennials did understand it. They were the last cohort to taste it as the default.

What obsession buys

If you want to outperform in a workplace where everyone is technically competent, the lever that still reliably works is volume of effort. The person who is genuinely the hardest worker in the room, and who is also smart, will progress faster than the one matched in talent but unwilling to do midnights and weekends. This is just what compounding looks like in a competitive environment.

The founder who shipped six products in a year and had one of them catch is not luckier than the founder who shipped one. They got six rolls of the dice. The same applies in any salaried role. The associate who took every weekend deal was not blessed by the partner. They were in the room six times more often.

Yet there is a cost. People who run that pace pay for it in their bodies, their relationships, and parts of their inner life. The ones who can sustain it are, in a way, strong. They have made a trade and they are paying the bill.

You can’t have your cake and eat it

Where I get impatient is the third position, which I hear most often. It is the person who wants to be paid like the obsessed and live like their grandparents. Same promotion, same comp progression, same seat at the table, while clocking out at 6pm or working from home but barely working and protecting their evenings as sacred. If everyone in your workplace is roughly as effective as you are, that math does not math. You cannot have less output and equal reward in a competitive system.

If you choose the balanced path and the promotion goes to the colleague who is camped at the office, that is the trade. Do not call it a failure of work-life balance culture. It was a choice. It might have been the right choice. It might keep you sane, your marriage alive, your kids actually known to you. Those are real wins, and they are also not free.

The legal scene in Singapore is probably a good model to understand this. Every lawyer at any firm in Singapore is, by definition, a high performer. The only way to break out of the pack is to grind harder than the next high performer. So lawyers go into cutthroat practice in their 20s and early 30s, where the differentiation is hours and intensity. Then, when they want to start a family, they go in-house. The pay drops and their status in the legal scene stagnates until they go back to practice to continue building it again. The predictability goes up and the weekends come back.

This is the environment and system where everyone you are working with is already a high performer. The thing to notice is that industries are converging towards a similar shape. Tech, banking and consulting all have their own version of the same arc. The choice point arrives at different ages depending on the industry, but it arrives.

Why nothing dials back

The obvious objection is, why must everyone move so fast in the first place. The answer is structural rather than cultural.

The economy now runs on capital that has been raised at valuations the underlying businesses cannot yet justify. A SaaS company trading at 50x earnings, a growth company at 200x, a private company priced off a comparable that itself is priced off vibes. Once that capital is in, the business has to deliver growth that retroactively makes the valuation rational. Growth pressure is converted into hiring plans, OKRs, sales quotas, product timelines and ultimately into the calendar of each and every employee.

It is not just speculative tech either. The boring industrial conglomerate trading at 30x is being pushed by an asset manager who needs the holding to perform for their LPs. Even Singapore REITs are facing quarterly distribution pressure from a unitholder base that is hungry for yield growth. The pressure runs through the entire capital structure, including the parts that look stable from the outside.

You cannot opt your company out of this without losing access to capital or customers or both. And you cannot opt the system out without breaking the loop that produces things you secretly (or not so secretly) want. Those inflated valuations are part of the collateral that lets banks underwrite loans, that lets the government issue bonds, that funds the MRT you ride and the covered walkway you stand under when it rains. Cheap eggs at NTUC and Sheng Siong are downstream of the same system.

The choice that is left

So I am not really arguing that obsession is good or that balance is bad. The argument is that what used to be the default is now a choice and choices have costs. Pretending the choice does not exist (and complaining and being entitled) is the worst option.

If you want to get ahead in a world where the cycle has compressed and your colleagues are willing to push, you have to push too. That path pays but it also taxes you personally.

The other path is legitimate too. Take your evenings if you want them. The slow Friday work from home day. The Saturday lunch with your parents that they secretly look forward to all week. The second kid. Those are worth the trade. Just do not be surprised when the promotion goes elsewhere, and do not call that injustice.

Work-life balance was the default condition of a slower world. Now in late stage capitalism, it is now a choice.

We were driving through the countryside hours into a road trip. To keep myself awake, I asked ChatGPT to run me a trivia game. Then, somewhere after increasing the difficulty a couple of times and changing topics a few times, it asked me this:

How has China’s digital surveillance infrastructure shaped global conversations about privacy and tech ethics?

I gave a generic answer. Then ChatGPT replied with: Digital surveillance is a double-edged sword, it said. It ensures crime prevention and public safety, but it infringes on personal privacy and enables state control.

I shot back at it because it is the answer you get from a very western-educated person but it is biased and that bothered me. The “double-edged sword” assumes the two edges are equally weighted. It treats personal privacy and physical safety as two values on a scale, but they cannot be equally weighted, or even tipped scales towards personal privacy. It doesn’t make sense on first principles.

The lineage

The liberal grammar of privacy emerged from a specific argument in a specific century. The problem was the absolutist state. The answer was the autonomous individual.

Hobbes, writing in the wreckage of the English Civil War, where life in the natural condition was solitary, poor, nasty, brutish, and short, so the sovereign existed to put a stop to it. You surrendered freedoms in exchange for survival. Hobbes is uncomfortable reading for liberals because he starts from primary needs and reasons forward to authority.

Locke built the off-ramp a generation later. Natural rights, life and liberty and property, precede government and are not granted by it; government exists to protect them, and its legitimacy hinges on that protection. Mill extended the architecture: the only justification for coercion of an individual is to prevent harm to others.

Rawls is the patron saint of how this lineage gets dressed up as universal reason. His original position asks us to choose principles of justice from behind a veil of ignorance, not knowing our future race, class, gender, talents, or generation. From that vantage, he argued, that liberties come first, Rawls argued, because without them you cannot pursue any conception of the good life.

Rawls assumed rational people would want, above all else, the freedom to make their own life plans. That is the anthropology of someone who already has food, shelter, a working civil order, and a household that will still be there in the morning. Rawls, in my opinion, constructed it blind to the worst case and to most of human history, with a western bias.

A person reasoning honestly from behind it, knowing they might end up anywhere in any era, would start not with free speech but with whether they would be alive next month.

First principles

Here is where the cultural lens conversation usually end up with, that Western individualism prizes the autonomous self, and Asian collectivism prizes the group and the family. I find this exit unsatisfying. If we are going to argue about what governance owes us, we should argue from first principles, not just from the comfort of cultural difference.

The standard liberal argument rests on a single load-bearing axiom: absolute power corrupts absolutely. From this axiom follows the conclusion that mechanisms of correction should be pushed downward, toward the individual, toward the smallest possible unit of sovereignty. Privacy is one expression of this principle. The axiom holds, but the leap to the individual in my opinion, does not.

Pushing the corrective mechanism all the way down has its own failure mode: When every individual is a sovereign actor with their own private sphere, you are not running one experiment with one chance of error. You are running a Monte Carlo simulation with as many trials as there are people. Each private sphere is a coin flip on whether something bad happens inside it. The more thoroughly you privatise agency, the more coin flips you accumulate, and the more bad outcomes follow.

Domestic abuse, child neglect, the slow radicalisation of a young man in his bedroom, the planning of violence, the dissolution of a household, the disappearance of a vulnerable person: these are events that occur precisely where surveillance does not reach. The liberal answer is that the cost is worth paying, because the alternative is a state that knows too much. But the cost is paid in real lives and not just theory.

The liberal will say a few bad outcomes at the individual level are the price of protection against the larger catastrophe of state abuse. Yet it assumes distributed disasters are smaller in aggregate than concentrated disasters. That assumption is doing a lot of work, and it is not always true.

What comes first

So what would a person actually choose first, behind the veil? Free speech, property, and the right to publish grievances all come later. The starving, the stranded, and the assaulted are not theorising about any of them.

The hierarchy, stripped back to the species, is straightforward. I propose that these are the most fundamental primary rights:

1. Survival comes first: food, water, shelter.

2. Access to immediate family: children, parents, the small unit of biological attachment that determines whether a person grows up sane. The Harlow experiments on rhesus monkeys in the 1950s demonstrated that infant primates deprived of maternal contact developed lifelong psychological pathologies even when their nutritional needs were fully met. The need to be near one’s mother, in a primate brain, is a structural requirement, not a preference.

3. Physical safety: the ability to walk down a street and arrive at the other end with your body and your limbs and possessions intact.

These are primary rights that do not vary between cultures.

Free speech, property, mobility, opportunity, the right to influence public decisions: these are secondary rights. They are extraordinary, the privilege of a civilisation that has already solved the primary problems. They matter, but they are downstream. A governance theory that prioritises them ahead of the primary tier is not neutral. It is the political philosophy of people who have had more icing than cake.

In practice

The United States has secured an extraordinary degree of secondary rights and visibly lost the plot on the primary tier. Most readers can produce the inventory themselves: mass shootings, fentanyl, abandoned transit, retail districts shuttered by unprosecuted theft. The country has the most aggressive free speech protections in the developed world and one of the most anxious civilian populations in it. The trade is visible but not described as a trade.

Europe is a softer version of the same drift. Different specifics, same direction. The lived sense of urban safety the postwar generation took as a baseline is no longer the baseline, and the conversation is still conducted in the vocabulary of the 1970s.

China has run the opposite experiment. More than 700 million surveillance cameras, by widely cited estimates. The state knows where you are, what you bought, who you are with and what you said in a group chat. By any Western metric, the system is dystopian. And yet just thirty years ago China was a country with significant violent crime, organised gangs in second-tier cities, kidnapping rings, the kind of public-space lawlessness that was a feature of every developing economy at that stage of urbanisation. The transformation, by any honest accounting, has been remarkable. Cities that were dangerous in the 1990s are now, by the lived experience of their residents, among the safest in the world for routine activities, and the rate of violent street crime is a fraction of what comparable Western cities now report.

A woman can walk through a Chinese city at midnight. People can stand on the sidewalk with their phones out and their valuables in their pockets. Restaurants hold seats with a handbag on a chair. My wife has left her handbag in foodstalls and in DiDi cars and it has come back every time with valuables still inside. This is what the trade looks like from the inside.

The secondary effects are easy to underestimate. Night economies operate freely because nobody is afraid of the dark. Families occupy public spaces at 11pm in the evening. Women travel solo without thinking twice. Political expression has been narrowed, but arguably, that is much less important if physical and economic activity has visibly broadened.

The Benevolent State

Of course, digital surveillance only works functionally in a state with benevolent leadership.

In Singapore, conditions were specific: leadership oriented toward population welfare rather than factional enrichment, a meritocratic bureaucracy that let competence rise, housing policy that gave four in five citizens a real stake in the system, ethnic integration that prevented segregated grievance pools, a judiciary clean enough to enforce contracts and prosecute corruption regardless of political affiliation. Under those conditions, state capacity (including surveillance capacity) is not an infringement on rights but a guarantor of them. It secures the primary tier so the secondary tier has something to stand on.

The objection is that who can guarantee the benevolence? Leaders change, surveillance built for safety gets deployed against dissent, meritocracies ossify into dynasties, and redistributive housing policies get captured by the people who already own. The pessimistic case is that all such systems decay, that LKY was a once-in-a-generation accident, that the Chinese model is one bad succession away from horror and the Singaporean model is one bad cabinet away from drift. The pessimistic case is what Dalio would call the long arc of the cycle and Taleb the fragility of any system that depends on competent succession.

But the optimistic case for the Western model is no longer obvious either. Countries that bet everything on the secondary tier are watching their primary tier degrade in front of them, and they are still describing the trade as if they were holding the high ground. Ironic is the American who lectures the Singaporean about freedom while sending his children to school behind bulletproof glass.

Chamath Palihapitiya made a point on Joe Rogan recently. Love him or hate him, he, a billionaire fund manager made the case against the capitalist system for labor. He said every Western tax code has been built around the same idea: capital should pay less than labor. In the 60s and 70s the rationale was straightforward. Build factories, hire people, take the lower tax rate as the reward for doing useful work. Capital was creating value: factories, infrastructure, jobs. The bargain made sense then.

Yet the bargain stuck around. The “useful work” behind it did not. This gap is also visible in Singapore, which has committed more fully to capital-favoured tax design than almost any developed economy.

Most of what those of us with legal or financial backgrounds know about organising companies, taxes, capital, and labor is layered on top of a single axiom: capital deserves preferential treatment because capital builds. The axiom has been treated as fixed for so long that whole careers and entire industries are built on its consequences. What Chamath pointed out was new to me, that it need not be fixed.

The math

The headline corporate tax rate in Singapore is 17%. A qualifying startup, taking advantage of the Start-Up Tax Exemption, pays an effective rate closer to 6% on its first S$200,000 of profit. The top personal income tax bracket is 24%, applied to chargeable income above S$1 million. There is no capital gains tax, no dividend tax (Singapore runs a one-tier system, where dividends paid out of already-taxed corporate profits are exempt in the hands of shareholders), and no inheritance tax.

A wage earner at the top of the income scale pays 24% on the marginal dollar. That is low by international standards. It is still 24% on every additional dollar of income. A business owner who reinvests profits, pays themselves through dividends, accumulates equity, and eventually passes the lot to their children faces only the 17% the company pays at the corporate level, often less if they have a competent tax adviser. The dividend leaves untaxed, the equity sale exits untaxed, and the assets cross generations untaxed.

The argument for this design is straightforward. Singapore is a small island with no natural resources. What made it work was attracting capital and the people who allocate it. The corporate rate drew HQ relocations. Fund managers came for the absence of capital gains tax. Family offices arrived after estate duty was abolished in 2008. The whole structure says: come here with your capital and you will keep what you earn. It worked spectacularly for the country in aggregate. But cracks have started to show.

Empty shophouses

More than 3,000 F&B outlets closed in Singapore in 2024, the highest in nearly two decades. The first ten months of 2025 produced another 2,400. Prime retail rents rose 3% in 2024 and another half-point in Q1 of 2025. Shophouse rents in Little India and Geylang jumped close to 13% in a single quarter. Flor Patisserie, a decade-old bakery in Siglap, had its rent raised by 57% on renewal. Most of this is trite for any Singaporean reading the news.

This is the wealth transfer Chamath named, in its Singapore version. The landlord pays 0% on the eventual capital gain and 0% on the inheritance to the next generation. Wealth flows in one direction. And you would think that with the free market, the rent would correct itself. Yet Mainland Chinese capitalists, looking for a city outside the grasping reach of the CCP, have started showing up to pay these high rents on consumer and F&B businesses that are sure to lose money (some of them don’t even bother to put bilingual signboards). Losing money does not matter when what they are actually buying is permission to live here long term. Local tenants and old household names take the fall and unfortunately the more institutionalised the landlords are (i.e. REITS), the more they could not give a fuck. That is also why every middle-class Singaporean who can afford it wants to be a property investor. Singaporeans buy property partly because they expect it to appreciate forever, which is an assumption I have written about in another article. The system has largely confirmed it for the people lucky enough to have already entered.

The system rewards entry, and entry is increasingly inherited. The 20th-century boom made one generation of Singaporeans into capital allocators. The current bargain lets them pass that position to their children, who become capital allocators themselves because their parents already are. The rest of the working and middle class stay on the labor side. Being on the labor side in late-stage capitalist Singapore means watching wages rise more slowly than living costs. That is what Chamath was describing in macro terms. In Singapore it shows up as wealth flowing from the young, who are usually middle-class taxpayers, to the old, who usually own the capital, and from the old to their direct descendants. The result is a non-meritocratic inequality.

Singapore’s Gini coefficient climbed from 0.412 in the late 1990s to 0.458 in 2016, above the UN’s 0.4 alert line. During the country’s strongest GDP growth periods, the wage share of national income actually fell. Most of the gains went to capital, not wages.

Singapore washing

The original raison d’être of the corporate tax discount was that businesses create jobs. You get the tax break because you do the social work of hiring, training, paying, and building. The bundle of justifications stacks neatly enough: more inbound businesses, more innovation, more domestic jobs, and a stronger national brand that pulls in more foreign capital. The fourth depends on the first three being real. If they are not, the brand eventually catches up.

Each of the four has weakened. The clearest example is the inbound-business one, and the cleanest recent illustration is Manus.

On 27 April this year, China’s National Development and Reform Commission killed a $2 billion deal. Meta had agreed to acquire Manus, an AI startup that spent the previous summer relocating its headquarters from Beijing to Singapore. Beijing called the whole thing off and ordered the parties to unwind, citing technology, data, and national security sovereignty.

This is what people are now calling Singapore washing. Move the cap table, switch the operating entity, restructure data residency. The Chinese company becomes a Singaporean one in legal substance, and the label gets you preferential treatment in capital markets, exemptions in trade frameworks, and a 17% corporate tax rate.

The pattern is broader than just Manus. Take a foreign business, switch the legal address, claim Singapore status, capture the regulatory and reputational dividends, and run a forty-person office on top of an entity that fundamentally still reports to its origin country. In my professional experience, the slice of inbound capital that looks like this is growing.

The job-creation justification fares no better. The tax break was supposed to compensate businesses for the social work of hiring, training, paying, building. The structure increasingly does not require that work. With AI enabling one-person company architectures, with the ability to incorporate purely to capture the 17%, the bargain is being collected without the work being done. A “one-man unicorn” enjoys the same corporate discount as a 500-person operating company. AI is starting to remove labor completely out of the equation.

The harder problem

The instinct from here is to just raise the rates. Raise corporate tax. Introduce capital gains tax. Close the family-office structures. Move the dial back toward labor. But brute-force fixes do not work.

Imagine the simplest version: corporate taxes go up, personal taxes come down. In theory, more entrepreneurs would prefer salaried employment. In practice, almost everyone who can will incorporate themselves and have the company employ them, capturing the lower personal rate while still using the corporate vehicle. The arbitrage just shifts. The fix cannot be brute force or cosmetic.

The structure underneath the rates is also wider than the corporate income tax line. Raising the 17% would make Singapore less competitive on the headline number while leaving the actual architecture (no capital gains, no dividend, no estate) intact. The wealthy individuals and family offices that came here did not come for the 17 alone. They came for the whole arrangement that favours capital. Touch any one component and the equation that brought them breaks. Singapore as a nation state still relies, fundamentally, on serving rich transnational individuals and multinational companies. The professional and ancillary service industries (legal, accounting, banking, advisory) are the outer ring of that core. Remove the incentives that bring rich people and multinationals here, and Singapore loses the only USP it has built up at scale. The only viable direction, then, is not to drag capital down but to lift labor up to something closer to equilibrium with it.

China itself is another data point. China has higher corporate taxes and harder compliance than Singapore, and it remains a magnet for capital because of the size of its domestic market. Market size, more than tax architecture, holds capital there. Singapore does not have that lever. We are 6 million people on an island.

Narrower fixes do not escape the gaming problem. Differentiating corporate tax by local employment is one of the smarter ideas: companies hiring more Singaporeans get a lower effective rate. It is directionally right. Anything defined at ground level gets gamed at ground level. Padded payrolls, fictitious roles, grandparents converted into nominal employees.

There is a deeper question underneath all of this, which is whether the corporate structure itself is the problem. The default purpose of a company under modern law is to maximise shareholder value. There is a long-running argument that companies should instead optimise for stakeholder value, where stakeholders include employees, customers, suppliers, and the community in which the company operates. The argument is sound in theory. Nobody has yet shown how to implement it without either re-creating the gaming problem at a different layer or destroying enough of the company’s competitive position that the workers it was meant to protect lose their jobs anyway.

The closest historical attempt at a stakeholder rebalancing is collective labor action. Unions are the older answer, and unions have a record. The Hollywood guild strikes through Covid and into 2023 are the cleanest recent case study: distribution upstream got destroyed, the box office never properly recovered, and the film industry now sits in something close to a death spiral propped up by a handful of directors and animation studios. The labor side won the strike on paper yet the industry the workers depend on got hollowed out in the process.

If unions are too blunt, the next instinct is to legislate stakeholder mechanisms directly. Profit-share floors, broad-based employee equity, mandatory ESOPs above certain headcounts, mandatory pension contributions on the company side. These also open new windows for arbitrage and raise the compliance burden, which reduces Singapore’s attractiveness relative to other low-friction jurisdictions. The compliance cost will fall hardest on the local SMEs, while the multinationals it was meant to constrain don’t consider Singapore at all.

There is no single knob that fixes the system because unfortunately at this stage, the problem is the system. Capitalism’s design has been progressively optimised over generations by people who built lives, businesses, and wealth on top of the axiom. They are not going to surrender it cheaply.

The structural bind

I am not against capitalism. In fact capitalism made us. The Singapore that lifted itself out of postwar poverty, the modern infrastructure, the millions of households who joined the middle class, all of it sat on top of a deal that rewarded capital because capital was, at the time, doing the work that justified the reward. The capitalist of sixty years ago looked like the person who built things. Today’s capitalist increasingly looks like the person who owns the shophouse, raised the rent by 57%, and pays nothing on the gain when they sell. Or the founder who raised a Series C on a software business he doesn't believe in, exited at nine figures, and now winters in Florida.

Unfortunately the bargain has tilted. Almost every part of the tax code in most developed nations including Singapore, and not just CIT, has been arranged in capital’s favour while labor’s share continues to thin. What to do about it is the harder question, and being a small country with a small population narrows the answers significantly.

There are two paths from here. The first is structural reform: rethink the ownership of companies, the role of companies in society, what counts as productive work, and what we want a life and a career to be for. That is the work of generations, and it requires reopening some of the deepest beliefs we hold about ambition and purpose. The second is to wait for the free market to correct on its own. Capital will eventually have to pay more for labor. The trigger could be falling labor quality, young people opting out of the system, or the squeeze becoming intolerable enough that wage earners stop accepting it.

Singapore in particular has structural constraints that make a domestic resolution very hard. We are an island of around 6 million people without natural resources, and our entire economic identity is built on attracting and retaining transnational capital. Without expanding our territory, growing our population to support a meaningful domestic market, or executing some piece of ingenious policy engineering that nobody has yet articulated, it is close to impossible for a nation state of this size to favour labor over capital.

I am on a road trip with my family right now, and it is not my first time here. I drove across parts of the US last October. On road trips, you get to see how normal people live, and get to talk to people and understand how they view their life. And at the same time, the long 6-7 hour drives makes me think about all that I have gathered from what I have seen and adjust my mental model about life in general.

On this particular trip so far, what I keep noticing is how Europeans actually live. The roads are good. People paint, garden, fix bicycles, take long walks. They sit at cafes in the middle of a Tuesday afternoon with a coffee or a glass of wine and they do not look guilty about it. They have hobbies they have been at for years. They have time for their dogs. They have time for their children.

I asked my wife somewhere between Germany and Austria on one of those drives. Would you pay around half your income in tax to live like this? A decent house. A decent car. Hobbies. Time. And the texture of a normal human life.

She did not give me a real answer.

A case against Late-stage Capitalism

I used to be more libertarian than I am now, that skin in the game would allow markets to allocate correctly. That when power is given to the state, it would corrupt. I still believe most of that. But the further I travel and the more closely I look, the harder it is to ignore that strong capitalism, taken to its logical end, has produced societies where a meaningful share of the population struggle.

Inequality itself does not bother me much. I think inequality is a vanity metric. Some people will always be richer, more capable, more ambitious. That is fine. What bothers me is the floor. When the bottom of the distribution sits below a level at which a person can survive, the society has failed at its basic function which is much more fundamental than fairness.

In the US, the Ludwig Institute for Shared Economic Prosperity has built a metric called the Minimal Quality of Life Index, which tracks not just survival expenses but the cost of the goods and services a household needs to participate in normal economic life. Their finding is that the bottom 60% of American earners cannot afford a minimal quality of life. That bottom 60% holds around 22% of disposable income but would need close to 39% to clear the threshold. In America, it is expensive enough that some people opt out of insurance entirely, and a single unforeseen event then derails a life that was, until that morning, perfectly viable. In much of Europe, this category of risk has been institutionalised away. In Singapore, we have built a middle path that mostly works.

Show me the incentive and I’ll show you the outcome

The argument against socialism, and I am using the word loosely here, is that it removes skin in the game and slowly erodes the engine that produced the prosperity in the first place. The Soviet Union and its breadlines are the easy reference. Communism and socialism are not the same thing, but they are derivative of a similar first principle: that the state should control some portion of what a society needs.

I do not refute this argument, as the saying by Charlie Munger goes: Show me the incentive and I'll show you the outcome. When you decouple effort from outcome, things break.

What I have started to believe is that the right answer is probably not a clean choice between the two. It is some uncomfortable balance. Let capitalism go hay in the domains where private ambition does not affect public interest. Hold it back in the domains where the absence of incentive actually increases welfare. Healthcare is one of those domains. Basic infrastructure is another. Education, probably. Housing, arguably.

Yet I find it interesting that I would stutter at the question of what it costs, and whether I would actually pay it.

What 40% in tax buys you

Top marginal rates in Denmark, France, Austria, Belgium and Sweden run above 50%. The average wage earner in Germany hands over around 39% of their salary to income tax and social security. A Danish worker pays around 36%. According to OECD data summarised by Euronews and the Tax Foundation, the average top personal income tax rate across European OECD countries is now above 43% percent, and the EU’s tax burden as a share of GDP sits near 40%.

What you get for that, in the better-run countries in the region, is a society where the median life is not a struggle. The roads like the autobahn is nicer to drive on than American roads. The hospital doesn’t charge you a bomb for cancer treatment. People take their full holiday. They walk their dogs in the middle of the day. They start a farm or a vineyard. They make things by hand and the things are good because the maker had time to learn.

And the level of care for their craft. The bag in the small leather shop. The pride they have in their bread and olive oil. A Saturday afternoon painting in the lawn. This seems to be what happens when a society does not require every citizen to spend the prime of their adult life sprinting toward financial safety. People start caring about their craft again. They start being human in the older sense of the word.

There is an old story about a CEO who meets a fisherman in a small village. The fisherman catches a few fish in the morning, plays with his children in the afternoon, takes a nap, then spends the evening drinking wine with his friends. The CEO tells him he should work harder, scale up, build a fleet, go public, and then in twenty years he could retire to a small village where he fishes in the morning, plays with his children, naps in the afternoon, and drinks wine with his friends. The fisherman looks at him. That is what I already do.

The parable is usually told as a joke about corporates but underneath it is a first-principles claim about what living is actually for. The fisherman simply skipped the decades of accumulation that the CEO treats as a prerequisite to the life he already has. If the point of all the work is to eventually arrive at a life you enjoy, and someone is already there, the logic of working harder to get there later does not seem to hold.

The democratic socialist system in Europe is not perfect, but it has let a large number of ordinary people arrive at the fisherman’s life without requiring them to build the fleet first. Yet if you strip away the chase, I am not confident we would know what to do.

I am not romanticising it as there are real costs. Productivity is lower. Innovation, by certain measures, is slower. Yet, does all these matter when on first principles, the government’s sole duty is to ensure that the average human is able to lead a decent lifestyle?

China, and the strange shape of the future

There is another model I keep thinking about, which is the lifestyle as a working-class urban Chinese household can now access, and what that lifestyle costs. For the price, the average urban Chinese life is, in certain respects, more affordable and more functional than what most of the world manages.

This makes even more sense as China is its own factory. The competition is insane and the scale of the domestic market is enormous, which means a manufacturer who fails to drive cost down loses to a manufacturer who succeeds, and there are tens of thousands of them. The result, over a couple of decades, is that the daily-life economy, the furniture, the appliances, the clothing, the food, the small functional objects that fill a home and a life, has been driven down to prices that do not really exist anywhere else in the world. I furnished my rental flat in China for under $10,000 which is a single mid-tier piece of furniture in a Western capital. The floor of material life has been raised because the cost of clearing it has collapsed.

The same logic shows up in infrastructure. High-speed rail moves you between major cities at a cost and frequency that the United States, for all its wealth, has not managed to build. Mobile payments, e-commerce logistics, food delivery, the boring connective tissue of an ordinary day, all of it works at a price point that is difficult to leave behind once you have lived there for a bit. The things you need to function are abundant, fast, and cheap. The things you need to enjoy a weekend are abundant, fast, and cheap. The system has its costs, some of them severe, but the lifestyle delivered for the price cannot be denied.

The real question

If Elon is right about where AI is heading, the long-run shape of a developed economy under heavy automation might end up resembling something like that. The productive AI and robotics base becomes the machine. The cost of physical goods continues its collapse, this time driven by automated production at scale. Energy, if the optimistic scenarios on solar and storage hold, becomes close to free. Routine work becomes optional. The state, in some form, redistributes the surplus, because the surplus has nowhere else to go - basically UBI or UHI as he likes to call it - Universal High income. You end up in a socialist utopia. Your time is your own.

Yet I am honestly uncertain whether this would be a good thing.

The real question, the one underneath the question I asked my wife, seemed to be this: If you removed money, or removed most of the financial reward of working, and still be able to have a decent lifestyle, would you still be able to find meaning?

The accumulation of private wealth as a measure of status is a recent invention in the long arc of human history. Money is more recent still. For most of our existence as a species, status was relational rather than material, and meaning came from what you contributed to the people immediately around you. We have spent the last few centuries building an entire civilisation on the assumption that financial reward is the primary engine of effort, and that effort is the primary source of meaning.

If that engine were removed, or even significantly weakened, would we still find meaning in life? Would we truly blossom with art and creativity like it has been theorised that we would, or would we crumble and corrupt the system again so that we can form some sort of zero-sum game which we humans like to play.

I do not know what most of us would do. I do not even know what I would do.

I didn’t think about the impact of population until last year. I was at the All In Summit, surrounded by an unusually dynamic crowd of founders, operators, and capital allocators. The hosts described the developed world’s trajectory with a word I wasn’t ready for:

Suicide.

The collective demographic suicide of the developed world. The West and East Asia are not just failing to replace themselves; they are actively choosing not to. This is happening through a slow generational surrender driven by preference, economics, and exhaustion.

Elon Musk, the Tesla and SpaceX CEO who was also speaking at the conference, framed it differently. Having children, he said, is basically about having hope for the future. That stuck with me, not because it was poetic, but because it reframed fertility as something more than a personal lifestyle choice. It made it structural, even civilisational.

It turns out this isn’t a new observation either. Rousseau, writing in 1762, had already identified population as the ultimate barometer of a society’s health. In The Social Contract, Book III, Chapter 9, he writes:

“The rest being equal, the government under which, without external aids, without naturalisation or colonies, the citizens increase and multiply most, is beyond question the best. The government under which a people wanes and diminishes is the worst.”

By that measure, most of the developed world is governed by the worst possible systems just because the conditions those institutions have produced make people not want to reproduce (I will talk about this later).

It’s Not a Labour Problem. It’s a Consumption Problem.

My first instinct was to think of depopulation as a manpower issue. Fewer people meant fewer workers, and surely AI and automation could fill the gap?

I was wrong; the crisis isn’t really about who makes things, but about who buys them.

Our global economy is, at its foundation, a consumption engine. Every business, no matter how far upstream it operates, ultimately depends on an end consumer. The company servicing oil rigs? Its revenue traces back to the person filling up their car. The enterprise SaaS platform? Its customer’s customer’s customer is a human being buying something. Strip away the layers of B2B abstraction and you always arrive at: a person, with money, spending it on something they want or need.

When that person doesn’t exist, the whole chain collapses upstream.

Eric Schmidt, the former Google CEO, also spoke at the All In Summit and made a related point. When your population shrinks, he argued, business valuations in that market generally dip.

The logic is straightforward. Valuation is fundamentally a bet on future cash flows; fewer consumers means less revenue, lower projected cash flows, and lower valuations. Which then translates to a lower investment returns for institutions and retail investors which in return affects the consumption power. A huge vicious cycle.

Consider the race between ChatGPT and Anthropic. Both are sprinting to capture market share. Suppose, hypothetically, one of them captured every person on Earth at $20 a month. That is the ceiling. The only way to grow revenue from there is to raise prices, but not everyone will pay more, and many will churn. Revenue could actually decrease. The growth story ends because there are only so many humans and there will be fewer of them.

Spotify offers a real-world case study. At some point, they effectively dominated music streaming. A middle manager at the company once told me, in a moment of unusual candour, that there was essentially nothing left to do but raise prices or expand the market. They had saturated their core offering. This is why Spotify launched podcasts, then audiobooks. Not because they were passionate about the spoken word, but because the only alternative to finding new consumers is finding new things to sell to existing ones. When the pool of existing consumers starts shrinking, even that strategy runs dry.

The Numbers Are Brutal

The replacement fertility rate, the number needed to maintain a stable population, is 2.1 children per woman. Here is where the developed world stands.

South Korea sits at approximately 0.68, the lowest in the world. Each generation in South Korea is roughly a third the size of the one before it. Japan hovers around 1.2. Singapore’s at 0.87. China, despite relaxing its one-child policy years ago, has a total fertility rate at or below 1.0. Italy and Spain are clustered around 1.2 to 1.3.

Even the United States, long considered the demographic exception among developed nations thanks to immigration and a relatively younger population, hit a record low fertility rate in 2025 of approximately 1.57 children per woman, down nearly 20% from two decades ago. The CDC’s provisional data shows 3.6 million births in 2025, and the trend line points only downward. Social Security’s trustees now project the US total fertility rate won’t reach even 1.9 until 2050.

Seven countries and territories worldwide currently have fertility rates below 1.0. The majority of Europe, East Asia, and increasingly Latin America are below replacement. By 2050, an estimated 151 countries will have sub-replacement fertility.

If you want to see what demographic decline looks like when it compounds over decades, look at Japan.

Growing up, I remember Japan being an expensive country to visit. The yen was strong, Tokyo was aspirational, Japanese brands commanded global premiums, and Japan, in many ways, was the future.

But following a multi-decade stagnation where GDP growth flatlined and deflation set in. The yen, once a symbol of economic might, weakened dramatically, hitting a 37.5-year low of 161 yen to the dollar in July 2024.

The collapse of the yen also had a second-order effect that few people outside finance circles talk about. It enabled the yen carry trade, where global investors borrow yen at near-zero interest rates and use that cheap capital to invest in higher-yielding assets elsewhere in the world. For decades, Japan effectively became the world’s piggy bank. The country that once exported Walkmans and Toyotas now exports cheap money. That is a sign of a domestic economy so anaemic that capital cannot find anything productive to do at home, so it leaves to find returns elsewhere.

There are many proximate explanations for Japan’s malaise: bad monetary policy, zombie banks, misallocated fiscal stimulus. All valid, but underneath these lies a demographic reality that makes recovery structurally difficult. Japan’s population has been shrinking since 2008, its workforce is contracting, and its consumer base is aging and spending less. The fiscal burden of supporting an elderly population with a shrinking tax base creates a vicious cycle that no amount of quantitative easing can escape.

Japan’s government debt ballooned past 140% of GDP by the early 2000s and has only grown since. Repeated stimulus packages, totalling hundreds of trillions of yen over the decades, provided temporary relief but did not produce sustainable growth. You cannot stimulate consumption when the consumers themselves are disappearing.

Even China Can’t Escape It

I’ve made multiple trips to China’s tier-2 cities over the years, places where you’d expect traditional family values to hold firmer than in Beijing or Shanghai: mid-sized cities with lower costs of living and more conservative social norms.

Even there, the conversations surprised me. Multiple people told me that they weren’t even thinking about marriage. Not that they couldn’t afford it, though many cited cost. More fundamentally, they described marriage as troublesome. And children? Why have kids?

This, in a country whose government has swung from enforcing one-child limits to desperately incentivising reproduction. China’s fertility rate has collapsed to around 1.0 and, by some measures, is below that. The government recently announced subsidies and yet the decline continues regardless.

China’s demographic projections are staggering. The country’s population is forecast to decline by nearly 48% by 2100. There is only so much a policy can do.

Can Technology Save Us?

Here is the question I keep circling. If the problem is consumption, and consumption requires human desire, can we just build machines that consume?

The short answer I believe is no, not yet, and probably not ever in any form that solves this particular problem.

Consumption, in the economic sense, isn’t just the movement of money. It’s the conversion of resources into subjective utility. A human buys a coffee because they want the warmth, the taste, the ritual; they book a holiday because they crave novelty and rest; they buy a bigger apartment because they need space for a growing family or simply because they want one. Desire is the engine, scarcity is the constraint, and subjective experience is the fuel.

A robot doesn’t want a coffee, and an AI doesn’t crave a holiday. Even if we programmed machines to “purchase” goods and services, the money would circulate within a closed corporate loop where the company that owns the robot pays the company that supplies the robot’s inputs. There is no new wealth creation because there is no genuine utility being derived. The fiscal multiplier of robot spending would approach zero compared to human spending, because humans have diverse, unpredictable, and perpetually expanding wants.

The only way technology actually solves the consumption problem is if we create entities that are genuinely sentient, autonomous, and possess rights. Not corporate property, but economic agents with their own desires, freedoms, and spending power. Whether that’s sentient AI or cloned humans with full legal personhood, the bar is the same: they need to want things, and they need the independence to pursue those wants.

Here’s the paradox though. The very conditions that would make such entities economically useful, namely sentience, autonomy, desire, and rights, are precisely the conditions that make the solution indistinguishable from just having more people. If your AI has to want things and have the freedom to act on those wants in order to be a meaningful economic participant, you haven’t invented a shortcut at all; you’ve just invented a more complicated and ethically fraught way to create a person.

Until we cross that threshold, we have a consumption problem with no technological exit.

Money Is a Ponzi Scheme (and That’s the Point)

Before we talk about solutions, we need to be honest about what money actually is.

Fiat currency is, at its core, a mechanism of future borrowing. Governments issue debt, central banks print money, and both are implicit bets that tomorrow’s economy will be larger than today’s, that future growth will retroactively justify present spending. This is not a conspiracy theory, this is the actual architecture. The entire global financial system is predicated on the assumption that the future will be bigger, more productive, and more consumptive than the present. This works as long as the assumption holds, and the assumption holds as long as there are more people, or at least more productive people, generating more economic activity over time.

When population shrinks, the assumption breaks. Debt issued today cannot be repaid by a smaller future economy. Asset prices, which are valued on discounted future cash flows, decline. Tax revenues fall. Pension systems, designed when there were six workers per retiree, collapse when the ratio hits two to one. The entire fiscal architecture of the modern state begins to buckle.

Japan has been living it; China is about to; Europe is well into it. The question is not whether this reckoning comes, but when, and whether it will be managed or catastrophic.

There Are Only Two Solutions

If we strip away the noise, there are exactly two ways to address demographic decline in any timeframe that matters to current economic systems.

The first is biological, nine months at a time. Have more children. Every pronatalist policy in the world, from cash incentives to extended parental leave, is an attempt to make this happen. The evidence is discouraging. Countries that have spent trillions on such policies, from Japan and South Korea to Singapore and Scandinavia, have seen at best marginal and temporary upticks in fertility. The structural forces pushing fertility down (cost of living, career opportunity costs, cultural shifts in gender roles, housing, and frankly the sheer difficulty of raising children in modern economies) appear to be stronger than any government incentive.

The Immigration Paradox

The second is immigration. Import the consumers and workers you cannot produce yourself. Immigration is the faster lever, but it comes with a cost that the developed world is only beginning to confront honestly.

Across Europe, anti-immigration sentiment has hardened into policy. The EU is expanding deportation powers and offshoring migration processing. In 2025 alone, over 80,000 pushbacks of migrants were recorded across the continent. The United Kingdom saw sustained anti-immigration protests throughout 2025, with over 180 arrests and dozens of police injuries, and the Labour government’s adoption of anti-immigration rhetoric has emboldened the far right.

London, to many of its longtime residents, is unrecognisable. This isn’t really about race, but about pace. The speed of demographic transformation has outrun the capacity for cultural integration. The same pattern plays out across Western European cities.

In the United States, the H-1B visa programme, ostensibly designed to attract high-calibre talent, has been plagued by abuse. Companies have used it to import cheaper labour rather than genuinely scarce expertise. Immigration enforcement under the Trump administration has led to a historic decline in net migration, with population growth dropping from over 1% in 2023 to just 0.3% in 2025.

Sweden, once the poster child for progressive immigration policy and one of the safest countries in Europe, has experienced a significant rise in violent crime that has been widely linked to failed integration of immigrant communities. The Swedish model, generous in its welcome but insufficient in its integration infrastructure, has become a cautionary example.

The core problem with immigration isn’t volume; it’s assimilation.

Singapore approaches this differently. The country allows immigrants of high calibre and requires them to go through structured assimilation processes such as National Education tours, community contributions, and integration programmes. It is selective, demanding, and unapologetic about it. Arguably, it may still not be enough, since Singaporeans complain regularly about feeling displaced in their own country despite these filters.

China takes a harder line still. Through programmes like the United Front, the Chinese government maintains influence over its diaspora, encouraging and in some cases pressuring overseas Chinese to maintain allegiance to the motherland or to return. This is less an immigration policy than a civilisational retention strategy.

Yet despite the messiness, despite the backlash and the legitimate grievances, immigration remains the only lever that operates on the timescale that matters to economic cycles. Business valuations and fiscal health move on one to eight year cycles, and you cannot wait eighteen years for the babies born in 2026 to enter the workforce, earn money, and start consuming. The economy cannot pause while a generation gestates.

This is the impossible tension at the heart of the population crisis: the biological solution is too slow, and the political solution is too volatile.

I Am Part of the Problem

I’m married, and we have two dogs. They call it DINK, dual income no kids, and we fit the profile of the very problem I’ve just spent three thousand words describing.

Why aren’t we having kids yet? I’m not entirely sure. It’s a confluence of factors, as it is for most people in our demographic. There’s the cost, the loss of freedom, and the sheer grinding hardship that I watch my peers endure, especially the ones who subscribe to gentle parenting philosophies and seem perpetually one tantrum away from a breakdown. There’s the uncertainty of bringing a child into a world that feels like it’s accelerating toward something none of us can name. And then maybe, honestly, there’s just the fact that we just don’t really like kids.

What I can’t escape, and what writing this essay has made sharper, is that cognitively I understand the necessity. I can see the math and trace the fiscal architecture, and I know that the persistence of the economic system I benefit from (the transnational capitalism that funds my lifestyle, my career, my dogs’ premium kibble) depends on the next generation of consumers and workers and taxpayers existing. My freedom, in a very real sense, is borrowed from someone else’s willingness to bear the cost of reproduction.

Rousseau would say I’m living under the worst kind of government, one under which the people wane and diminish. Musk would say I’ve lost hope for the future. Both framings are too clean, and the truth is messier than either. You can have hope and still not want to change diapers at 3am. You can understand your duty to civilisational continuity and still choose the dogs.

Maybe that contradiction is itself the diagnosis. A species that understands perfectly well what it needs to survive, and chooses not to do it, isn’t really facing a policy problem so much as a crisis of will.

I don’t know how you legislate your way out of that.

My friend Shulin Lee went on CNA last week. She runs a legal recruitment firm, one of the more prolific ones in town, and she said something that should not have been controversial. Companies are moving their hiring out of Singapore. The reason, in her direct experience placing candidates across the region, is that Singaporeans are less hungry than the people they are competing with.

People attacked her personally. They attacked her religion. Someone who was seeming on her side said that it was the MPs who should be doing something about it. I commented, snarkily as I always do, that if you have time to be this triggered by a recruiter about not being hungry, you are kind of proving her point.

I have spent the last few years building businesses and traveling across the US, China, Southeast Asia, and parts of Europe. What Shulin described is not a recruiter’s anecdote. It is what you see the moment you step out of the bubble. These hungry people are working harder than us and are not doing it for Singaporean pay. Many of them are doing it for a fraction of it.

We were them, once.

What we used to be

There is a Charlie Rose interview from 2009 where Lee Kuan Yew explains, with characteristic flatness, why Singapore became what it became. We had decent schools. We learnt from the Europeans, the Americans, the Japanese. And critically, we were cheap. The substance of what he said, across that interview and many others over the years, is that this combination is what turned Singaporeans into General Managers and Managing Directors across the region. We were the bridge labour, fluent enough in the West’s frameworks to be useful, hungry enough to outwork the people who taught us.

China has been running that exact playbook on us for twenty years as well. I have met more than a few Singaporeans who spent the bulk of their careers in Shanghai or Beijing on full expat packages, and almost all of them tell some version of the same story. They were valuable until the moment they were not. The Chinese counterparts they were sent to mentor learnt what they needed, and then the company quietly worked out that you could pay one local hire a third of what the Singaporean was costing and get someone who would stay later. A few of these people are now back home driving Grab.

The moats are thinner than we think

The standard pushback to all of this is that Singapore’s geography, governance, and head start will continue to insulate us. They will, for a while. They will not forever. Education, the thing we have built our whole self-image around, is being democratised at a speed that makes the question of where you got your degree less and less load-bearing every year. The premise that a Singaporean is structurally smarter and therefore deserves to be paid more is starting to look, from outside, like the kind of thing only Singaporeans believe.

Governance is the other pillar, and the honest read is that we have been lucky. The system has been effectively patriarchal in a way that has worked, and it has worked for long enough that we have started treating the outcome as a property of the country rather than of the configuration. Look at how quickly things can come off the rails elsewhere. London a generation ago was untouchable. Myanmar was opening up. Iran was modern. None of these unwound over decades. They unwound in a handful of years.

Categorically, it is not the job of a government to keep you competitive. You can drag a person to the edge of the labour market and you cannot make them swim in it. The other available move, the one we are quietly already making, is to convert Singapore into a place where capital congregates and where the legal and tax infrastructure does the earning instead of the people. Import the manpower. Lease out the jurisdiction. Become a kind of GCC, or a Cayman with better food. The catch is we do not have the resources the Gulf has. Our only resource was supposed to be us.

Look at accountancy. The profession in Singapore is now substantially staffed by Malaysians who cross the border to do work that Singaporeans no longer want to do. Accountancy is grindy, unglamorous, high-grit work. The fact that the local pipeline has thinned out is not a labour market quirk. It is the leading edge of a pattern.

LKY warned us about Iskandar. His line was that at the stroke of a pen they could take it over, and the broader anxiety, repeated across his speeches and interviews, was that Johor’s proximity and ambition would eventually cost us our edge. The only reason that anxiety has not aged into a prophecy is that the Malaysian state has been consistently inefficient enough to bail us out. That is not a strategy. That is a coincidence we have been pricing in as a moat.

Affluence rots quietly

Jiang Xue Qin, who has written carefully about Chinese education and society, makes the broader point. Affluence makes societies lazy, arrogant, and stupid. The downfall of most great civilisations follows roughly that arc. The societies and groups that displace them are the ones that stayed energetic, hungry, and open. The pattern is fractal. It applies to nations, but it applies just as cleanly to states and cities. Austin pulls people in because it is open and not yet self-important. The Bay Area, for all its rot, still gets a constant inflow of people willing to sleep on a couch for the chance at the startup dream. The places that win are the places that the hungry are still moving to.

The legal fictions we live inside

What unsettles me about the Shulin reaction is not the anger. It is the assumption underneath the anger, which is that everything around us is fixed. That the certifications are real. That the property values are real. That the institutions will keep producing the outcomes they have produced. None of that is real in the sense people think it is - its is all legal fiction, transient in nature. A degree from NUS is a coordination signal that employers have agreed, for the moment, to accept. Strip the agreement and the signal is just a piece of paper. An employer from first principles has every right to hire a “non-qualified” person versus one with the “higher qualification”. The true qualification is whether you are able to do the job with the level of competence and eagerness as is required. HDB prices, ownership tenures, the entire residential ladder, are a fiction of demand and supply and policy outputs. Change the policy, adjust the demand and the ladder changes shape. Treating any of this as bedrock is fragility dressed up as common sense.

I see this in myself too. The reflex to assume the system will hold is not something other people have. It is something we all carry, and the only way to notice it is to spend enough time in places where the system did not hold.

The window we mistake for the world

Ray Dalio has a line about how each of us only really internalises the history we have lived, and a bit of what our parents and grandparents told us about the history they lived. For a Singaporean alive today, that window covers roughly the 1970s to now. It has been a golden age. By the standards of human history it has been an aberration. Most of recorded time is tumult, displacement, and contingency. We have no muscle memory for the other thing because the other thing has not happened to us yet.

No one is coming to save you. The MPs are not. The certifications are not. The property is not. The country has been carried for fifty years by a generation that knew it was poor, and we are the generation that has the privilege of forgetting.

The only thing left, at the level of a country, is to stay energetic, open, and hungry. The unsettling part is that this is not something a policy can mandate. It is something a people either still are, or are not.