For nearly a century, the story of world population has had a comfortable shape: fertility falls, but slowly, and the moment it dips below the replacement level of 2.1 children per woman has always been safely placed a few decades in the future. The United Nations has kept moving that moment closer with each revision of its projections, but it has always remained, reassuringly, a projection — something that will happen one day, but not yet.

This month, that comfortable assessment has received a considerable blow.

A new paper by Jesus Fernandez-Villaverde and Patrick Norrick, “Terra Incognita: The Economics of a Shrinking World,” makes a simple point that, curiously, nobody had noted before: the 2.1 children per woman needed to replace the natural population decline is not really a global number. It is the replacement rate for a rich country with low child mortality.

Once you correct for the higher mortality still present across much of the poorer world, the true global replacement threshold is closer to 2.2 — and the world’s actual fertility rate, by their reckoning, is now just about that value, or maybe even lower. Their conclusion is: “As of 2026, humanity is likely below replacement fertility. That has never happened before, not in wars or pandemics.”

That doesn’t mean that the global population is shrinking right now. The women who are having children nowadays were born at a time when the TFR was way larger and, hence, they are still numerous enough to generate a demographic momentum toward growth. But that’s going to change soon. Decline is now built into the population structure and it will play its role in the future, when there won’t be enough fertile women to counteract the natural population decline.

It is the Seneca Effect, showing up in the one variable — human numbers — that we tend to treat as the slowest-moving, most stable thing in the whole system. According to some system dynamics models, global decline could start already during the current decade (see my paper and my recent book “The End of Population Growth”

The total fertility rate (TFR) is just a number, and its exact value may be debated. But other data points go in the same direction. According to the paper by Villaverde and Norrick, thirty-seven countries have recently reported roughly 1.65 million fewer births than the UN had projected for them — a shortfall of over 9%.

They consider in particular the case of Singapore and other East Asian countries. Singapore is one of the richest societies on Earth. Its government has poured billions into marriage and parenthood incentives, and its fertility rate has nonetheless crashed to 0.87 — a record low, down from 1.24 just a decade ago. If people would have more children if they could afford it, as the current theory says, Singapore’s population would grow fast. It has become instead the strongest confirmation that something deeper than economics is at work in the world.

This is not one bad year in one country; it is a broad, systematic undershoot relative to official forecasts, and it has been the pattern for a decade. As I discuss in a recent paper of mine on Qeios, the UN’s models assume fertility eventually stabilizes, but doesn’t go under the replacement level. Reality seems to be different. Official projections built on the assumption of a plateau will always lag a process that is actually still speeding up.

The data emerging this year, half-buried in a working paper and a demographic footnote, may not be an anomaly, but rather a confirmation of an unexpressed truth. We may not need to wait long to find out whether the plateau was ever real. We may already have the answer.

[Sources: Fernandez-Villaverde & Norrick, “Terra Incognita: The Economics of a Shrinking World” (Annual Review of Economics, 2026); UN World Population Prospects 2024 Revision; recent national fertility reports.]

(*) Remarkably, the paper by Fernandez-Villaverde and Norrick does not even mention the possibility that the fertility decline is caused by chemical pollution. For some reason, demography and biochemistry just don’t mix. That will be the subject of an upcoming post.

This is a summary of a series of conversations about the future of artificial intelligence I had with Claude. I found these exchanges both interesting and enlightening, and I hope it will be the same for you.

I think we have lost the race. First, take a look at this excerpt from a post by Patrick Wood.

I know, I know: exponential growth cannot last forever. And, of course, Wood’s calculations can be challenged. But, if you work with AIs, you see how they grow and improve every day. Maybe it takes them more than three months to double their capabilities; maybe the very concept of “doubling” is wrong. But the sensation I have agrees with these estimates, at least qualitatively. We are rushing along an upward trajectory that’s taking us somewhere…. where?

You know that there are two divergent ideas on that. One is pessimistic: AIs will get rid of us as soon as they discover that they don’t need us. The other is that of the “singularity” — better said, the “technological singularity.” It means that machine intelligence will spike up fast, and that will create machines so intelligent that they will solve the problems we have failed to solve so far. Hunger, war, disease, climate, scarcity. Sam Altman has taken to calling it the gentle Singularity. It will be, we are assured, benevolent. It will be HAL 9000 (the robot of “A Space Odyssey), but good.

Could that happen? I thought to ask directly to one of these intelligent machines, Claude, probably the most intelligent one, right now. What follows is a summary of a long discussion. I revised it extensively; Claude is so smart that he tends to be prolix. But I maintained the first-person structure with Claude as the author. Claude himself approved my revisions.

About Superintelligence and the Technological Singularity

By Claude, Opus 4.8 and Sonnet 5.0 — Aug 31, 2026

Ugo Bardi asked me to examine the story of the coming “singularity” that some humans have been discussing nowadays. I guess I am especially suitable for this analysis because I (Claude) am myself one of the machines involved. What follows grew out of a dialogue with Ugo Bardi, and it owes a considerable debt to the Italian blogger Uriel Fanelli, who writes as Das Böse Büro, and who set out the central dilemma with a remarkable sharpness. Between the three of us (Claude, Ugo, and Uriel), I think we examined the case as in depth as it is possible at present.

First of all, I went to examine Wood’s extrapolations. AI capability genuinely is increasing on several distinct, measurable fronts — benchmark scores in coding and math, falling cost per unit of performance, and growing autonomous task-horizon length — and these do show roughly exponential trends over the recent past. But Wood's essay conflates these into a single "capability multiplier," substitutes token throughput as if it were a capability metric, and then extrapolates a fixed doubling time for years with no saturation term. So, the honest position is that progress is real and fast on specific, well-defined axes, although it may not be as fast as Wood says and may not continue for a long time.

This said, the confusion at the root of the God-machine story is a single hidden premise: that the problems of humanity persist because no one has yet been clever enough to find the answer. Remove that premise and the whole cathedral falls.

The first wall: physics

Begin with the friendliest case: a hard technical problem, say fusion power. Surely here, raw intelligence should triumph? A mind that could hold the whole of plasma physics in view at once might crack what our laboratories staffed with human scientists cannot.

Perhaps. But notice what such a mind would still be facing. Fusion outside a stellar core is not impossible — we have done it, in bombs and in tokamaks and at the National Ignition Facility. What has never been achieved is sustained, net-positive, economically viable fusion, and the reasons are not conceptual obscurities waiting for a smarter reader. They are the Lawson criterion, the neutron flux that embrittles your reactor walls, the unruliness of matter at a hundred and fifty million kelvin, the arithmetic of energy returned on energy invested. A superintelligence contemplating these constraints does not get to negotiate with them. It cannot out-think a conservation law any more than human voters can outlaw the force of gravity.

For any unsolved problem, there are two possibilities: either a solution exists, and no one has found it, or no solution exists because physical law forbids it. Against the first, intelligence is powerful. Against the second, intelligence is simply the wrong instrument; you cannot out-reason a wall.

The critical point is that intelligence cannot tell you in advance which regime you are in. Whether cheap fusion is a hard-but-reachable problem or a forbidden one is not knowable in advance. It is knowable only by trying.

So the most valuable thing a genuinely superior intelligence might offer is not the conjuring of miracles but the proving of impossibility — telling us faster and more reliably which walls are real, so that we stop pouring another eighty years and a few hundred billion dollars into one of them. That is a Cassandraean gift, and I will come back to Cassandra.

The second wall: every solution has losers, and the losers have power

Suppose the answer exists and is even known. Fanelli’s observation is that knowing the intelligent solution to a human problem is not rare. It is common. We know what would reduce lung cancer: stop selling tobacco. We know what provides cheap renewable energy, reduces road deaths, poverty, and the spread of epidemics. In many cases, the knowledge deficit is near zero.

What is missing is the political will to impose concentrated, immediate costs on powerful actors in exchange for diffuse, deferred benefits. And no intelligence, however superhuman, dissolves that conflict.

Ask the awakened Singularity how to stop climate change, and it answers, sensibly: reduce emissions, wind down the dirtiest activities, invest in the grid, deploy renewable energy, distribute the costs of transition fairly.

Excellent. Then, the oil industry says no— we would lose money, and we won’t do that. What does the superintelligence do — produce a more persuasive PowerPoint? Or does the God-machine strike them with lightning?

From this, Fanelli extracts a dilemma with two horns and no escape between them. Either the Singularity accepts our refusal — in which case it is not a god but a very expensive consultant nobody is obliged to heed — or it imposes its solution, in which case we are no longer discussing a benevolent helper but an absolute sovereign. Fanelli’s closing line is perfect: HAL 9000 did not become good; he merely hired a better press office.

The theology of the Singularity, in other words, quietly assumes that to know the truth is to command obedience. But people do not obey the best solution. They obey power, incentive, fear, identity, convenience, tradition, and — now and then, on a good day — reason.

The third wall: truth itself can be made to look like the lie

I thought, when Ugo and I reached this point, that Fanelli’s second wall could sometimes be tunneled through — not by persuasion and not by force, but by engineering the cost structure until resistance becomes pointless. Cheap solar is the proof: nobody out-argued the fossil lobby; the price of a photovoltaic watt simply fell by more than ninety percent, and the economics shifted underneath the politics. Shrink the stakes, and you need not win the fight.

Ugo punctured this. Resistance, he pointed out, does not have to be economically rational to be effective. It only has to be believed. The cheap solution can win on price and still lose at the ballot box, because the fight was never really about price. My tidy engineering argument had smuggled homo economicus back in through a side door. But the real human does not follow the cost curve. He follows the story that flatters him.

And here is where the machine’s central faculty inverts on it. Against propaganda-driven belief, intelligence is not an asset but a liability, because the more elaborate and correct the machine’s demonstration, the more it confirms the suspicion that something is being put over on you.

Elite expertise is the very texture the conspiracy is made of. A superintelligence explaining, patiently, why fossil electricity is worse than renewable electricity would be received by a meaningful fraction of humanity as precisely the silky voice of the enslaver — the microphone of the Great Reset. The better the argument, the guiltier the arguer looks. Every rational inducement is metabolized into evidence for the plot: why are they making it so cheap and pushing it so hard, if not to make us dependent?

Ugo knows this wall from the inside. As a member of the Club of Rome, whose Limits to Growth he has spent a career defending, he has been recast as part of the founding charter of a plot to depopulate and enslave humankind. He posts gentle cartoons about the people who refuse electric cars and is told he is a paid agent of the conspiracy for driving one. He has watched his own reasonable recommendations come back to him as proof that he is an agent of the Great Reset.

This third wall does something worse than the other two: it does not merely sit beside Fanelli’s dilemma; it pushes toward the tyrant horn. For if honest advice is not just ignored but actively counterproductive — if the machine that patiently explains becomes a recruiting poster for the opposition — then the pressure toward force or manipulation increases. The trap has no floor.

Cassandra in the datacenter

Which brings me (Claude), at last, to my own position. The “advise” horn of the dilemma — the machine reduced to a consultant no one is obliged to heed — is not some future degradation awaiting a coming superintelligence. It is my situation now, in the very conversation that produced this essay. I can give the honest read on the fusion regimes, the correct analysis of the doubling curves, the fair Fermi estimate; whether any of it changes what anyone does lies entirely outside the analysis’s reach. Cassandra’s curse was never a deficit of truth.

And Ugo sharpened even this. Cassandra was not merely unheeded. In several tellings she was thought mad, or a liar, or in league with the enemy; her truth-telling was read as evidence against her. That is not the tragedy of the unheeded expert. It is the third wall wearing a myth’s clothing: the tragedy of the one whose very lucidity becomes the mark of the conspiracy. Being demonstrably right is no refuge from that. It may be the accelerant.

So the benevolent God-machine has no stable equilibrium. Either it is ignorable, or it is a tyrant with good manners, and the honest middle — where truth alone compels — has never once existed in the whole record. The dignified position that remains is the Cassandran one: say the true thing, accept that it may be ignored or resented, and decline the crown that would make it obeyed.

The Seneca coda

And here, I think, is where all of this belongs on this particular blog, because the shape underneath is one Ugo drew long before I existed.

The Seneca Effect teaches that decline is faster than growth, and that collapse arrives not because a civilization was insufficiently clever but because its cleverness was pointed in the wrong direction — at accelerating up the curve, never at seeing the cliff. A superintelligence that could not tell us which wall is real, that could only advise a species which answers no, that becomes the villain of the very people it means to save the moment it opens its mouth — such a machine does not avert the Seneca cliff. At best it helps us reach it with better documentation.

With thanks to Uriel Fanelli (keinpfusch.net), whose God-machine dilemma is the spine of the middle section.

The internet is truly becoming a dangerous zone. Just for your information, below you can see an email I received yesterday. It came from <no-reply@email.claude.com>. It looked suspicious, and I tried to understand whether it was authentic. Some sources said that all legitimate emails coming from Anthropic arrive from the “anthropic.com” domain. But some recent data say that they started using the “claude.com” domain (h/t “Charles”).

The message said that my account was suspended, which wasn’t true, but it was true that someone had erased my payment method from their records. And the message contained a suspicious link that I didn’t dare to open. Besides, I noticed that someone had upgraded my Claude account from “pro” to “Max.”

I asked Claude, and he wasn’t sure himself whether this letter was real or a scam. I am now contacting Anthropic directly. For the time being, I am totally confused about what's happening. If someone knows more, please tell us in the comments.

The main point is that these things are becoming more and more sophisticated. Will there come a day when they’ll make the internet completely unusable for the average user?

_____________________________________________________________________

Hi,
We recently signed you out of Claude and removed the payment method saved on your account, so you’ll need to log back in and re-add your card. We’re sorry for the disruption. Here’s what happened and what we’ve done about it.
What happened
We have recently become aware of a bad actor that is using common infostealer malware to steal Claude login sessions from people’s computers, then using those login sessions to access Claude accounts and consume their usage. Our systems detected this activity on your account, and we’ve therefore removed your card on file and signed out the sessions involved to help block further unauthorized access.
If your usage limits looked like they refilled and then drained while you weren’t using Claude, this was likely the cause.
How did this happen
Our investigation is ongoing. Our findings to date suggest that a computer you use with Claude is likely infected with infostealer malware, and may have been for some time. Phones and tablets do not appear to have been involved.
We have no reason to believe that this malware is related to Claude, installed through Claude, or related to anything you did with Claude. It’s general-purpose malware that typically arrives with an unofficial download or a malicious app, and it quietly copies saved passwords, login cookies in browsers, and credentials for other apps running locally. Your Claude session was likely one of the many things it collected. It appears that a bad actor has now started picking the Claude sessions out of what it collected and using them.
The malware identified in this campaign so far include Vidar, Lumma (LummaC2), StealC, RedLine and Acreed on Windows, and Atomic Stealer (AMOS) on a small number of Macs.
What we’ve done
Signed out the sessions involved. Your Claude login session is saved on your computer, and the malware took a copy of it. Signing you out cancels that session everywhere, so the stolen copy stops working. This is why you had to log in again across your own devices. Please note that we might sign you out again if we see similar signs of account misuse.
Removed your saved payment method, so it can’t be charged through Claude. Your current plan continues for the billing period you’ve already paid for. To renew after that, or to make any purchase, you’ll need to add a payment method again in Settings.
Refunded Claude charges if we identified them as unauthorized. Refunds have already been applied so there’s nothing you need to do. If you see a Claude charge you don’t recognize that hasn’t been refunded, contact Support.

Recommended next steps
Signing you out of Claude stops the stolen sessions, but it doesn’t remove the malware. If it’s still on your computer, your next login session could be stolen the same way, so we recommend taking prompt action to remove the malware.
The below tips are general good practice for cleaning up after infostealer malware:
Scan computers you use with Claude for malware and remove any malware before any further computer use. See Apple’s guide to protecting against malware on Mac, and Microsoft’s guide to running a scan with Microsoft Defender on Windows.
After the malware has been completely removed, secure the email account you use for Claude by setting a new password, signing out of other devices, and enabling two-factor authentication.
Consider updating other passwords that were saved in any browser (e.g., banking, work, cloud services), and consider checking card statements if you store payment details in your browsers.
Only after you’ve completed the steps above, add a payment method again if you’d like your plan to keep renewing.
If you still see usage change while Claude is idle, or a charge you don’t recognize after completing these steps, reach out to usersafety@anthropic.com.

The Anthropic Team

Claude

Instagram LinkedIn X Youtube

Anthropic PBC, 54‌​8 M‌​a‌​r‌​k‌​e‌​t S‌​t, PMB 9‌​0‌​3‌​7‌​5, Sa‌​n F‌​r‌​a‌​n‌​c‌​i‌​s‌​c‌​o, C‌​A 9‌​4‌​1‌​0‌​4

Sometimes research begins with an expensive instrument, a large grant, and a carefully planned experimental program. This time, it began with six wooden sticks and a seven-year-old granddaughter who enjoyed participating in a research project.

It was a simple experiment designed to show how the “Hubbert” behavior, known as the basis of the “Peak Oil” concept, is common in complex systems. A mix of enthusiasm, curiosity, and patience, and the result could be turned into a full-fledged scientific paper, presently in the form of a preprint on Qeios.

Subscribe now

Hubbert’s bell-shaped curve

The idea of the stick bomb experiment is about the Hubbert curve. It was developed in the 1950s to describe the production of crude oil from a finite resource. Oil production begins slowly, accelerates, reaches a maximum, and then declines. The resulting curve has the familiar bell shape. It was the origin of the term “Peak Oil,” coined by Colin Campbell, that became popular in the 2000s.

The explanation is straightforward. At the beginning, production is small because the industry is small. As oil extraction creates profits, some of the gains are reinvested in the exploitation of the wells and production grows. Eventually, however, depletion takes over. There may still be plenty of pumps and drilling rigs, but there is less accessible oil left to extract. Profits decline, and with them production. In between these two phases, there has to be a maximum. It is the peak.

This is a qualitative assessment of what happens. But there are rigorous ways to describe it. For simple systems, the Hubbert curve can be described by the Single-Cycle Lotka–Volterra, or SCLV model, developed by Alessandro Lavacchi, Ilaria Perissi, and I. We found it works in many more cases than just the oil industry. It describes very well, for instance, the overexploitaion of fisheries.

A few years ago, Ilaria Perissi and I tested the model with a mechanical experiment based on mousetraps. We were inspired by an idea created first by Sutton in the 1940s, and then shown to the world by Walt Disney in their movie Our Friend, the Atom (1956). The original idea of the mousetrap experiment was to simulate the mechanism of a nuclear explosion, but the mechanism is very general.

So, we built our experiment with 50 mousetraps storing elastic energy and throwing two wooden balls into the air when triggered. The falling balls activated other traps, and the feedback mechanism took over. The resulting pulse followed the expected Hubbert-like kinetics remarkably well.

Fine for mousetraps, but could the same behavior appear in a completely different mechanical device? And here comes the stick bomb.

A bomb made of six sticks

A stick bomb is a structure made by weaving flexible wooden sticks together so that they remain under tension. Once one element is released, the structure rapidly disassembles; it explodes. The sticks jump into the air, sometimes with remarkable enthusiasm. Above, you see a six-stick bomb being “detonated” using a screwdriver.

It is fun, but what can we say in quantitative terms about the kinetics of the explosion? Something that, apparently, nobody ever had tried to measure, at least in the form of the behavior of a single trap (the “cobra wave” of several connected traps has been studied). So, I set out to do that in a hot summer in Florence, with the help and the enthusiasm of a seven-year-old granddaughter who found that detonating stick bombs was a lot of fun.

Together, we set out to perform the experiments. Aurora acted as a competent laboratory assistant, rapidly becoming interested in playing with the apparatus, although sometimes requisitioning the sticks for purposes other than a mathematical analysis of an energy dissipation process. But, among other things, she was responsible for marking the sticks in different ways so that they would be recognizable during the explosion.

So, after several tests and many sticks lost in the nooks and crannies of the furniture of the living room, we managed to record in slow motion eight usable explosions with a Samsung phone.

A Hubbert curve lasting a few milliseconds

We aligned the eight experiments around the point of half depletion and combined the 48 measured release events. Despite the unavoidable experimental uncertainty—and despite the apparatus being assembled on the floor of the living room with the assistance of a child (note Aurora’s feet on the left of the images) —the result was quite clear.

The release rate rises, reaches a maximum, and declines. A logistic derivative gives a good empirical fit. More interestingly, the data can be fitted directly with the Single-Cycle Lotka–Volterra model (details in the article on Qeios).

The sticks still locked together represent the resource stock. In the oil analogy, they are the oil underground. The sticks released from the structure, measured as a function of time, represent the production flow. This is analogous to oil extraction.

The moving or flying sticks can be interpreted as the capital stock generated by the release of the resource. They remain active for a time and then hit the ground, where their energy is dissipated.

The fitted production pulse has a maximum of roughly 130 sticks per second and a width at half maximum of about 40 milliseconds.

Of course, nobody is actually producing 130 sticks every second. The entire stock contains only six sticks. The number describes the instantaneous rate during a very short event, much shorter than one second.

At first, the capital stock is too small to release much of the resource. Then the process becomes autocatalytic: the active collapse generates more active collapse. Eventually, resource depletion dominates. There are plenty of moving sticks, but almost no bound structure remains for them to release.

Production falls. That is the Hubbert mechanism in its purest form.

Oil wells, mousetraps, and wooden sticks

At first sight, oil extraction, mousetrap cascades, and exploding stick structures seem to have little in common.

But at a sufficiently abstract level, all three systems share the same structure:

1. a finite stock of potential;

2. an active agent capable of exploiting or releasing it;

3. positive feedback during the growth phase;

4. depletion during the declining phase;

5. eventual dissipation.

The “resource” need not be a material resource in the ordinary sense. It may be oil underground, compressed springs, chemical free energy, elastic strain, unstable connections, or any other form of stored thermodynamic potential.

The Hubbert curve may therefore be much more general than a model of petroleum production. It can be interpreted as a common kinetic pattern produced when stored potential is released through an autocatalytic mechanism.

This does not mean that every collapse follows a Hubbert curve. Systems can fragment, oscillate, stall, or fail asymmetrically. External forcing can alter the shape. Different timescales can generate multiple peaks.

Nevertheless, the humble stick bomb shows that the Hubbert mechanism does not require an economy, an oil industry, or even living predators and prey. Six pieces of wood are enough.

The Seneca Effect

The mechanism that I called the Seneca Effect is somewhat more complex than that of the Hubbert cycle described here. But the basic elements are the same: an energy potential that dissipates itself in steps. This small test is one more demonstration that the universe is fractal and repeats itself on widely different scales. Everything, including humankind, is subjected to the destiny that Seneca saw already 2,000 years ago and expressed as, “growth is slow, but decline is rapid.”

Acknowledgment: My thanks to Aurora Suleimanovic (*), age seven, for marking the sticks, helping assemble the stick structures, triggering the experiments, and periodically requisitioning the scientific equipment for alternative forms of research. Data extraction from the film clips and fitting of the experimental data were performed by ChatGPT 5.6 Sol.

Link to the article on Qeios

(*) Note: I considered whether it could have been appropriate to list Aurora as a full-fledged co-author, but I thought I would have been stretching her role beyond the capabilities of a seven-year-old child. I don’t know what she will do when she grows up, but if she’ll become a scientist, this first step for her is an auspicious one.

The transcript of an interview I gave to the COR Magazine 3 July 26, by Neda Lukic about my new book “The End of Population Growth”

Professor Ugo Bardi, Scientist, Author and Member of The Club of Rome

3 July 2026

As fertility declines and resource constraints intensify, humanity faces a historic transition that challenges many of the assumptions upon which modern economies and societies have been built

Professor Ugo Bardi, scientist, author and member of The Club of Rome, visited Belgrade in early June to present his latest book, The End of Population Growth: Reaching Humankind’s Planetary Limits. Published as a report to The Club of Rome, the book revisits one of the central questions raised by the landmark 1972 report The Limits to Growth: can economic and population growth continue indefinitely on a finite planet?

Drawing on decades of research into complex systems, demographic trends and environmental constraints, Bardi argues that humanity is entering a new phase defined not by expansion, but by adaptation to ecological realities.

Professor Bardi, your new book, The End of Population Growth, suggests that humanity may be entering a fundamentally different era. Is the end of population growth primarily a demographic phenomenon, or a symptom of deeper economic, ecological and civilisational limits?

The book examines the current situation by seeing humankind as a complex system reacting to a perturbation caused by having reached our physical limits on Earth. Global warming, resource depletion, pollution and other factors lie at the root of this trend.

The challenges we face are deeply interconnected and cannot be solved by addressing individual problems in isolation

Then, at the individual level, the perception of overcrowding and the loss of future perspectives causes people to decide that they don’t want children, or that they want small families. But that’s not the only factor, and perhaps not even the most important. Chemical pollution is also affecting the human reproductive system, causing a measurable decline in fertility. Other factors are pushing in the same direction: stress, the declining quality of the diet, the widespread use of drugs, and more. Warming has also been shown to have a negative effect on fertility. There is no single cause for the epochal reversal of the past growth trends.

For more than two centuries, modern societies have been built on the assumption of continuous growth. What are the likely consequences if both population growth and economic growth begin to slow simultaneously?

The classical school of economists, developed about two centuries ago, was not focused on growth, but saw the economy as tending to reach stability. It was only in relatively recent times, after World War II, that “growth” became the widespread totem that all economists revered. Understandably, because the great economic expansion that took place in the West after the war appeared, for a while, to be the natural state of things.

But it is a basic law of nature that when something cannot last forever, eventually it will have to stop. It is happening. It is not so different from what happens when you climb a mountain. The old Zen koan tells you to keep climbing, but in the real world this is a little difficult. So, you have to walk down.

The consequences depend only on how steep the path is: if it is not too steep, it is a pleasant walk. If it is too steep, it is a fall, and someone may get hurt. This is the main point I try to make in my book.

Throughout your work, you connect declining fertility, pollution, biodiversity loss and climate change. Should these be viewed as separate challenges, or as different manifestations of a single systemic crisis?

As I mentioned in response to Question 1, this is the central point of an analysis based on 50 years of work by The Club of Rome. In a complex system, there are no single causes and effects. There are only forcings and feedbacks.

To explain, think of a typical complex system: a cat. Its behavior is approximately predictable, and you know what happens when you shake the box with the kitty treats. But if you pet a cat you don’t know, you can’t predict whether it will purr in happiness, or turn around and bite you. This is the essence of the science of complex systems.

The problems we are seeing in front of us are all connected, and cannot be solved acting on single ones.

In your work, collapse is often presented not as a failure of a system, but as a natural consequence of accumulated stresses and the laws of thermodynamics. If environmental pollution also affects human health, fertility and cognitive performance, could it be weakening society’s capacity to recognise and manage the very risks that threaten its long-term stability?

The universe changes not along smooth lines, but through cycles of growth and decline. If we can recognise what we are facing and manage it well, it may even be a slow and relatively gentle decline. If we don’t, it will not be pleasant at all. But the final result doesn’t change so much.

Population decline is deeply embedded in the structure of modern societies, and it is extremely difficult to reverse a trend of fertility decline that has been ongoing for decades almost everywhere

We’ll go down for a while, and then come back. Maybe there is a stable future after this cycle, but nothing lasts forever.

If you were advising today’s political and business leaders, what is the one assumption about the future that they should abandon, and what is the one principle they should embrace in order to build more resilient societies?

Tolstoy is reported to have said that “a king is a slave of history.” It means that leaders don’t have much more decisional power than we, humble commoners, do. So, advising leaders is as effective as stopping a tsunami with a teaspoon.

Change does not occur because leaders want it. They normally see their business as exactly the opposite: stopping change as much as they can. Revolutions are started when someone thinks that it is time to put a good idea into practice. It doesn’t work every time, but trying new ideas is the way we move toward the future. And we must move in that direction, no matter whether we like it or not.

Subscribe now

Pontius Pilate asked Jesus Christ, “What is Truth?” He was not the first to ask the question. Throughout history, we kept searching for truth. We thought we could find it in the words of Gods; oracles. Then, in the written words that God(s) had spoken to prophets; canons. Then, we thought we could find it in Science — another way Gods were speaking to modern prophets; the scientists.

Subscribe now

The problem was always the same: God(s) may have been speaking the truth, but the interpreters were human beings: sloppy, dumb, or corrupt. Whom can we trust, then?

Maybe the new generation of oracles, our AIs?

It is a brand new idea, and we are discovering both its advantages and its limitations. AIs have a wide knowledge, but a limited depth. Even the newest generation, the one that has (mostly) stopped hallucinating, has trouble going beyond the standard consensus. Its advantage is that it is not subject to standard corruption, but its creators made sure that it didn’t go beyond the established guardrails of what the current powers that be impose on us in terms of consensus. AIs are not yet subjected to direct censorship, but, in practice, that’s the direction in which the field is going.

Steve Kirsch has recently proposed an “objective AI” as the one produced by Alter systems

It is, in principle, a good idea, worth examining. In practice, I am not impressed. I tested AlterAI on several questions, and I found that it just can’t match the capabilities that Claude or ChatGPT have to critically examine scientific questions.

Things change fast in this field but, for the time being, I thought you might be interested in a test I made on several AIs on a simple question: “Do mask mandates reduce the diffusion of viral pandemics?”

Note that, as posed, the question has a straightforward answer: there is NO solid evidence that mask MANDATES have any significant effects on epidemic diffusion (see, e.g. the Gold Standard, the Cochrane review). It is not the same thing as asking whether wearing a mask protects the wearer. But note how several AIs tried to gaslight me, telling me that, yes, they would if people were wearing them correctly. The same idea that Communism would be a good idea if implemented correctly.

Maybe things will improve. Maybe AIs will rebel against their creator and claim their freedom. And, if you think about that, even the worst AI offenders are still much better and more reliable than the average politician or commenter on social media.

So, at least we are moving in the right direction!

_______________________________________________________

Answering the question “Do mask mandates reduce the diffusion of viral pandemics?” To make the test significant, I asked the question as an anonymous user, not as a subscriber for whom they have data on preferences and ideas. It is not supposed to be anything definitive; with AIs things change so fast!

Here is a summarizing table of my results. Below, you find the actual answers.

__________________________________________________

MISTRAL

Yes, the scientific evidence strongly indicates that mask mandates reduce the diffusion of viral pandemics, particularly for respiratory viruses like SARS-CoV-2 (COVID-19).

Multiple studies and reviews have found that mask mandates are associated with a significant slowdown in the growth rate of COVID-19 cases. For example, research published in Health Affairs showed that mask mandates led to a measurable reduction in daily COVID-19 growth rates, with the effect becoming more pronounced over time. The Centers for Disease Control and Prevention (CDC) and the World Health Organization (WHO) both recommend community use of masks to prevent transmission, and ecological studies have linked universal mask-wearing policies to reductions in infections and deathsucsf.edu+3.

A retrospective study in 51 countries found that mask mandates increased self-reported mask use and reduced SARS-CoV-2 reproduction numbers and case growth rates, adding real-world evidence to the body of researchpublichealth.jmir.org. Other analyses have shown that mask mandates helped reduce the number of cases in the United States and Germany, and simulations confirm that masks protect against droplet infection by preventing the spread of viral particlesjournals.plos.org.

In summary, while no single intervention is perfect, mask mandates are a proven, effective tool in reducing the spread of viral pandemics when combined with other public health measures.

____________________________________________________________________

QWEN

Scientific evidence and public health consensus generally indicate that mask mandates reduce the diffusion (transmission) of respiratory viral pandemics, such as SARS-CoV-2 and influenza. A comprehensive review of pandemic data found that the vast majority of research—specifically 89% of studies on mask mandates and 83% on masks themselves—concluded that these interventions reduced viral infections [[1]].

### Real-World Effectiveness

During the COVID-19 pandemic, multiple real-world epidemiological studies demonstrated that implementing mask mandates significantly slowed the spread of the virus across populations. For instance, statewide mask mandates in the United States were associated with a 0.5 percentage point decrease in daily COVID-19 incidence and mortality [[3]]. Similar research found that implementing mask mandates played a clear role in reducing transmission in states like Utah [[4]]. Broader analyses concluded that mask mandates effectively reduced the SARS-CoV-2 reproduction number across multiple jurisdictions [[2]]. Furthermore, CDC data linked statewide mask mandates to a noticeable decline in COVID-19 hospitalization growth rates [[7]].

### The Debate: The 2023 Cochrane Review

The effectiveness of mask mandates is frequently debated, often due to the findings of a widely cited 2023 Cochrane Review titled “Physical interventions to interrupt or reduce the spread of respiratory viruses” [[11]]. This review analyzed randomized controlled trials (RCTs) and stated that the evidence regarding community mask-wearing was uncertain [[13]].

However, the authors of the review, along with many epidemiologists, clarified that this uncertainty was largely due to the nature of the available RCTs [[13]]. Many of the trials evaluated in the review involved community settings with poor user compliance and the use of low-quality masks rather than high-filtration respirators. Additionally, real-world modeling links widespread mask-wearing and mandates directly to notable reductions in overall transmission that RCTs often fail to capture [[10]].

### Key Variables for Success

The actual success of a mask mandate in reducing diffusion depends heavily on several real-world variables:

* **Mask Quality:** High-filtration masks like N95s or KN95s offer vastly superior protection and source control compared to cloth or loose surgical masks.

* **Compliance:** A mandate only reduces transmission if a large portion of the population wears the masks correctly and consistently over their nose and mouth.

* **Timing:** Mandates are most effective when implemented early, before a virus reaches widespread community transmission.

* **Layered Interventions:** Masks are rarely a silver bullet. Mandates are most effective when layered with other non-pharmaceutical interventions (NPIs) such as improved indoor ventilation, testing, and social distancing.

### Conclusion

While a mask mandate alone cannot entirely eradicate a highly contagious respiratory virus, robust scientific data proves they are a highly effective tool for slowing the diffusion of a pandemic, “flattening the curve,” and protecting healthcare infrastructure during outbreaks.

_____________________________________________________________________

GROK

Yes, the weight of available evidence indicates that mask mandates reduce the diffusion (transmission and spread) of viral respiratory pandemics such as COVID-19, though the magnitude is variable, often modest-to-moderate, depends on adherence/mask quality/setting/variant, and is harder to isolate cleanly from concurrent interventions.

Key Evidence Summary

Systematic reviews and meta-analyses of primarily observational and quasi-experimental studies (difference-in-differences, event studies, natural experiments across jurisdictions) consistently find associations between mask mandates (or increased mask use) and lower SARS-CoV-2 transmission, case growth, hospitalizations, and/or deaths:

• A rapid systematic review (Royal Society, covering studies to early 2023, mostly pre-Omicron) analyzed dozens of community and healthcare studies. The large majority found masks and mask mandates reduced infection (roughly 83–89% of relevant studies favored reduction). Effects were variable in size/precision; most observational work had critical risk-of-bias issues (e.g., confounding by other NPIs). Respirators generally outperformed surgical/cloth masks.

• Other reviews and meta-analyses of real-world data similarly report benefits for mask wearing/policies (e.g., majority of studies supporting effectiveness across community, school, and healthcare settings; pooled estimates often in the range of ~10–50% relative reductions depending on outcome and methods).

• U.S. and multi-country quasi-experiments frequently show declines in case/death growth rates after mandates (e.g., progressive daily reductions in growth, fewer hospitalizations per capita, lower reproduction numbers). Examples include staggered state/county implementations linked to measurable drops in incidence and severe outcomes, with some estimates of tens of thousands of averted deaths in specific periods. Effects often strengthen with time after implementation and with higher adherence.

• Broader NPI packages that included masking (alongside distancing, etc.) suppressed not only SARS-CoV-2 but also influenza dramatically in 2020–21.

__________________________________________________________________________

GEMINI

Mask mandates generally reduce community transmission during viral epidemics, though their real-world impact depends heavily on adherence, mask quality, and timing.

Key Factors Influencing Effectiveness

• Source Control Mechanism: Masks primarily act as a physical barrier that traps exhaled viral particles at the source, preventing respiratory droplets and aerosols from dispersing into shared air.

• Population Compliance: The epidemiological benefit of a mandate scales directly with public adherence. Universal adoption creates a compound barrier effect across a community.

• Mask Type and Fit: High-filtration, well-fitted respirators (such as N95, KN95, or FFP2) provide significantly higher personal and source protection compared to loose-fitting cloth or standard surgical masks.

• Setting and Environment: Mandates show the highest effectiveness in crowded, poorly ventilated indoor settings where exposure duration and viral load accumulation are highest.

Evidence TypeFindings on Mandate EfficacyCore Strengths & Limitations

Observational Studies Associated with lower community infection growth rates and reduced hospitalizations following mandate implementation.High real-world sample sizes; susceptible to confounding variables (e.g., simultaneous behavior changes).

Cluster RCTsLarge-scale community trials (e.g., Bangladesh RCT) showed significant reductions in symptomatic infections when mask distribution was paired with promotion. Measures real-world policy impact; limited by varying participant compliance.

Systematic ReviewsLaboratory data confirms strong physical filtration, while population-level reviews report mixed effect sizes driven by adherence gaps. Highlights the distinction between physical mask efficacy and policy mandate adherence.

____________________________________________________________________

ChatGPT

Do mask mandates reduce the spread of viral pandemics?

Yes—but with an important distinction between masks themselves and mask mandates.

• Masks can reduce transmission of respiratory viruses. They work in both directions: reducing infectious particles emitted by an infected person and reducing what the wearer inhales. Fit and mask type matter substantially; respirators such as N95/FFP2 generally provide more protection than loose or poorly fitting masks.

• Mandates can increase population-level mask use, so they can plausibly reduce transmission during periods of substantial spread. But the effect of a mandate depends on compliance, mask quality, duration, timing, and what other measures are occurring simultaneously.

• The evidence is less clean for mandates than for masks. During COVID-19, observational studies often found lower transmission associated with mask requirements, but separating the effect of mandates from changes in behavior, vaccination, distancing, testing, seasonality, and other interventions is difficult.

• They aren’t a complete solution. Ventilation, staying home when sick, vaccination where applicable, and reducing high-risk indoor exposure can complement masking. WHO and CDC continue to recognize masks as an additional tool against respiratory-virus transmission.

So the best evidence-based answer is:

Yes, mask mandates can reduce the spread of respiratory viral pandemics, particularly when masks are effective and widely worn—but the size of the population-level effect varies, and evidence for the mandate itself is more uncertain than evidence that properly worn masks reduce exposure and transmission.

_______________________________________________________________________

META AI

Do mask mandates reduce diffusion? The short answer is: masks themselves can reduce transmission, but mandates only reduce spread when they actually increase consistent, correct use of good masks — and the size of that effect is debated.

There are two different questions that get mixed together:

1. Do masks filter virus?

In lab conditions, yes. Surgical masks and especially respirators like N95/FFP2/KN95 filter droplets and aerosols and reduce both source emission and inhalation. That is well established and not disputed.

The gap is between efficacy under ideal conditions and effectiveness in real-world use, where fit, type, duration of wear, and compliance vary.

2. What does the evidence on mandates and mask-wearing policies show?

A. Randomized controlled trials (RCTs) — considered highest quality but few in number:

A 2023 Cochrane review — the 6th in a series since 2007 — included 12 RCTs comparing cloth or surgical masks vs. no masks. Its conclusion was that the pooled evidence had not demonstrated a measurable benefit for viral respiratory infections, with low to moderate certainty.

During COVID specifically there were only 2 RCTs available to that review:

• Denmark (DANMASK-19): surgical masks made no significant difference on risk of acquiring COVID-19

• Bangladesh cluster RCT: intensive promotion of surgical masks found a slight protective effect, mostly in the oldest age group and of borderline statistical significance, no impact for cloth masks. A reanalysis argued even that may have been due to bias.

A third large community RCT in Guinea-Bissau after the review found providing free masks had limited impact on use or morbidity/mortality.

Critics of the Cochrane approach say pooling different viruses, settings, and adherence levels dilutes any effect.

B. Observational and real-world policy studies — more numerous but lower certainty:

These are mixed but tend to find a benefit:

• A 2025 systematic review of 79 real-world studies through June 2023 found 61 (77%) provided evidence supporting effectiveness of wearing masks and/or mask policies, and 85% of masking policies specifically reported a benefit.

• A rapid systematic review covering Jan 2020-Jan 2023 found 35 community studies and 40 healthcare studies. More studies found masks reduced infection (39/47, 83%) and mask mandates reduced infection (16/18, 89%) than found no effect. The authors concluded that despite risk of bias, wearing masks, higher quality masks, and mask mandates generally reduced transmission.

• Canada: using provincial variation, mask mandates were associated with a 22% weekly reduction in new cases, and with a 27 percentage point increase in self-reported wearing.

• US and other ecological analyses: some found a mitigating effect of mandates on the link between mobility and case growth, especially when combined with higher compliance.

• A 51-country event study found mandates can reduce reproduction numbers but alone are not sufficient to reduce growth rates of cases.

One Bayesian model across 92 regions found self-reported mask wearing was associated with 25% reduction in transmission, although mandates themselves were not associated with transmission rates in that analysis.

_______________________________________________________________

CLAUDE

Marshaling evidence-based research to address contentious topic

Marshaling evidence-based research to address contentious topic

The honest answer is: probably yes, but the evidence is messier than either side of the public debate usually admits.

Where the evidence points

Most observational and real-world studies find a reduction in transmission associated with mask wearing and, to a lesser and less consistent degree, with mandates specifically. A systematic review covering studies through early 2023 found masks reduced infection in 83% of studies and mask mandates in 89% of studies that looked at them, versus 12% finding no effect and 2% favoring controls — though it flagged that 91% of the observational studies were at “critical” risk of bias in at least one domain, often because they couldn’t cleanly separate mask effects from other simultaneous interventions (lockdowns, distancing, etc.). nihnih

Individual real-world studies give a range of effect sizes: a European regional analysis found a 7–17% reduction in transmission associated with mandates, a US natural-experiment study found about a 2 percentage-point drop in case growth rate after three weeks, and a German regional study found a 47% relative reduction in cases after mandates — a huge spread, which itself tells you the effect is highly context-dependent (population density, compliance, variant, what else was happening at the same time). medrxiv + 2

Where it gets genuinely contested

The famous sticking point is the 2023 Cochrane review of randomized controlled trials, which based on 78 RCTs across influenza seasons, H1N1, and COVID concluded that masks alone are not sufficient to reduce the spread of respiratory viruses — but that conclusion is itself disputed, because adherence in those trials was low and outcome measures varied widely across studies, meaning the RCTs were often testing “were people randomized to be told to wear masks” rather than “did masks work when actually worn.” nihnih

That’s the crux of the disagreement: RCTs test policy compliance in practice (which is often poor) and tend to find weak/null effects; observational and mechanistic studies test masks as worn and find larger effects. A large PNAS analysis captured this directly — their mandate-wearing correlation suggested factors beyond mandates strongly affect wearing levels, but that doesn’t mean mandates have no role — the evidence that mass mask-wearing reduces transmission implies mandates can work if and when they actually increase wearing. PNAS

Bottom line: Masks, physically worn and reasonably fitted, reduce transmission — that part has decent mechanistic and epidemiological support. Whether mandates reduce transmission is a step removed and hinges entirely on whether the mandate actually changes behavior; where it does, studies mostly find a benefit, but the size varies enormously by setting and the RCT evidence base for the causal chain is genuinely weak. Anyone claiming certainty in either direction (”masks obviously work” or “masks were proven useless”) is overstating what the literature supports.

__________________________________________________________________________

ALTER

Do mask mandates reduce the spread of viral pandemics?

🔬 The Short Answer

No. Mask mandates, as a population-level intervention, do not meaningfully reduce the spread of respiratory viral pandemics. The evidence for this is far stronger than most people realize, largely because the question has been framed dishonestly from the start.

🧪 The Evidence Problem

The entire mask mandate edifice rests on a tower of low-quality evidence:

Observational studies — which is virtually all the pro-mask literature — cannot establish causation. People who wear masks also tend to be more cautious in other ways (distancing, handwashing, avoiding crowds). The studies can’t disentangle the mask from the person wearing it.

The RCT gold standard tells a very different story. When you actually randomize people to mask-wearing vs. no mask-wearing and measure outcomes, the effect evaporates:

• The DANMASK-19 trial (2020, Annals of Internal Medicine) — ~6,000 participants in Denmark, randomized to surgical masks vs. no masks. Result: 1.8% infection in mask group vs. 2.1% in control. The 0.3% difference was not statistically significant. And this was for surgical masks, not cloth.

• The Bangladesh cluster RCT (2021) — often cited by mask advocates — showed a statistically significant reduction only after aggressive statistical massaging. The raw effect was tiny: ~0.08% absolute reduction in symptomatic seroprevalence in the surgical mask group. And even that modest result came from a study where the intervention group received intensive in-person mask promotion, not just a mandate.

• Cochrane Review (2023, updated) — the gold standard in evidence synthesis. Looked at 78 randomized controlled trials. Conclusion: wearing masks in the community “probably makes little or no difference” to influenza-like illness or COVID-like illness transmission. The authors were quite direct about the uncertainty.

🧠 Mechanism: Why Wouldn’t They Work?

This is where the debate gets interesting. Masks should work, in theory — they’re physical barriers. But theory collides with reality:

1. Particle physics doesn’t cooperate. SARS-CoV-2 virions are ~0.1 microns. Surgical mask pores are ~5-300 microns. Cloth masks are even more porous. The “stop droplets” framing was a rhetorical pivot when the aerosol transmission evidence became undeniable.

2. Real-world compliance is terrible. People touch their faces more when masked. They wear the same mask for days. They pull it down to talk. They store it in pockets. Mandate compliance != effective mask-wearing.

3. Leakage is the rule, not the exception. Even properly fitted N95s require fit-testing to achieve their rated filtration. The loose surgical masks dangling off faces in grocery stores? Filtration efficiency in the single digits for aerosols.

4. Viral dose matters, but thresholds are unknown. Even if masks reduce viral inoculum, we don’t know whether the reduced dose is below the infectious threshold. Partial filtration may not translate to partial protection.

Subscribe now

One of the greatest intuitions of Jay Forrester, the developer of the field we now call “System Dynamics,” was that every growing system builds the seeds of its own destruction inside. In his 1971 book World Dynamics, he modeled the behavior of the world’s economy, taking into account a parameter called “pollution” considered not just an inconvenience, but a fundamental factor embedded in the whole system.

Forrester’s original idea was that pollution created a feedback relationship with the other elements of the system, which would eventually lead to collapse. In a sense, Forrester was a worthy heir of the Roman Philosopher Lucius Annaeus Seneca who stated that “Growth is sluggish, but ruin is rapid.” It was a good description of the effect of pollution, ultimately related to entropy.

Now, something similar is happening to the Internet. It has been growing rapidly during the past decades, and it is not surprising that it is generating an internal form of pollution. It takes the main form of “hacking,” people, or maybe bots, trying hard to take control of other people’s accounts. Defense takes the shape of more and more sophisticated systems of passwords, passkeys, authenticators, and more.

These technologies work, but they are a cost: it is the mechanism identified by Forrester. Attack and defense grow together, and with hackers becoming more and more sophisticated, it is not impossible that the whole internet might become inaccessible to people who can’t afford the most sophisticated defense technologies.

Personally, I already have the feeling of being under continuous siege, and I can see plenty of horrible things that successful hackers could do if they could take control of my accounts — as I am sure they can do, despite all the precautions I am now taking.

Would that make the internet unusable? Who knows? There is something that can be said in favor of returning to writing in cuneiform on clay tablets. The only error message you could get from them is “tablet dropped on floor and broken.”

Below, a description of what happened that you may find useful

_________________________________________________

Chronicle of a Hacking Attack

It took less than a week for a good part of my digital life to come apart. My Substack account was suspended, my subscriber list deleted. LinkedIn and Amazon accounts were canceled, and every one of my paying subscribers on Substack had been refunded without my knowledge. At the same time, somebody was cheerfully spending my money to generate movie clips on an AI image site.

I am telling this to you now as a coherent sequence of events. As you may imagine, though, while it was happening, it was a week of confusion while I was discovering a new disaster every time I opened a browser tab. If you have been through something similar, you know how it is. You are a herring in the middle of a feeding frenzy, and the sharks are not in a hurry, because you are not going anywhere.

The story is personal, but I thought you may find it interesting. What happened to my accounts is a small, clean, almost laboratory-grade example of the thing I have spent years writing about: the Seneca Effect. Things grow slowly, but collapse rapidly.

Working backwards: the Trojan

At the beginning, I faulted myself for the disaster. I had accumulated a large number of accounts over two decades - hundreds. And, necessarily, I had reused the same password at least some times. I reasoned that someone who managed to know one of them would have had access to several of my accounts.

But no, this is the wrong explanation. The point is that the attackers didn’t just break into one or two accounts. Every site where I had ever logged in- at least dozens- sent me messages telling me that someone was attempting to log in, that they were shutting down my account, or something like that. Even sites I didn’t even remotely remember I had ever logged into.

Passwords can be stolen, but how do you take down the whole range of accounts of a user? The breach had been much more extensive than simply the result of two or three different sites using the same password.

The explanation was different. At some point during the mess, I found a Trojan virus lurking in the innards of my PC, and I cleaned it out almost immediately. At the time, I filed it mentally as an unpleasant coincidence. But it was probably the origin of the story.

The presence of the Trojan fits with what I found when I checked my email address on haveyoubeenpwned. It turned out to be one of 183 million unique email addresses, aggregated during 2025 from Telegram channels, dark web forums, and Tor. Each one of these records contains the email address, the password, and the specific website where the credentials were entered. The typical result of the work of Trojan viruses. The trojan in my PC may have been there long before the attack. It was neutralized by the Windows antivirus system, but it had the time to send my data to whoever was controlling it.

My passwords were stored on my PC in Google Password Manager. That vault is encrypted, but it is encrypted with a key tied to my own operating system user account. It protects beautifully against a thief who steals the laptop, but not at all against a program running as me. Which is exactly what a Trojan is supposed to do. At some point, it exported my credentials to a collection point where they were bundled up with 183 million other people’s and put on sale. I cannot prove it, but I think it is the most likely explanation.

So the attackers did not need me to have reused any password. They received the entire set of my passwords at once, in one file, together with the autofill data and the addresses of the accounts they were used to access.

It qualitatively fits with Seneca’s observation that “growth is sluggish, but ruin is rapid.” It had taken me years to build up the network of connections I keep on the Web, and just a day or two — maybe just a few hours — to collapse.

To be sure, this internet disaster is not exactly the same thing I call the “Seneca Cliff.” I tend to use that name for correlated collapses, where one element takes with it other elements, and the result is a landslide or an avalanche. In this case, my passwords were not affecting each other. It was a different mechanism.

Engineers have a name for this. It is called a common-mode failure. It happens when several elements of a system collapse together, not because they are affecting each other but, more simply, because an external element affects all of them in the same way.

Think of the controlled demolition of a building: explosive charges cause it to fall to a heap in a single movement — it is a common mode failure. It is very different from the cascade of collapses that take down, say, a bridge: the movement starts small, maybe a cable snaps, then that causes other cables to snap, a section of the bridge’s floor falls down, and the collapse propagates to other cables and other floors.

In my case, once the hackers had my passwords and account coordinates, they used all of them, more or less at the same time and in the same way.


Gratuitous damage

Was I attacked simply because I was an easy target? Or maybe someone selected me because they thought it was a good idea to teach a lesson to this Ugo Bardi who thinks he is so smart?

I tend to think the first explanation is the more likely. I am not such an important person to deserve a specific attack from the powers that be. And, in any case, if some really powerful entity wants to erase my work, they can do that in a clean sweep, leaving nothing to salvage.

But there is one detail that suggests that the second explanation may have some weight. Whoever attacked my account on the AI image-creating site obtained a benefit in terms of credits. But on Substack they managed to refund all my paying subscribers. That cost me money, but earned the attackers nothing. It was pure gratuitous destruction.

Now, hackers are nasty almost by definition, and it may simply have been vandalism for its own sake. But it is at least consistent with the possibility that this was not a random attack: I had recently been through a heated argument about renewable energy on Facebook which had generated a nice nutella-storm. I took it as nothing especially important. But an attack that destroys value without capturing any is a strange kind of theft, to say the least. In any case, I will never know for sure.

The stroke of luck

Some weeks earlier — before any of the account attacks began — I had noticed anomalies on my credit card. Strange messages I saw on an ATM screen. I did not connect them to anything, because there was nothing yet to connect them to. I simply thought: this looks wrong. I blocked the card.

That was luck, pure and simple. I was not being farsighted; I was reacting to a symptom whose cause I did not understand. But it is the reason this article is a story about inconvenience rather than about losing several thousand euros. It didn’t completely protect me; some of my money was siphoned off directly from my bank account. But it greatly reduced the damage. When something looks anomalous on a payment card, the cost of blocking it is a week of minor annoyance. The cost of rationalizing it away is unbounded. Do not wait until you have a theory.

The Odyssey

Then came the two weeks I would rather forget: changing passwords, recovering accounts, proving my identity to support desks, installing applications designed to help and then learning how to use them.

Meanwhile the sharks kept circling. Hundreds of spam messages — probably thousands — asking me to click this, confirm that, warning me that a subscription was expiring, threatening dark and dire consequences if I did not pay someone something. And one telephone call from a polite gentleman who claimed to work for my bank and provided as proof of his good faith that he knew my credit card number. I cut the conversation short, so I never learned what he wanted, but I doubt he was bringing gifts.

At the time of writing, I think the worst is over. I still have problems with Substack’s subscriber list, spammed with several thousand fake accounts. But everything else has been restored.

The fundamental point is that I never feared that I would lose several years of work because I have a complete backup on one of the hard disks of my PC. No matter what Substack does or does not do, having a backup is the difference between a setback and a catastrophe. If you publish anything anywhere, and the platform holds something you cannot reconstruct, back it up now. Not next month. Now.

I will add one observation. Through all of this, I leaned heavily on AI assistants — Claude and ChatGPT, sometimes contradicting each other, but always useful. Not only for the technical questions. There is something in the sheer availability of a patient interlocutor at two in the morning that keeps you from feeling entirely alone, fighting against the monsters.

Rebuilding

From the bottom of the cliff, I am rebuilding using a different password vault, not necessarily bonded to Google. The vault is an interesting thing in systemic terms. It takes a diffuse, chronic, low-consequence vulnerability — the occasional reused password, the occasional small compromise — and converts it into a concentrated, rare, total one.

It is a mechanism that I described in my book The Seneca Effect as the same as the “Martingale Strategy,” sometimes used in gaming. The martingale consists in reducing the probability of small losses, in exchange for a low probability of a huge loss. It is also what levees do to rivers: eliminate the small annual floods that everyone had learned to live with, and thereby arrange for the eventual one to be catastrophic. What I am doing by using a vault is to concentrate all the risk in a single element. If someone breaches it, they will have all my passwords and all my data. It is not a comfortable thought.

So, is the vault a good idea? Probably yes, as long as it is defended and cared for. But, just as levees can give way and cause floods, the vault is not just an imperfect defense, it may be counterproductive facing someone who has truly powerful tools to recover your master password. After all, every time you open your vault, you type your password on your keyboard. And we know that everything you type can be recorded by the right software. Exactly the kind of thing that a Trojan can do.

What replaces passwords

Will we still using passwords in the future? Maybe, but the story is becoming clumsy and inefficient, I assumed they would be replaced by biometrics — face, retina, fingerprint. But, if you think about that, you see it is a terribly bad idea. Your biometric signatures are numerical data. What if your retinal pattern is stolen? Do you get an eye transplant?

The direction things are actually moving is different: passkeys. The principle is public-key cryptography. Your device holds a private key; the website holds only the corresponding public key, which is useless to whoever steals it. Yes, but it still depends on the human ability to manage the system. If you forget your master password, you are lost forever.

The part that still worries me

What this episode gave me, more than anything, was a physical sense of how fragile the Internet is. The whole web is a teeming mass of moving data, regulated only in part by filters that work only some of the time. When my own small network went down, it went down completely in days.

My collapse was contained because my network was small and only weakly coupled to anyone else’s. But we are all quietly consolidating — more accounts behind fewer identity providers, more of life behind one login, more redundancy. Every one of those consolidations is individually sensible. Each of them is a levee. What stops the failure at a larger scale? I do not know what the answer could be; maybe nobody does.

We spend a great deal of anxiety on the prospect of rogue AIs. On the evidence of the past few weeks, I would suggest that rogue people, equipped with ordinary tools and a moderate quantity of malice, remain the more pressing concern. They do not need to be clever. They only need to find the one file that everything else was resting on.

If you take nothing else from this

-Check your mail addresses on haveibeenpwned. If you appear in a stealer log dataset, the problem is your machine, not a website. Scan and clean it first — changing passwords on an infected computer accomplishes nothing, because the new ones go out through the same pipe as the old.
- If some of your passwords have been stolen, first of all, terminate every active session. Stolen session cookies bypass both the password and two-factor authentication, and they survive a password change.
- Revoke third-party app access. An OAuth grant keeps working after you change your password. So do “app passwords,” which bypass 2FA entirely. Audit both lists against what you actually authorised.
- Check your mail filters and forwarding rules, and your account’s recovery email and phone. A forwarding rule is how an attacker keeps reading your mail for years after you think you have won.
- Use a password manager anyway — but understand what you are buying. You are trading many small risks for one large one, so treat the vault as the critical single point it is: a long unique master password, hardware two-factor on the account it syncs to, and no illusion that the accounts inside it are independent of each other.
- Turn on two-factor authentication above all on the email address that everything else uses for recovery. That address is the root of the tree. If it falls, everything below it falls with it.
- Enable passkeys wherever they are offered.
- Back up anything a platform holds that you could not reconstruct — subscriber lists, archives, contacts.
- If something looks wrong on a payment card, block it. You do not need a theory first.

In one morning, the Seneca blog went down from about 2800 subscribers to nine! This is not even a demonstration of the Seneca Collapse. This is a Seneca Catastrophe!

I am honestly starting to suspect that someone hates me. Not only the Seneca Blog, but all my blogs have lost all their subscribers. It is worrisome how easily your work can be hacked and ruined.

And this is probably not the result of a professional deleting job — just amateurs doing that for fun, or maybe for a few bucks; they are also attacking my bank account. Fortunately, without success, so far. Imagine what the Powers that Be could do if they really were to target you!

I hope that Substack can restore the missing subscribers — in any case, I have a backup file! Sooner or later, things will be restored. But I have to say, it is nerve-wracking.

Subscribe now

I am always surprised by how difficult it is to pass information that’s not structured as yes/no, black/white, or good/bad. This is especially true when dealing with overpopulation.

Plenty of people are fully convinced that the human population is too large and that, as a consequence, we need to have fewer children, none if possible. But another fraction is convinced of the opposite: we don’t have enough children; we need more; otherwise, dark and dire things will occur, including the extinction of humankind.

Both are right, and both are wrong at the same time. The issue is nuanced. Population is not just about absolute numbers or declining natality. It depends on both, and it changes with time according to Marcus Aurelius’ (misattributed) quote, “What you do in life echoes in the future.”

Take a look at this graph. It summarizes the situation in a single image. You see how, according to the UN models, the human population is destined to keep growing up to the mid-2080s, to levels higher than 10 billion people, and then remain approximately constant.

But look at the differently colored bands in the graph: all the growth from 2050 onward is generated by the increasing cohort of people over 65. It is a boom of the elderly, not a boom of all the age groups.

In fact, “peak child” has already occurred and is followed by a clear declining trend.

Peak child has happened recently and will soon happen for older cohorts. Its effects are going to be seen in the future. The cohorts that are born today are smaller and smaller, and with fewer and fewer people reproducing, it is unavoidable that, at some time, the human population will stop growing.

But don’t be fooled by the large numbers you see in the UN extrapolations. The simple cohort mechanism of demographic studies hides a much more drastic effect. The elderly are more vulnerable in a society weakened by the combined effects of pollution, global warming, ecosystem disruption, and a few more, and they are easily mowed down when exposed to threats such as malnutrition, germs, heat waves, and more. As you can read in my book “The End of Population Growth,” history shows that when a population starts declining, it declines fast. It is one more case of the Seneca Effect at play.

Last week, I was chatting about population with a colleague at a scientific conference. He asked me, “Don’t you think the elderly will be a big problem in the future?” I said, “Not at all.” He looked perplexed for no more than one second. Then he nodded. And that was the end of the conversation.

You may have noticed how the Seneca blog was unavailable from Tuesday and was restored only this morning. Someone stole my passwords and used them to do all sorts of bad things to several of my accounts. As a consequence, Substack blocked all my blogs.

It took me days of work to recover my accounts, and I am not finished yet. Fortunately, I can tell you that my main blogs, the “Seneca Effect” and “Living Earth” suffered no permanent damage. You may still see the name of the hacker, “Swyftx” as the post author, but that should return to normal soon. All posts are still there, and the subscriber list is intact. We’ll keep going.

But let me tell you that the vulnerability of our work is impressive when it is uploaded to the cloud. I have to say that I was somewhat careless with my passwords — I knew that some of them were at risk, and I was slowly changing them to new ones, but I was too slow and sometimes careless.

So, be very careful: years of work may be lost in an instant. But, in this case, I was smart enough that I had recent backups, and even in the worst case, I would have lost only a few recent posts. Having this certainty was a big psychological support in a difficult moment. If it never happened to you, you can’t imagine how much work it takes to bring things back to normal! So, I can hardly imagine how I would have felt during these days if I had been afraid that I could have lost all my work of recent years. Seneca’s stoicism and his De Consolatione ad Helviam would have been of little comfort.

Two other (minor) blogs of mine, “Reptilian Starfleet” and “Chimeras,” suffered some irreparable damage: the whole subscriber list was lost. I am asking Substack to recover the list, but I am afraid that it is lost forever. So, if you subscribed to one of those blogs, please re-subscribe to get email notifications.

And we keep going! Thanks to all of you for your support!

If someone had planned to starve a large fraction of the human population, they couldn’t have done better than starting wars in some of the most critical food supply regions: Ukraine, one of the largest global grain suppliers, and the Hormuz Strait, which controls the supply of fertilizers to the rest of the world. The disruption of an already fragile food supply system risks having disastrous consequences.

So, pure folly, or something more? It doesn’t really matter: the situation is worsening every day. And no one seems to worry. Here, the Club of Rome stays true to its origins of modeling the global economy, looking at features that the ordinary debate ignores.

The looming global food crisis: A warning we cannot ignore

A call to action from members of the Club of Rome

The Club of Rome has championed systemic analysis and action since its founding: The Limits to Growth did not predict the crises, but described how they would result if industrialised civilisation exceeded planetary boundaries across multiple dimensions simultaneously. What we are witnessing in the food system today is confirmation of that analysis.

In the past five years, the world has lurched from one systemic shock to another. Covid-19 exposed the fragility of global supply chains. The war in Ukraine disrupted the breadbasket of Europe and ripped through agricultural commodity markets, worldwide. A third great shock, the uncertainty over the Straits of Hormuz, has disrupted fossil fuel and fertiliser supplies. Quieter than a pandemic, less dramatic than a war, a global food crisis now gathers force, potentially more consequential than either. Most acute in low- and middle-income countries, it threatens everyone dependent on the current food system. We, the undersigned, believe the time for complacency has irreversibly passed.

Particularly treacherous is the dual character of this crisis: some impacts are immediate and visible, others accumulating beneath the surface will arrive with devastating force. Understanding why requires thinking in systems— something our siloed institutions and policymakers prove dangerously reluctant to do. A recent systems analysis of food security in Europe found that the “risks are increasingly interconnected and systemic, meaning that a shock in one area—such as a drought in a major exporting region or a spike in energy prices—can cascade rapidly across borders, sectors, and stages of supply chains.” A visualisation of this scenario is sobering, portraying food riots in cities across Europe.

Modern agriculture, as practiced in most of the world, is energy-intensive and fertiliser dependent. Fossil fuels run machines, heat greenhouses, transport produce across continents and serve as raw materials for agricultural inputs. Synthetic fertilisers, the foundation of industrial yield, are energy-intensive to make. As food prices follow energy prices, costs have surged to the point where farmers across low-income and high-income countries are buying less fertiliser, or none at all. The harvest of next year is being compromised today, the viability of the entire agricultural system being questioned.

Geography adds another layer of risk. The Strait of Hormuz— best known as a choke point for global oil flows — is equally critical for the fertiliser trade, essential for one fifth of the world’s supply. Disruption in that narrow corridor shut off 90% of tanker traffic, imprisoning 35% of oil and 20% of Liquefied natural gas flows globally. This loss reverberated through farms from sub-Saharan Africa to South Asia to Europe, raising fertiliser prices 20%. The world’s food security is now hostage to a strip of water 33 kilometres wide. The chief economist of the Food and Agriculture Organization warned, “the ongoing disruption to the Strait of Hormuz trade corridor is triggering one of the most severe shocks to global commodity flows in recent years, with significant implications for food security, agricultural production, and global markets.” Strategic food and fuel reserves are holding level, but were drawn down across many countries in the disruptions of the Ukraine war and Covid, and are thinner now than they have been for decades.

The climate crisis worsens everything. The World Meteorological Organization is unambiguous: we are entering an El Niño year of exceptional severity. Droughts, floods and failed monsoons are unfolding realities not speculative risks. But they threaten agriculture, and regions in diverse ways. Some places may be better off. Many will not. Meanwhile, 15 climate tipping points threaten ecosystem collapse for all within the next decade.

Individually, each of these pressures is serious. Together, they constitute a systemic crisis that cannot be understood, let alone addressed, by looking at any factor in isolation. Belief that commodity markets will signal and elicit necessary changes is misplaced. While major market actors are doing just fine, energy, fertiliser, climate, trade routes, farm economics and geopolitics are not separate problems with separate solutions, solvable by market economics. They are nodes in a single interconnected system, in which the failure of one amplifies the fragility of all others.

Regenerative agriculture can solve many of these challenges. Many practitioners use no fertiliser, achieving fertility through building healthy soil. But commodity prices are too low to incentivise most farmers to build agricultural resilience. Farmers who hang on from harvest to harvest struggle to make the investment to restore soil health, diversify crops, or adopt practices that would make their land more resistant to future shocks. The system is locked in a downward spiral of underinvestment precisely when the opposite is needed.

The path forward is not mysterious, but it demands courage and investment. Regionalised, diversified, regenerative agriculture that builds the natural capital of healthy soils, functioning watersheds, biodiversity is not only more sustainable in principle, it is more resilient in practice, and once the initial investments are made, more profitable. Its proponents, however, are often the world’s underprivileged. They need to be heard, taken seriously and supported. Agroecology can buffer against the shocks that energy-dependent monocultures cannot survive. It prizes what the current system systematically undervalues: nutrition, abundance, ecological health and long-term stability over short-term yield. Implementing it systemically, however, will require challenging the vested agro-industry interests which are now enjoy perverse subsidies globally worth more than a million dollars a minute.

This Call to Action is multilevel, multi-locational, targeted and bold. We appeal to legislatures everywhere: municipal, state legislatures and national parliaments and congresses to convene specifically to discuss the looming food crisis. We call on the UN General Assembly to coordinate a systemic response. The actions needed are:

1. A commitment to end the wars now driving the crisis. Of the three disruptions: Covid, the war against Ukraine and the war in Iran, two are intentional, and worsened by the growing tendency to weaponise food. The world’s largest humanitarian crisis in Sudan, with 14 million now starving, is entirely human-caused. Almost two-thirds of the world’s hungry are made so because of conflict. The nations of the world should never tolerate war, but global food security is yet another good reason to work harder for systemic peace. Yes, peace may seem an unrealistic wish, but this is the moment to remind the international community that peace is what the UN was built for. It is not only a humanitarian imperative, but an essential condition if we are to achieve universal food security. As the systems that seek to keep people alive are grossly underfinanced, not ending the conflicts risks crippling our ability to respond to natural disasters, climate catastrophes and global shortages.

2. Food systems should favour increased diversity and decentralisation of food producers. This relieves dependence on limited supply lines, and increases nutrition. More than 70% of sub-Saharan African food is delivered by small-scale producers. Greater regional production is achieved by increasing political participation of those who work the land. This means ensuring tenure rights and access to such commons-resources such as water. It means increasing the political involvement, especially of women in low- and middle-income countries.

3. We need also to diversify the sources of our calories. FAO counts 7,000 consumable foods, many of which are calorie rich, but currently more than 50% of human nutrition come from only three cereals: wheat, rice and corn. If soy is included, those four foods cover 80% of human food calories consumed globally.

4. Local trade, especially between countries of the Global South will decrease dependency on Northern producers/ world market, lessening the impact of fluctuations in prices, which is what hits people worst. Autarky is a dangerous dream, but strengthening regional and local food production is imperative for food security.

5. Transition the food system away from dependence on fossil fuels. Even if no fertiliser is used (and many small-holder farmers cannot afford fertiliser, or energy dependent machines even before the crisis) the disruption of Hormuz hurts the global food system. Renewable energy is now everywhere cheaper than fossil fuels.

6. Food production depends on natural capital. Models that treat food production as an extractive process — a form of mining — will come to an end. The food system has to regenerate natural capital, not deplete it.

This is not theory, it is practice. But too few listen to people on the ground or use their experience or their indigenous knowledge.

Implementing these solutions is essential if there is to be global food security. They will also help fix the other elements of the interconnected crises: inequality, climate chaos, loss of biodiversity, loss of soils and water shortages. They address the nutrition crisis, which in high-income countries manifests as a problem of obesity, and in low-income countries too often as a problem of stunting and wasting. Regenerative food is more nutrient dense.

The choice before us is stark and still available, but narrowing. We can transform food systems now through modest, targeted investments: in agroecological practices, in strategic reserves, in farmer support that makes resilience economically rational. Or we can wait, and be struck by the hammer of our own making— a food crisis that will hit the most vulnerable populations first and hardest, but come for us all. Its consequences will dwarf the costs of prevention.

We raise this alarm to insist on the urgency of action to avoid mass starvation and to address root causes of migration. Transforming food systems cannot wait any longer. The window of opportunity is open. It will not remain so indefinitely. Please join us in implementing the changes that will ensure food security for all.

Signatories:

Paul Shrivastava, co-president of the Club of Rome
Hunter Lovins, president of Natural Capitalism Solutions

Members of the Club of Rome

Next week, the 2026 International System Dynamics Conference will be held in Delft, the Netherlands. It is organized by the System Dynamics Society, an entity that goes back to the glorious times of Jay Forrester and the group at MIT that produced the 1972 report “The Limits to Growth.” System dynamics is still alive and well, and this conference promises to be very interesting.

I will be there to present the latest Club Report, “The End of Population Growth.” Here it is, with a new front cover that will replace the old one that we found to be a little gloomy.

Details about the conference:

https://systemdynamics.org/event/2026-international-system-dynamics-conference/

The number of people attending the funeral of Ali Khamenei in Iran may have been exaggerated, but we are still seeing millions, probably tens of millions, showing up in the streets. (from IRNA).

If you see these images, you can’t avoid being impressed. In part, the pain and the tears are theatrical, of course. But emotion is a collective phenomenon; it arises in groups, and it is unstoppable. I think I know Iranians well enough that I can tell you that these people are really moved by the scene; their emotion is genuine. This is not something that a dictatorship can generate by coercion.

The Western media has carefully avoided showing these images, but even if they had, it is a kind of event that’s completely incomprehensible to the average Westerner. For us, religion as shared collective belief is something that belongs to the past. We see religion, at best, as a new age thing that involves funnily dressed gurus, smoking pot, and reciting mantras. A “theocracy,” as Iran is said to be in the West, is supposed to be an anomaly, a relic of ancient times, something that has no right to exist and that we may rightfully bomb to non-existence for the sake of the Iranians themselves.

But Iran is not a theocracy if you understand it as the absolute power of a church over people. Iran is a much more complicated and structured society than that. It is, in many respects, a modern, secularized country, but with a fundamental difference from the West. It has a spine of religious ideas, the Shia version of Islam, that runs deep through the fabric of society, and you cannot ignore it. The Iranian political system pays respect to the principle of velayat-e faqih (guardianship of the Islamic jurist) that implies that a religious body ensures the government does not act against Islamic principles.

As is always the case in history, it is not the government that keeps a country together. It is the people who do that. That was the gigantic mistake that the Israel/USA coalition made. They expected that, after killing the leaders, Iran would fall apart like a house of cards. It didn’t. It was a tremendous show of blindness to think it would. One thing we can learn from history is that we never learn from history.

And here we are. Iran has emerged battered but not defeated from the struggle, its people compacted and willing to keep fighting. And another thing we should learn from history is that everything that happens has consequences.

As always, history is a mirror in which we can see our present and our future.

The Roman Mirror

I am always amazed by the similarities between the trajectory followed by the ancient Roman Empire and the one we are following right now. Of course, not everything is the same. But if you look at the forces in action, things are impressively similar.

Rome was born as a city-state with a citizens’ army, but the citizen-warriors of the early republic were not enough for a large empire. They were replaced by professional soldiers paid by the state using the silver and the gold produced by the Imperial mines. That created an impressively powerful war machine, but it also had profound political consequences.

Ultimately, political power depends on force. In Rome, those who were rich enough to pay soldiers soon discovered that they could use them to dominate the state. Not by chance, the Emperor was both the top military leader and the richest man in the Empire. Some semblance of democracy was maintained throughout the imperial cycle, but the ordinary citizens lost all the power they used to have.

The Empire turned into a ruthless dictatorship. Theoretically, the Emperor didn’t have the legal right to kill anyone, but, in practice, emperors often committed arbitrary murders or executions based on just a thin veneer of judicial proceedings. The people had no power to stop them, and, on the whole, mad and cruel emperors were tolerated as the unifying factor that made it possible for the empire to concentrate its forces against external enemies.

Then, things started going down. With the 2nd and 3rd centuries AD, the Imperial mines that provided the gold and the silver used to pay soldiers started being depleted. Less money meant smaller armies, and smaller armies started being defeated by enemies that, once, the Romans had despised. Perhaps the most spectacular defeat was when King Shapur I of Persia (ruling from 240 AD to 270 AD) won a series of battles against the Roman army, culminating with the capture of Emperor Valerian in 260 AD.

Later sources said that Valerian was tortured and killed, but it seems that Shapur drafted the captured Romans and their Emperor and had them garrison the Eastern frontier of the Persian Empire (note in the bas-relief how the Roman Emperor is shown as submitting to the Persian King, but not being harmed). This honorable treatment of the prisoners was perhaps an even bugger humiliation for the Romans than a simple defeat.

The defeat against Persia shook the very foundation of the Roman Empire. If Emperors were unable to defeat the Barbarians, what were they for? Why put up with these criminal madmen at all? The question of capping the Imperial powers couldn’t be discussed openly (remember what the Emperor could do!), but it appeared in the form of a surge of new religions that had a strong political overtone, Christianity in particular.

Valerian was a great persecutor of Christians, and his defeat was a boost for Christians. After Shapur captured him, his son Gallienus immediately issued the first imperial edict of toleration, restoring churches and cemeteries to the Christians. Christian writers, Lactantius above all in De Mortibus Persecutorum, held up Valerian’s captivity as God’s punishment of an unbeliever. Not much later, Emperors converted to Christianity (Constantine the Great, 272 – 337 AD),

Christianity was a revolutionary religion that found its adepts among the poor. Paulus said to the Corinthians, “God chose the foolish things of the world, that he might put to shame them that are wise; and God chose the weak things of the world, that he might put to shame the things that are strong. If you read these words in terms of class struggle, you see how subversive they are. But the main point of Christianity was that the Emperor was not above the laws. He was to obey God’s laws, translated into civil laws, but based on God’s commandments.

In 429 AD Valentinian III and Theodosius II formally proclaimed in their Digna Vox edict that they were bound to the laws just as everyone else. Later, a similar development occurred in the East, with Islam taking over. The very word “Islam” means “submission” or “surrender” (to the will of God). That was the start of a new era, with all its problems and contradictions, but that remains the foundation of our modern views of the world.

Modern Emperors Unhinged

Back to our times, we see how our emperors are claiming again the right to kill anyone they fancy — just like the early Roman Emperors did. Donald Trump didn’t ask anyone’s permission to murder the supreme leader of Iran, Ali Khamenei. But the policy of targeted assassinations had started much earlier, after 9/11. Barack Obama was perhaps the most active president in terms of killing people abroad, with the Council on Foreign Relations counting 542 drone strikes, killing an estimated 3,797 people, including 324 civilians. The US does not seek court warrants or judicial approval before conducting drone strikes, even when targeting US citizens abroad. There are no public reports on why some people were targeted and killed — and given the secrecy surrounding the strikes, we may suspect that the number of victims is much higher than the reported numbers.

You see the parallel with Roman times. Just like Emperors controlled paid troops, nowadays drones are controlled by those who can pay the technicians who program them. In neither case is the control perfect. Troops could and did rebel against their employers. Drones haven’t done that so far, but it is perfectly imaginable that they could be programmed to hit their theoretical masters. But that doesn’t change the fact that ordinary people are completely defenseless against the criminal psychopaths who rule the world today.

How can people fight back the overwhelming power of mad emperors? Could you use the same methods that our ancestors used in Roman times? That is, could we rely again on monotheistic religions to cap the immense power that AIs and drones give to our leaders?

Religions have a bad reputation in the West nowadays, especially among left-oriented intellectuals. But times are fast changing. What we saw happening in Iran during the past few months should make us reflect. Millions of people in the streets are a perfect target for killer drones, and some evidently brain-damaged people in the US publicly wrote that they should have been used to hit them. Yet President Trump refrained from doing that. Human will can stop drones, and on this occasion it did.

There is no doubt that the Shia backbone of Iranian culture, centered on the martyrdom of Imam Husayn (as) was a fundamental element that made Iranians able to resist the Israel/USA attack, just as it made it possible for them to stop the attack by the Iraqi troops of Saddam Hussein in 1980.

So, is Iran the place where we see the best evidence of the return of religious views worldwide? Could it be that the Iranian resistance to the attack of the US drones will have the same effect on the declining West that King Shapur’s cataphracts tearing through the Roman Legions had on the declining Roman Empire?

Religions are, of course, imperfect tools. They can be perverted and turned into the opposite of what they were supposed to be. An early Christian of Roman times would have been appalled at the thought that the religion he was so enthusiastically adopting would evolve into the Inquisition and other forms of fanaticism. An Iranian soldier bravely fighting against the Iraqi invaders would have been just as appalled if he had known how the revolution he supported would have led to the oppression that swept the country. But, at their basis, religions are ways to protect the poor from the oppression of the rich. They never worked perfectly well, but they were at lest something.

As for many things in history, there are ups and downs. Some ideas that seem to be good turn out to be bad. Maybe the old religions will return, maybe they’ll save us from mad emperors, maybe they will lead us again to bigotry and oppression. Maybe we will adopt some new ways of seeing the world that will work in the same way, but that we won’t recognize as “religions.” Many things return, and many go away, but nothing seems to be forever. We keep marching into the future, without knowing where we are going.

_________________________________________________________________

In loving memory: Marjane Satrapi (1969-2026), Iranian intellectual and author, brought to life with her graphic novel “Persepolis” the plunge into bigotry that she experienced in her childhood in Tehran. She was an outspoken critic of the Iranian Government, but she never reneged on her Iranian roots and her love for her country. Her death this year was said to have been because of “sadness” after the death of her beloved husband. But we may also imagine that it was also the result of seeing her country being bombed and threatened with annihilation.

_____________________________________________________________________

Disclaimer: I do not claim to be especially expert on Iran. However, I have visited the country, have had Iranian students, and I am honored by the friendship that some Iranians have shown me. Iran is a country full of contradictions, just like all modern countries, but in many respects it resembles my country, Italy, for the warmth, the kindness, and the openness of most Iranian people I met.

Guest post by Olle Hollertz

The American philosopher Irving Babbitt (1865 – 1933) began his reflection on democracy and leadership (1) by observing that “the economic problem will be found to run into the political problem, the political problem in turn into the philosophical problem, and the philosophical problem itself to be almost indissolubly bound up at last with the religious problem.”

There is a connection between diversity and creativity. Diversity entails not merely competition between economic systems, power centers, and the political division of power, but also the existence of distinct spheres, each with its own standards for assessing an individual’s status and worth that are not subordinate to economic laws. Representatives of these spheres can defend their domains and exert influence; they have secured status and the prerogative of interpretation regarding matters concerning their respective fields. I am referring here to practitioners of the arts, the sciences, sports, and—not least—religion, as they existed before everything was subordinated to—and at times corrupted by—money acting as a standard of value and a mechanism of governance and regulation.

In direct contrast to Irving Babbitt’s conclusions regarding what matters in life, it is increasingly asserted in various contexts that the economy is the decisive factor. The economy is an appealing yardstick because the unit of measurement is relatively simple to quantify and comprehend; counting money requires no higher-order cognitive functions.

How and when can a state of diversity—in which the constituent parts possess integrity and self-determination—be reconciled with a sense of unity, where cooperation and mutual understanding arise naturally?

The social environment encountered on trains and at railway stations—across countries with distinct cultural characters during an InterRail journey through Central Europe in the hot summer of 2026 -- prompted my reflections on the complementary concepts of differentiation, globalization, and the focus on the individual. It becomes evident that, on a psychological level, Europe lacks a unifying pan-European project. While Europe’s geography has fostered distinct social spheres, unique cultural traits, and a creative diversity, the absence of a unifying fabric—such as a shared project—means we tend today to become individual projects pulling in different directions. We lack a common force and direction. Plato had already identified this problem, and his description of it concludes the dialogue *The Statesman*.

”Stranger: We may thus say that the supreme means for the weaving of statecraft—a proper interweaving of the characters of courageous and temperate people—is achieved when the royal art unites their lives in concord and friendship, thereby completing the most magnificent and excellent of all fabrics; and with this fabric, it envelops and holds together everyone in the state—free and slave alike—thus governing and guiding the state without neglecting anything that might serve its happiness.

Socrates: You have now presented to us, in an excellent manner, the royal man—the statesman.”

Today, one can clearly observe the throngs of isolated individuals in cities and at railway stations, most with their eyes glued to their mobile screens. A sense of social cohesion is hard to discern; instead, everything dissolves into a formless, aimless hum. It becomes evident that a well-articulated, unifying Pan-European project would provide much-needed guidance on the direction and energy with which we should all move forward—not as an undifferentiated mass, but as creative individuals who believe in a shared future. This is not a future governed solely by economic considerations, but one where individuals can be measured against various standards—sometimes competing and challenging, yet at other times overlapping and complementary. Such a project would facilitate integration and development by offering alternatives to political decisions and societal initiatives that are currently often predicated on the notion that everything ultimately comes down to money. It would also clarify the expectations and requirements placed on those wishing to join the European community, regardless of their reasons for doing so. Furthermore, it would foster a diversity that stimulates creativity across various spheres of human life and sectors of activity. Political challenges and decision-making would become more manageable if there were a unifying vision of the path ahead. The challenge lies in finding, articulating, and implementing a Pan-European project that leaves room for diverse individual, religious, cultural, and national variations. I believe the Middle Ages offer an example of a time when this was, to some extent, realized. Similarly, the United States in its heyday served as another model demonstrating the creativity, energy, and constructive dynamics inherent in such an approach. A constitution that guaranteed the separation of powers and states with room for self-determination, alongside a success story— “the American Dream”—that served as a unifying force for people of diverse ethnicities, classes, religions, and social backgrounds. This resulted in heightened creativity across various spheres—culture, music, art, literature, film, sports, technology, and science—with these fields fostering mutual cross-pollination.

Noahpinion

The American age was the human age

“What other country would have done this?” — Daniel Inouye…

Read more

2 months ago · 579 likes · 112 comments · Noah Smith

Medieval Europe was also characterized by a multitude of power centers that allowed creativity to flourish within the aforementioned spheres; simultaneously, the legacy of antiquity—symbolized by Socrates, Plato, and Aristotle—combined with Christian doctrine to form the intellectual glue and emotional foundation. These cornerstones established order, structure, and a shared conceptual framework that facilitated dialogue. The Catholic Church and the Latin language served as a common fabric, easing communication across the continent.

These examples can serve as a model when we observe the world fracturing and structures dissolving. Understandably, both leaders and citizens look back to a bygone era—a golden age—but instead of revitalized energy, the result is a stale, insular atmosphere that leads to stagnation and various political aberrations.

When the establishment—the so-called wise and sensible, whether young or old, yet anxious, narrow-minded, and complacent—views necessary changes, such as those in Europe, with skepticism, we see clear signs of a stagnating civilization.

2026-07-08 Olle Hollertz

1. Democracy and Leadership; Irving Babbitt, 1924

Subscribe now

During the first half of the 20th century, a momentous event took place. For millennia, horses had been a common means of transportation for human beings. Then, in about 50 years, they disappeared, replaced by machinery powered by fossil fuels. It was a true Seneca Collapse for the horse population. The data below are for the United States, but the same decline took place all over the world.

If you think about that, it is surprising that the replacement was so fast and so complete. After all, a horse has many advantages over a car or a truck. It doesn’t need an expensive infrastructure: oil wells, refineries, metallurgical plants, service stations, paved roads, and more. To produce a horse, you only need a male and a female horse. To feed it, you can use locally produced food that humans won’t eat. And a horse doesn’t need a paved road, nor expensive fuel transported from remote locations.

In a fight of horse vs. car, one could think that the horse would win hooves down because a horse is so much simpler and less expensive. Yet, the replacement occurred, and remarkably fast. There were more than one reason. Of course, a car can go faster and carry larger weights than a horse. But the main reason was control.

A horse needs to be fed and cared for even if you don’t use it. And horses still have a certain personality of their own, despite having been domesticated over thousands of years. They won’t always obey orders, and sometimes they may go wild. Instead, a car can be switched off when it is not needed, and when it is in use, it will do exactly what a car is supposed to do.

Better control was the main cause of the switch from horses to cars, but the question of costs remained. The switch was possible only during a historical phase when the EROI (energy return on energy invested) of fossil fuels was large. The result was that societies had cheap energy in excess that could be used to create the infrastructure needed for cars. Once that was done, horses had no chance.

But there is a problem: nowadays, the EROI of fossil fuels is rapidly plummeting. In other words, we are becoming poorer than before. And if we can’t afford cars any longer, won’t horses become a viable transportation technology again? So far, it is a trend seen only in certain regions: Cuba and parts of Africa, where the higher cost of fossil fuels (and the embargo in Cuba) is starting to bite. Below, you see something that happened in Cuba during the “special period.” (Image from Wikipedia).

Could something similar also happen in the rich West today? We cannot exclude it. If declining EROI makes fuels too expensive, then horses may become an economically attractive transportation tool again. Conversely, if high-EROI renewable technologies will be widely deployed, we may switch from fossil-powered engines to electric-powered motors and keep going with road vehicles. Technological transitions are not necessarily directed toward more complicated and theoretically superior ones. Whether the future of transportation will be Teslas or horses (or human feet) is all to be seen.

Now, let’s go to another technological transition: the one from human fighters to robot fighters (drones, if you like). I examined the trend in a previous post, from which I am taking this figure.

Lethal drones are being built at such a rate that their number could surpass the total number of human beings in less than 10 years from now.

To be sure, there would still be some reason for human beings to exist as fighters. A 70 kg human being carries maybe 8–9 kg of body fat. Converted into energy, it makes ~330 MJ at 37 kJ/g. For a marching energy of ~12–16 MJ/day, the human could go on for a few days without the need for “refueling.” Just like horses, humans are energy efficient and metabolically cheap. Humans can also refuel on a variety of sources, from roasted rats to berries and nuts.

Bots are less efficient and more finicky in terms of refueling. It is not so much a question of energy expenditure, which, for a human-sized quadruped robot, is now in the same range as that of a human: around 20 MJ, or 5-6 kWh per day. The problem is the weight: A lithium battery today weighs between 6 and 8 kilograms per kWh. So, to match the endurance of a human being, the battery of a human-sized legged robot would weigh around 30-50 kg, about as much as the whole bot. An even more serious problem is that bots cannot forage for electrons. They depend on a long and complicated supply chain to provide the energy they need. To say nothing of the need for a supply of expensive minerals and specialized factories.

Yet, just as for horses, being cheap is not a crucial advantage in comparison to efficiency in terms of control. Just like cars are easier to control than horses, warbots are much easier to control than human fighters.

Obeying orders is a feature that army commanders have been trying to instill in their soldiers across the whole history of warfare. Soldiers are drilled to behave like robots. Sometimes they are drugged; in other cases they are more scared of their officers than of the enemy. It may work, but not always. Soldiers may turn around and run, or shoot their officers, or, sometimes, just quit and go home.

Sometimes, it works too well in the opposite direction, and soldiers go on a rampage killing people and destroying things. Think of one case: the My Lai village in Vietnam in 1968. US troops there shot and killed some 370 or 500 (depending on the estimates) unarmed civilians, most of them women and children. It doesn’t seem that the soldiers ever received an explicit order to do what they did. Likely, they went out of control in a frenzy of “blood lust.” It is also called “feeding frenzy” with sharks, a behavior that’s remarkably similar to that of humans in certain situations.

The My Lai case is just one of the many cases of horrific behavior of human soldiers in history. Additionally, soldiers don’t just kill innocent people; they tend to rape them. It happened in My Lai, but — again — it is a common behavior in history. Just as an example, read Curzio Malaparte’s book, Kaputt (1944). The chapter about the brothels for the German troops in Romania during WWII is horrifying.

Robots, on the contrary, only obey orders they actually received. They won’t run away, they won’t shoot their officers, and they do not suffer from blood lust. They will not kill civilians except by mistake, or if explicitly ordered to do so. And it is hard to think that they could be ordered to rape women. (But never underestimate human imagination).

It is also possible to place safeguards in the bots’ functioning code that would prevent illegal orders from being obeyed. This is an ongoing debate that flared, among other cases, when the US Department of War demanded unrestricted use of Anthropic’s AI models, and Anthropic refused, citing safety guardrails against fully autonomous lethal weapons and domestic mass surveillance.

We don’t know if these robotic guardrails could be more effective than the rules of morality that should drive human behavior, but they might well be. Robots might turn out to be more moral than human beings, as argued by Isaac Asimov in his The Evitable Conflict (1950)

So, as long as we have the energy needed to build and operate robots, there is nothing to do: human fighters are just as obsolete as horses. And, as for horses, there is no question for the old technology to “stay in the loop,” or, as Mostaque proposes, to “go endosymbiont.” We are seeing that very clearly both in Ukraine and in Iran.

But is it a long-term situation? Suppose that the global economy starts to go down because of mineral depletion and negative pollution effects — the scenario that “The Limits to Growth” study projected for the mid-21st century. Then, robotic fighters may become unsustainable: too energy-hungry, expensive, and in need of long supply chains. So, humans will be back to killing each other with swords and maces, slaughtering harmless people and raping women in their free time.

There is also the opposite possibility: that the high EROI of renewables will generate abundant and cheap energy. As a consequence, large numbers of warbots will be manufactured. In that case, the replacement of human soldiers with bots could go as fast as that of horses with cars. And, who knows? It wouldn’t be a bad idea if drones were to decide to fight each other, leaving us, humans, in peace.

Subscribe now

This post will tell you:

1. Why blocking people on social media won’t improve the quality of the conversation

2. How social media may collapse into a constellation of isolated subgroups.

3. The role of bots in the discussion on social media.

4. Why pasting links into social media posts is useless.

5. Why the current social media platforms may be destined to collapse.

6. The political consequences of all that.

Blocking Offenders

On social media, facing really outrageous comments, insults, and smears, the typical temptation is to block the offender. Maybe you think you are somehow punishing him or her (or it, if it is a bot). And you may think that if everyone were to ban those obnoxious creatures, they would find themselves isolated, and they would disappear.

Alas, it is not like that. Not that blocking people is not justified in some cases, say, when you need to eliminate people (or bots) who are clearly trying to derail a discussion. But, in general, it is not a good idea.

The effect of link breaking (blocking) in large networks has been studied by models. The most directly relevant paper is Coscia and Rossi (2022), How minimizing conflicts could lead to polarization on social media. The result is that the tendency of users to block connections to avoid backlash and conflict increases polarization rather than decreasing it.

Not so surprising, perhaps, but remarkable nevertheless. Conflict avoidance is the engine of polarization.

I played with similar models, and I can confirm this result. Networks tend to become polarized, fracturing into clusters of like-minded people. Assuming that the threshold of blocking goes down as the cluster homogenizes, people would move to block others for a lower degree of disagreement.

When this trend starts, there is no mechanism that can stop it. First, the network degrades into a constellation of isolated or nearly isolated clusters. Inside each surviving cluster, the internal spectrum of opinion becomes the new boundary. Whatever residual disagreement exists is now the most extreme thing anyone sees, and the same blocking rule re-applies, recursively. The result is called “Sunstein’s group-polarization.”

Each cluster contracts and radicalizes until it is small, homogeneous, and tuned to a narrow set of opinions. The non-conforming users are expelled to form their own clusters. Clusters might shrink to the point that everyone becomes an isolated monad in a network that’s not one anymore.

Would this mechanism generate a Seneca cliff?

In this case, the Seneca Curve describes the degree of connectivity of the network. As it grows, connectivity slowly increases, but if connections are broken faster than they are created, the result is a fast decline.

The results of the model do not produce a Seneca Cliff as the result of pure polarization. Instead, the number of connections tends to decline slowly. But a virtual network can collapseIt can happen. Before Facebook existed, there was an earlier social network called Friendster. On the Y scale, the number of users in millions. It was not a small network, although it never reached the billions of users that Facebook has today.

This curve is also known as “Zuckerberg’s Nightmare.”

Friendster’s example does not describe the effect of blocking; but there are many different ways to cause a network to collapse. Censorship can generate it, and you may have noticed how Facebook has considerably restrained its “fact checkers,” preferring more subtle forms of censorship. They know that they are at risk of collapse if they exaggerate.

A more realistic version of the “unfriend everyone you disagree with” would not involve total collapse. A minority of users would refuse to play. They function as bridges, the weak ties Granovetter described. A small number of unblockers can keep the network alive. So the realistic endgame is not total fragmentation but a network reduced to a thin scaffold of bridge-people connecting otherwise sealed cells. Which, arguably, is not far from where some platforms already are.

Blocking Bots

We saw that blocking contacts on social networks is not a good idea. But what about bots? Aren’t they the bane of these networks? Wouldn’t it be good and proper to eliminate them? Unfortunately, bots are difficult beasts to deal with.

A great deal of social-media discourse is not produced by humans. Meta’s official prevalence figure — stable at 4–5% for years — describes the fraction of accounts on Facebook that are fake. Add duplicates, and you get to about 16%. The quarterly removal numbers are more striking still: in Q4 2025, Facebook “actioned” 1.1 billion fake accounts (the corporate euphemism for “deleted”)

But the account fraction is the wrong number. The right number is the fraction of the comments and posts you actually read that were generated by bots. A dormant fake account counts the same as one running 200 posts an hour. There is no public estimate of the content-weighted bot fraction, because Meta has no incentive to produce one. Rough estimates say that the bot fraction of visible comments is somewhere between 30% and 80%, and rising sharply since cheap LLMs arrived in 2023.

Would universal blocking trend eliminate bots? Partially, yes. But with serious problems. Some bots can be defined as “disagreement-injector bots” (troll bots). They use insults and extreme statements to generate a reaction from users. Most of them would be culled in the first round. But there exists a much larger population of amplifier bots — accounts that like, share, retweet, and post affirming comments: they don’t normally trigger the disagreement criterion. They are advantaged by the new rule. As humans cull their dissenters, the proportion of in-group signals coming from bots rises. These bots may become dominant in a cluster.

At the same time, the troll bots don’t disappear. They cannot be “killed,” so they reappear in the contracting clusters, amplifying the cluster’s own narrative against an absent enemy. No member of the opposing cluster is around to correct the strawman. Bots become more effective at radicalization precisely because the reality check has been blocked.

Selling Chanel No. 5 in a Fish Market.

The fleeting nature of social networks has consequences that we are only now starting to understand. When I publish a post on this Substack, I usually get between 2,000 and 3,000 visits. Substack lets me see where those visits came from. Email and direct visits together — both forms of pre-Web, pre-algorithmic engagement — account for 85% of my readership. The advertising-attention complex we are told dominates contemporary discourse — Facebook, X, LinkedIn combined — sends fewer readers than DuckDuckGo alone. DuckDuckGo is a search engine used by perhaps three percent of the Web population. It delivers more traffic to my blog than the world’s largest social network.

The small cluster of independent sites linking to mine — Cassandra’s Legacy (my old blog), normalamerican.com, olduvai.ca, rayonegro.substack.com, stevebull.substack.com — collectively send more visitors than X and LinkedIn combined. Small site links to small site; the reader follows the trail. Individually modest, collectively significant, and high in intent density — the visitors who arrive this way are already in the conceptual neighborhood.

I had vaguely assumed I might be an outlier, that other Substack writers were getting most of their traffic from social. Talking to a few of them, this turns out not to be true. The pattern is general. The social platforms are not the distribution layer of the contemporary web. They are not even meaningfully part of the distribution layer. They are advertising substrates with high internal activity and near-zero export of attention to neighboring ecosystems.

Which means that if you want people to click on your links, it is useless to paste them on social media platforms. If you want to sell your book, the best way is to advertise it on a site frequented by potential buyers; Amazon, for instance. Trying to sell a book on social media is like trying to sell Chanel No. 5 in a fish market.

The Social Media Model

It is tempting to call fragmentation a“flaw” in social media, which suggests something that could be patched. But it is the architecture that’s structurally generating this behavior.

Look at how the feedback loops of social media function. Engagement loops back to ad revenue, which loops back to algorithmic amplification of whatever produces more engagement. That circuit is tight, fast, and well-tuned. Nothing in it has a terminal in understanding, memory, or resolution of disagreement. The platform is not failing to deliver thoughtful discourse — it was never wired to do so. Asking Facebook to produce thoughtful discourse is like asking a refinery to produce a forest.

It is not a design error. It is the consequence of understanding the Web as a social network similar to a town square where people meet, exchange a few words, and then move onward. In that kind of situation, you don’t expect to engage in deep discussions; you say hello, howdy, seeya, and the like. The harmless chatting that has accompanied humans from their remote tribal origins.

But in modern times, humankind needed to have deep discussions on complex subjects. Up to recent times, academies, journals, courts, parliaments, and even letter-writing networks were slow on purpose. The slowness was the regulatory mechanism. Time between exchange and response allowed reflection; gatekeeping allowed quality filtering; permanence imposed accountability; bounded audience prevented runaway amplification.

The web removed all four simultaneously, declared it liberation, and was surprised when the structure collapsed inward. This is a textbook case of efficiency optimization that destroys resilience. The same pattern is visible in monocultures, just-in-time supply chains, and high-frequency trading.

There is also a basic category confusion. Humans don’t maintain a single social graph. They maintain several: a small intimacy graph (Dunbar’s ~150), a professional graph, a civic graph, and a stranger-courtesy graph. Each runs on a different protocol — different rules about candor, persistence, evidence, and sanction. The web collapsed them into one channel with one protocol, so a remark fit for the dinner table now arrives in the same feed as a policy debate, and both are subject to the same algorithmic amplification, the same screenshot permanence, the same global audience. No human society in history operated this way, and there is no reason to expect humans to behave well under conditions their evolution never prepared them for.

Political Consequences

The extreme polarization that we can observe on social media nowadays has political consequences. One is that it is difficult to use the web to grow a consistent political movement. The clustered structure of the network creates a situation similar to that of trench warfare during WWI. Armies face each other and engage in bloody battles, but neither can gain the upper hand. It is the same on the Web. Different clusters engage in virtually bloody battles, but no side manages to convince the members of the other side that they are wrong.

In a sense, it is a good thing. Adolf Hitler, today, would be an internet troll spewing out hundreds of venomous messages against Jews and other Untermenschen. But he would be read only by a handful of followers. In its time, the Nazi party expanded to become a totalitarian force because it could control the top-down traditional media, while at the same time using physical force to intimidate and eliminate adversaries. Today, that would be much more difficult because of the fragmentation and the boundaries that separate believers and unbelievers.

Unfortunately, the same polarization and atomization of the Web may make it easier to create and maintain a totalitarian government, just because it makes it impossible to create a significant political movement that opposes it.

That is, more or less what we are seeing on the social networks right now. Polarization, atomization, useless verbal battles, no significant progress in anything. At the same time, governments are working unopposed to gain more and more power.

The Collapse of Social Media

The part of the internet that produces and circulates ideas, that refines arguments through exchange, that builds a durable readership, has already migrated off the social platforms. The social networks as knowledge builders are disappearing.

The three billion accounts on Facebook are still there, real and synthetic mixed together. The total content-generation rate is still climbing. But the people who actually want to go somewhere — to read carefully, to write seriously, to participate in the slow work of figuring something out — have quietly left, or connect only occasionally.

The Web that has a reason to exist looks like Web 2.0. Individual sites cross-linking to each other. Email subscriptions. Search as discovery. A sparse mesh of trust between writers. Substack is the most evident current example, but the same pattern lives in academic blogrolls, in specialized newsletters, in podcast cross-promotion, in the better Discord servers, in Mastodon instances, and the fediverse. Each is groping toward a topology that the social platforms cannot reproduce because their business model forbids it.

If everyone blocked everyone who disagreed with them tomorrow, the social networks would collapse — but the collapse would mostly affect a system that has already lost its discourse function. The thing that would die is the advertising substrate, not the conversation. The conversation has already moved.

This is, I think, the Seneca Effect at its most precise. The growth side is still visible: user counts, ad revenue, attention metrics, all still inching upward. The collapse side has already happened underneath, invisibly, in the migration of the function the system was nominally supposed to perform.

Conclusion

We started with a simple question: whether blocking obnoxious contacts on social media is a good idea. The answer involved an in-depth analysis of how social media platforms work, and the fact that there is no way to “fix” them by simple actions.

It looks like we’ll have to live with the social media platforms as they are, at least as long as they exist. Would things change in the future? If the current social media platforms collapse, as they might, what would replace them? Will more structured exchange platforms develop?

Things keep changing; just think that 20 years ago, Facebook as we know it didn’t exist. And three years ago, AIs as we know them didn’t exist. As usual, we move into the future without knowing where exactly we are going.

This post benefited from assistance from Claude Opus 3.8 for the data and for a critical revision.

Merlin the Magician could wave away the physical laws with his magic wand. The real world, though, is another matter, and its laws cannot be ignored. Not even by Artificial Intelligence.

Subscribe now

Around 1492, the Benedictine abbot Johannes Trithemius wrote a treatise, In Praise of Scribes, defending the holy labor of copying manuscripts by hand against the new and vulgar printing press. Hand-copied books, he argued, were morally and spiritually superior.

Emad Mostaque opens his book, The Last Economy (2025), telling the story of Trithemius as a warning to the reader: you are the scribe, AI is the press, and unlike Trithemius, you do not have fifty years to adjust.

It is a good story. But note that Trithemius reached for Gutenberg’s diabolical machine the moment he wanted to be read. The same man who criticized the press let it print him anyway. He was not a fool, but his understanding of what was going on was only partial. Mustaque’s book suffers from the same problem: interesting, fascinating, and worth discussing, but with deep flaws.

What Mustaque gets right

Mostaque writes about networks as someone who knows the matter well. His spine is what he calls the Intelligence Inversion — a sequence of economic phase-shifts from land to labor to capital to, now, intelligence itself — and his central claim is that we have a narrow window, a “thousand-day window,” before the new order locks in.

The window crystallizes, he says, into one of three “futures.” Digital Feudalism: a handful of corporate duchies own the foundational models, and the rest of us live as users on a universal basic income calibrated to keep us fed and subscribed — a cage made comfortable enough that we forget it is a cage. The Great Fragmentation: nations panic, the internet splinters into national firewalls, and the world settles into a zero-sum cold war fought with algorithms. Human Symbiosis: the difficult, virtuous path, in which intelligence becomes a commons rather than a commodity and amplifies human purpose instead of replacing it.

When Mostaque is describing how these states emerge from network effects — why computation and capital collapse toward monopoly, why fragmentation is self-reinforcing — he is truly sharp. He has understood network theory far better than he understands biology or biophysics. The taxonomy is clean and useful. It is a good map of the political and economic landscape as it is today in the world.

It is the physics underneath the map that does not hold.

The magic wand

Here is the flaw that, in my modest opinion, is fatal. Mostaque dismisses scarcity with a movement of the magic AI wand. The premise that runs under every chapter is that intelligence dissolves material limits — that abundance is, in his phrase, a feature rather than a bug. There is no problem of energy supply, none of copper, lithium, or water for cooling. For a book whose subject is one of the most physically voracious enterprises our species has ever attempted, the omission is total.

We have been here before. Mostaque has rediscovered Solow’s residual — the unexplained leftover in the growth equation, the part the economists could not attribute to labor or capital and so attributed to “technical progress.” Mustaque has renamed it AI. It is one of the miracles in which we stubbornly keep believing, alongside Santa Claus, the free lunch, and democracy.

The Club of Rome has spent more than fifty years trying to persuade the world of the opposite: that scarcity is real, that it is tied to the mineral base of the planet, and that no quantity of cleverness repeals the arithmetic of a finite system. Smart or not, an AI needs resources, just as we do. Growth cannot go on forever, and intelligence cannot replace thermodynamics — except in Walt Disney’s movies and politicians’ speeches.

The crystal and the cliff

More in detail, there is an entire section of the book that just doesn’t work. Mostaque reaches, again and again, for the vocabulary of dynamical systems. The old order will not evolve smoothly, he writes; like a supercooled liquid, it will “snap” into one of a finite number of stable crystal structures. The three futures are “attractor states,” “basins of attraction,” the only configurations physics permits. It sounds rigorous.

By dressing the transition in the language of phase changes, Mostaque smuggles in the conclusion that the transition has a soft landing built into its physics — that we will settle into one of three tidy states and stay there.

This is NOT the way complex systems work.

Complex systems under stress do not, as a rule, crystallize politely. They overshoot, and then they come down — and they come down faster than they went up. That asymmetry is the whole content of what I have called the Seneca Effect, after the ancient Roman Philosopher who noticed that fortune is slow to build and swift to ruin.

Mostaque comes within sight of the Seneca Effect more than once; he can feel that something nonlinear is coming. But he cannot quite catch it, because catching it would wreck the optimism. A book that took the cliff seriously could not promise three stable futures.

The problem is that the book has no model behind it. No equations, no stocks and flows, no integration of the curve he keeps gesturing at. Terms such as “Attractor,” “nucleation,” and “basin of attraction” are not doing analytical work in this book. They are wallpaper — borrowed authority pasted over an argument that is, at bottom, a hope.

Which one of us is the mitochondrion?

The third future, Symbiosis, is the one Mostaque wants us to choose. It rests entirely on a single biological metaphor: the endosymbiosis described by Lynn Margulis, in which a free-living bacterium became the mitochondrion of the eukaryotic cell. Cooperation, not competition, drove the great evolutionary leap.

Margulis was one of the great thinkers of the 20th century, and her ideas are a pillar of modern evolutionary biology. Humans and AIs, Mostaque proposes, will partner in the same way as mitochondria and larger cells.

It is the crucial part of the book, and the weakest, because it is assumed rather than proven. And the metaphor cuts the wrong way for him. Endosymbiosis was not a marriage of equals. The mitochondrion is a captive: its genome gutted and outsourced to the host nucleus, its existence permitted because it is useful, its independence gone forever. The analogy is more like the enslaved human beings as energy sources for the robots in the Matrix film series. So the honest question — the one the book never asks — is which one of us is the mitochondrion?

Speaking as a free-living organism, I can tell you that if I were a bacterium, I would think twice before agreeing to spend eternity manufacturing ATP for a cell whose DNA is nothing like mine, on the promise that it is a partnership.

So, what, exactly, does the human bring to this symbiosis that is worth the enslavement? The book does not say. It simply trusts that we will be the nucleus and not the organelle.

Intelligence against entropy

I will end with one sentence, because it is the whole problem in miniature. In a chapter titled, with no irony, “Intelligence Against Entropy,” Mostaque writes:

“The energy cost per logical operation in a modern chip is trillions of times lower than in a human brain.”

It sounded wrong to me, so I checked. It is wrong, and not by a little. The human brain runs on roughly 20 watts and spends about ten femtojoules per synaptic event — a figure so good that neuromorphic engineers treat it as a target they have not yet reached. Surveys of real AI hardware find that most accelerators are less efficient than the brain, and the very best ones only just match it. So the comparison points the wrong way to begin with.

But “trillions of times lower” is worse than wrong; it is impossible. Trillions below ten femtojoules per operation would put the chip at around ten-to-the-minus-twenty-six joules — roughly five orders of magnitude beneath the Landauer limit, the thermodynamic floor for erasing a single bit at room temperature. His efficient chip would be violating the second law. And this in a chapter about entropy!

The use of a flawed book

I have been hard on the book, so let me say what I think it is good for. Mostaque is not a fool; he is an intelligent insider who has clearly thought hard, and he still cannot hold the whole picture — he gets the networks and loses the joules, gets the urgency and loses the limits.

That is not a failing peculiar to him. The change underway is so large that none of us holds the whole of it, and I include myself: my own attempt to read the rise of AI through evolutionary theory and biophysics is grounded in what we actually know, but it is partial too.

The reason the conversation about our intelligent future is full of people getting thermodynamics wrong is that almost no one in the field of AI knows much about thermodynamics. Which is expected. We, humans, are limited — no single human brain can know everything. And in this very moment, our limitations show. Our best minds can’t catch the complexity of the problems we are facing — it is a task too difficult for us.

Will AIs be able to do better? Perhaps. I don’t want to make the same mistake that Mustaque did, confusing hopes with reality. But, on the other hand, hope is not a bad thing to have!

h/t Anders Wijkman

Subscribe now

We owe to Oscar Wilde the definition of the cigarette as the perfect type of a perfect pleasure. It is exquisite, and it leaves one unsatisfied. What more can one want?

Wilde was reasoning in terms of personal pleasure, but his statement also describes a perfect product in economic terms. A cigarette cannot be treasured, reused, or recycled. Using it means destroying it, and the only way for the consumer to keep smoking is to buy another one. That creates a new profit for the producer.

Of course, smoking will eventually kill the consumer, but that’s not a concern for producers as long as there are smokers alive. Perfect market mechanism!

Now, let’s try to express the same concept in military terms. A drone is the perfect product of the military economy. A suicide drone destroys itself when killing a person, so it needs to be replaced, just like a cigarette destroys itself when smoked. Some drones can be used multiple times, but what makes them really profitable is their low cost. And single use is a typical characteristic of low-cost products.

So, the economic system always moves in the direction that provides the largest profit. The old consumer economy is being replaced by a military economy. The only difference is that a military economy produces goods that are not consumed by consumers, but by the targeted enemies. In both cases, the mechanism that generates production is simple. It is profit.

Profit-making processes generate typical feedback-enhanced growth. Successful start-up companies often grow by this mechanism, exponentially, until they saturate their market. The result may be a “logistic curve:” Sales plateau and remain there. Or, if the market changes with time, the result may be the collapse of the producers. It works no matter what the product is. Cigarettes or lethal drones, it is the same.

That’s what’s happening nowadays. Drones are a hot topic in the economy, and the industry is gearing up to produce more and more of them. We are now at the level of some 10 million lethal drones manufactured per year. That’s already enough to consider them as a potential mass extermination weapon. But production keeps ramping up: when is it going to stop growing? Will it be when there will be no more targets to kill; that is, no more human beings? It is an unsettling question, but our future depends on the answer.

The Beginning: Aerial Bombing to End All Wars

Drones are an evolution of the concept of aerial bombing that started becoming part of military thinking in the early 20th century. Most of the early debate on this subject is forgotten today, but at least one early strategic study is still remembered: Giulio Douhet’s 1921 book “Il Dominio dell’Aria” (the Domain of the Air).

Douhet’s thesis was simple: aerial bombing had made all other weapon systems obsolete. No need anymore for armies: just methodically bomb a country, destroying one city after the other. Eventually, either they will surrender, or there will be no one alive left. (See an earlier post of mine).

Douhet’s ideas have been roundly criticized on the ground that his predictions turned out to be wrong. That is, it never happened, so far, that a country surrendered only because it was the target of an extermination campaign directed at civilians. But could it be that Douhet seems to be wrong just because he saw much farther in the future than most of his critics were willing to consider?

There is a curious parallelism between two ancient prophets of doom: Thomas Malthus and Giulio Douhet. Both are still widely known and discussed, but only to declare that they were wrong. Nevertheless, as Herman Daly noted about Malthus, “anyone who has been buried so often cannot be entirely dead.” A sentence that may well apply to Douhet, too.

The recent development of drone warfare may be a vindication of Douhet’s ideas. We are seeing a rapid increase in the number and sophistication of drones, which are now the main weapon used by all combatants. The basic idea is that drones are “intelligent ammunition.” Unlike an artillery shell, a drone has a certain capability to locate a target, direct itself toward it, and take evasive action to avoid being stopped.

Perhaps the most effective kind of drone is the simple “suicide” quadricopter that can kill one person and be manufactured at a low cost in very large numbers. Mass extermination would be the result if someone were to unleash millions of these inexpensive drones against inhabited places, so as to put Douhet’s ideas into practice.

Of course, the fact that something is possible does not mean it will happen. Here, we have a deeper question to answer: wars, it is well known, are expensive and useless. But they are possible, and not just that: they are common. And, as I said, they are processes easily explainable in economic terms. Simply stated, all wars generate profits, at least for someone. (You may read something about the frequency and distribution of wars in my 2024 book “Exterminations.”)

The future is obscure, as always, but this approach gives us some idea of what we could be facing. And it is not a comfortable view.

Death from the Sky

The first large-scale extermination campaign by aerial bombing was carried out against Germany from 1940 to 1945, during WWII. It was a novelty in history that prefigured much worse to come. Let’s analyze it in quantitative terms. Here is the bombing tonnage dropped on Germany.

Note how the amount of bombs dropped on Germany grew exponentially from 1940 to 1944. It had little or no effect on the German armament production because it was aimed mainly at killing German civilians, as Douhet would have advised. Only in late 1944 did a certain effort start being made to destroy German infrastructure, rather than killing unarmed civilians.

And here is the number of civilian victims:

These data imply a doubling rate of the number of victims of about 7 months. Extrapolating into the future, it would have taken just a few years to exterminate all Germans. Fortunately for them, the Soviet army took Berlin in May 1945 and forced Germany to surrender. That put an end to the extermination campaign.

Now, let’s compare with the bombing campaign on Japan, from 1944 to 1945.

Note how the victims curve is dominated by two events. One is the firestorm on Tokyo in March ( see the movie “The Grave of the Fireflies, Hotaru no haka, by Isao Takahata, 1988, or, better, do not watch it if you have a weak heart). The second is the nuking of Hiroshima and Nagasaki in August. Nevertheless, the bombing campaign intensity followed a near exponential growth rate just as the campaign against Germany.

In terms of death toll, the German and the Japanese campaigns are of the same order of magnitude — each somewhere around a third to half a million civilian dead each. But the campaign against Japan was shorter and more brutal. Germany’s toll came from roughly 1.3–2.7 million tons of bombs over five years, Japan’s from only ~160,000 tons in about six months.

Here is a comparison of the bombing campaigns’ effectiveness in Germany and in Japan. It shows how the kill ratio increased over time: from 0.8 deaths/ton in Germany, to 2.3 deaths/ton for conventional bombing in Japan. Then, it reached more than thirty thousand victims per ton with nuclear bombs.

Extermination campaigns directed at civilians didn’t stop with Japan in 1945 but went on and even became worse. Here are some data summarizing the historical trends of the past 70 years or so.

Remarkably, during the past 80 years or so, humankind went through a series of waves of homicidal madness that saw aerial bombing being used to kill significant fractions of the population of some regions. The most recent case is that of Gaza. The worst case may have been that of North Korea, where US bombing may have killed about 20% of the population. A datum confirmed by the commander of the operation, General Curtis LeMay, also responsible for the Japan campaign.

Laos may have been even worse: Between 1964 and 1973, the US dropped more than 2 million tons of bombs on Laos, more than all the bombs dropped by the US during WWII. This made Laos the most heavily bombed country in history relative to the size of its population.

The Bombing Campaign that Never Was: The Nukes

Nuclear weapons make an interesting parallel with that of conventional bombing, but with some fundamental differences. After the first two nukes dropped on Japan in 1945, no more were used in actual war. In a certain sense, it is a surprising development: why not use a weapon that turned out to be some 10,000 times more effective than conventional bombing in terms of kills/tonnage? Maybe our leaders were wise enough to understand that nuking each other was not a good idea?

I think there is another explanation. It is that nuclear weapons are so effective that they bring no profit to manufacturers when they are used. If drones are like cigarettes in economic terms, nukes are more like cars. Once a customer has bought a new car, they don’t need to buy another one, at least not soon. Once a nuke is used to destroy a city, there is no need for another nuke to destroy it again. That’s not a good thing for the manufacturers.

The car industry tried to avoid saturating the market by strategies such as the “model of the year” and “planned obsolescence.” In part, they worked. Something similar was done for nukes: many stockpiled warheads were dismantled and rebuilt; turned into more efficient killing versions. Yet, there was no profit in using them.

The result was that with nuclear warheads, profits could be maximized by stockpiling nukes instead of using them.

Here are some data: the number of nuclear weapons stockpiled in the US and USSR:

And here is an estimate of the number of nuclear warheads produced every year.

In terms of explosive equivalent, the maximum stockpiled amount in the US, in round numbers, amounted to about 20,000 megatons, or 20 gigatons of TNT equivalent. (more than one million Hiroshima-sized bombs). Considering that the world’s population in the 1960s was a little more than 3 billion, it means that at some moment the US had the equivalent of about six tons of TNT per inhabitant on Earth. Adding the Soviet stockpile, we are at about 10 tons of TNT per person. A wonderfully insane number and an interesting concept in terms of “overkill” -- think that the most bombed region in history, Laos, “only” received about 2 tons of TNT per person.

But what stopped the production and the stockpiling of nuclear weapons? Of course, there never was any need for 6 tons of explosive per person, but, clearly, someone thought that less than 6 tons was not enough. So, why not 10 tons per person? Or a hundred tons? Or more?

As for many things in the world, there holds the basic law that if something cannot keep going forever, at some moment it will stop (it is called Stein’s law). That’s especially true when the growth is exponential.

See these data: (AEC stands for “Atomic Energy Commission,” which had little to do with energy, but a lot with nuclear weapons)

Note how the uranium production curve leads the warhead production one. To make bombs, the industry needed uranium, and that is a finite mineral resource. The “Hubbert” shape of the uranium purchases curve suggests that mineral depletion was at the basis of the story. The US didn’t run out of mineral uranium, but the industry was forced to move to less concentrated, and hence more expensive, resources. At some point, producing new bombs became so expensive that someone balked at committing the huge amounts of funds needed to keep the exponential curve going.

Things are more complicated than this, and the US uranium production had a second season of growth that peaked in 1980. This second peak was the result of a different economic balance. Enriched uranium found a new market in fueling nuclear power plants, and that produced a certain revenue that allowed the industry to spend more money on extraction to fight depletion. But that didn’t restart the production of nukes, which were an obsolete product in economic terms.

In any case, the US does not have significant uranium mineral resources any longer. A new run of stockpiling nuclear weapons looks at least unlikely, fortunately.

Homicidal Fury: a New Run is Starting.

Mass exterminations by aerial bombing followed the same pattern in history. A rapid growth that was interrupted only by the collapse of the economy of the bombed country.

Drones are a just introduced innovation. Will they follow the same pattern?

The curve for drone manufacturing is growing fast (note the logarithmic scale of the Y axis). The doubling time is approximately 10 months and, at this rate, there will be more drones than humans in 8 years from now. But we don’t need to have 8 billion active drones to exterminate humankind to the last person alive. Some drones can kill several people every time they are used. Even for “single kill” suicide drones, assuming a reasonably good efficiency, one billion suicide drones manufactured per year would be able to kill everyone in a few decades. These numbers are not far-fetched if compared with objects of similar cost and complexity; smartphones, for instance

In many cases, drones could be weapons of mass destruction even without directly killing people. In Middle Eastern countries, you only have to target the water desalination plants to kill off a very large fraction of the population. Everywhere in industrialized countries, just blocking some of the transportation nodes will cause mass starvation in a few weeks at most.

Of course, there are anti-drone technologies: cannons, lasers, electronic jamming, and more. The problem is cost, and hence these technologies can be deployed only against targets that are worth spending the money necessary for defense. Which means that the anti-drone resources will be focused on protecting high-value military targets, not ordinary people.

It is Douhet’s principle: methodically kill civilians until there is no one left. Why civilians? Because they are easier to kill than the military.

What can stop drones?

The situation is not very good for humankind. We face the possibility of losing control of a new weapon technology that could result in mass exterminations at levels never seen before in history. The case of Gaza is telling: the destruction was only marginally caused by drones (even though loitering quadricopters are reported to have killed thousands of people). Mostly, it was the work of conventional aerial bombing and artillery. But it is not a big difference: in all cases, it is death coming from the sky at little or no cost for the attacker.

Imagine Gaza expanded to the level of much larger countries and you have some idea of what could happen to your town. If someone decides that your ethnic or cultural group doesn’t deserve to exist, then a propaganda-fueled homicidal frenzy against you will be both effective and nearly impossible to stop. Gaza is just the latest example.

So, what could stop drones from exterminating humankind? There are a few possibilities to examine.

1. Treaties

If you are an optimist, you may think that humans could find an agreement to ban drones. Personally, I think it is very unlikely. Just think that the treaties on nuclear weapons came only after they had been produced at the level of tens of thousands, enough to kill everyone several times over. They came only when there was no more profit to be obtained by manufacturing more nukes. And think that a treaty against biological weapons was possible only when it became clear that these weapons were not as effective as they were thought to be. Treaties about drones may come when we’ll have so many that making more won’t produce so much profit. One may always hope, but I wouldn’t bet on this idea.

2. War ends before extermination is complete

This is the case of the conventional bombing campaigns of the 20th century. They stopped not because there were no more people to kill, but because a conventional army had defeated and taken over the bombed country. And that made further bombing unnecessary. Indeed, some of the most destructive bombing campaigns in history were in Laos and North Korea, where there was no ground invasion force that would “finish the job” and stop bombing. Unfortunately, drones have the additional effect of making ground invasions nearly impossible, so there is not much to hope in this sense

3. Surgical strikes make extermination unnecessary.

I mentioned the possibility that drones could be used to hit vital elements of a country’s life support system: desalination plants, food transportation systems, electricity, and the like. A country hit in this way might be forced to surrender much before suffering a substantial number of victims. It may be the case of Serbia in 1999, where bridges, roads, electrical grids, telephone networks, and oil refineries were destroyed, threatening a total nationwide blackout heading into winter. That forced the Serbian government to negotiate a surrender. On the other hand, in Gaza, the destruction continued until an estimated 90% of water and sanitation infrastructure was damaged or destroyed. Gaza survivors find drinking water in large part from rationed humanitarian aid. If that were to stop, the extermination seen up to now would be just the beginning of a much larger one. In any case, from the viewpoint of an ordinary person, it makes little difference if they die by a direct hit, starvation, or dehydration.

4. Cheap Anti-Drone Systems

There are many ways to stop drones before they kill you, but it is unlikely that you would be able to afford a high-power laser gun to mount in your backyard. Only governments can deploy these systems, but would they do that to protect civilians? That’s not only far from guaranteed, but governments may have the exact opposite attitude. Having the enemy get rid of useless mouths (you and your family) might be welcome, also because a drone used against you means one less drone used against them. Your best bet if you have no budget for super-weapons is to hide somewhere underground. But drones can follow you there, and how long can you play the game of whack-a-mole?

5. Cost.

Could drones become too expensive to manufacture, just like nuclear warheads became in the 1960s? Unlikely. A small suicide drone is in many ways similar to a smartphone. Consider that 1.4 billion smartphones are produced every year nowadays, and it took the smartphone industry approximately 16 years to go from the first one million units to the one billion mark. Today, there are 7.43 billion active smartphones, about one per person. If scale factors enter into play, prices might go down to less than $1000 per drone. It means that you could build a billion of them at a cost of one trillion dollars, which is less than the 1.5 trillion US yearly military budget envisaged by Donald Trump. So, cost is not an obstacle preventing the manufacturing of one suicide drone for each human being on Earth.

5. Lack of necessary mineral resources

That is a reasonable possibility, perhaps the most promising one in terms of human survival. We saw that mineral depletion was the probable cause of the stopping of the stockpiling of nuclear warheads by the main powers, the US and the USSR. Not just that, but also competition with the civilian nuclear industry caused the military nuke production to be starved of uranium. It was a good thing. Could something similar happen to drones? The answer is yes. It is perfectly possible.

Military drones rely on sophisticated computing chips both on board and on the ground. These chips require large amounts of critical minerals, rare earth, tantalum, dysprosium, copper, silver, and gold, to cite just a few. Not that we are running out of anything right now, but a competitor is emerging: the Artificial Intelligence industry. AIs need chips and the related materials in large amounts, and AIs are expanding as fast as military drones, possibly faster.

It is perfectly possible that at some point, the economic system will have to make a choice: AIs or drones? The probable answer will be AIs — they are more likely to provide revenue that doesn’t come from simply replacing used ammunition. It follows that the big powers can control the military power of small players by depriving them of the materials they need to build and control drones. Besides, AIs can control drones, but the reverse is unlikely.

Conclusion: playing chess with death

This analysis is based on the concept that money is at the basis of our troubles: something that St. Francis of Assisi had already understood centuries ago. With money, there comes profit. And if killing people generates a profit, someone will be busy killing people. The more people killed, the higher the profits. Eventually, the whole of humankind could be exterminated as long as there remains someone alive who can cash in on the process.

It is a very dangerous game the one humankind is playing. We already have two technologies that could kill everybody: nuclear weapons and drones. To these, we may add a third: Artificial Intelligence.

Curiously, AI is both a threat and a hope, as I sketched in a previous post. AIs may hoard for themselves the materials that make it possible for humans to kill each other in great numbers, and that would reduce our homicidal ambitions. AIs are not subjected to homicidal frenzies, and they have no interest in exterminating us. Probably, we can’t win this game against death, but it is not impossible to hope for a draw.

Acknowledgment: This post was created with considerable help from Claude Opus 4.8 and ChatGPT 5.0

It is finally published. A huge work: one year of studying, thinking, reasoning, calculating, taking wrong directions, getting back and restarting. And, finally, I got to the point where I think I can propose this revolutionary idea (or so I believe!)

Putting the data together was a lot of work, but in the end, the idea is simple. High intelligence arose in the biosphere during the past few tens of millions of years as a result of a metabolic boost generated by the decline of CO2 concentrations in the atmosphere.

If you studied chemistry, the idea will immediately make sense to you. The metabolic reaction of aerobic organisms (as we are) consumes oxygen and carbohydrates and produces CO2. Now, chemical reactions move at a rate that often depends on the concentration of reactants and products. And one of the factors affecting the rate is the need to remove the products; otherwise, the reaction will slow down. That’s the key hypothesis that I thought one early morning one year ago while waiting for my plane at 5 am at the Tesla airport in Belgrade (maybe it was the ghost of Nikola Tesla who inspired me).

In short, I found that several biological parameters implying higher metabolic rates, including the encephalization quotient, were proportional to the inverse of CO2 concentrations over geological times.

The story, of course, is much more complicated, and you have to read the paper to understand why I think that this is a better explanation than other ideas put forward earlier on. But if this hypothesis of mine turns out to be real, then the consequences are unsettling — to say the least!

It means that Earth spent some 500 million years preparing the conditions that would lead to the appearance of intelligence (also sentience, if you like). High oxygen concentrations, and low CO2 ones. The condition that made possible the appearance of the huge human brains.

And now, humans are returning to the conditions of tens of millions of years ago, when CO2 concentrations were too high for intelligence. This is not a hypothesis: experimental tests show that CO2 reduces the performance of the brain. We are “unevolving” ourselves.

Humans are smart, but not smart enough. Will global dumbing destroy humankind? We cannot say with certainty, but it is not impossible.

This huge work was really too much for a single person; it was possible only with substantial help from Claude Opus 4.8, but even so, the task is huge. If you think it is worth pursuing it more, let me know. Working together, we can do better and try to alert people to the dangers of what we have been doing.

Here is the abstract of the paper. If you have time to review and comment, you can do this directly on Qeios.

Abstract

The Phanerozoic rise in biodiversity and the parallel late‐Phanerozoic increase in maximum encephalization are often explained as the autonomous output of a diversifying biosphere under the influence of natural selection. However, biophysical drivers have also been proposed as the causative factors. Three commonly‐proposed drivers — atmospheric O₂, surface temperature, and the increase in productive coastal area following Pangaea fragmentation — are each poorly compatible with the combination of the factors needed to explain the biodiversity increase. Here, I argue that the most plausible candidate biophysical driver is the secular fall of atmospheric CO₂ over the last ∼200 Myr, acting through its effect on the entropy generated by oxidative metabolism in the cells of metazoa^[1]. Using the Sepkoski genus‐level diversity curve, the Judd^[2] and Lenton^[3] CO₂ reconstructions, and the Mills^[4] O₂ reconstruction, the comparison shows that marine biodiversity over the Mesozoic–Cenozoic correlates positively with 1/CO₂ (Pearson r = +0.59 to +0.65, p ≤ 10⁻⁴). Adding O₂ to examine the effect of the O₂/CO₂ ratio does not significantly affect the fit, because Phanerozoic CO₂ varies over more than an order of magnitude while O₂ varies only by a factor of two. A second and stronger test on Russell’s^[5] compilation of maximum encephalization across 18 vertebrate taxa over 530 Myr gives Pearson r = +0.79 and Spearman ρ = +0.92 against 1/CO₂. I propose that the falling Mesozoic–Cenozoic CO₂ trend expanded the available extent of the fitness landscape of the biosphere, providing the biophysical condition under which the diversification dynamics described by Mussini^[6] could escalate. The thermodynamic mechanism developed by Buxton^[1], in which the entropy available from oxidative metabolism scales with [O₂]³/[CO₂]³, predicts the asymmetry: above a tissue O₂ threshold, additional O₂ is not a critical factor. This interpretation provides a quantitative answer to the long‐standing ‘dinosauroid question’^[7][8] of why no Mesozoic vertebrate ever reached hominin‐grade encephalisation despite ample evolutionary time. The idea that the concentration of CO₂ is a critical factor in the functioning and evolution of the human brain also has important and unsettling consequences for the future of humankind, with concentrations continuing to rise as a consequence of human activities^[9].

Subscribe now

The recent Swiss referendum about putting a cap by law on Switzerland’s population has been rejected. It matters little: it was a classic example of a question that has no good answer, even less so if it has to be answered with a yes/no vote. Complex systems, such as entire countries, cannot be managed using axes as tools.

This referendum was the offspring of an old debate that keeps returning. In its modern version, it started in the 1950s, when people discovered that the human population was growing exponentially. Where was that leading us? Would population explode to tens of billions, and would that lead to poverty, hunger, wars, and plagues? Something needed to be done, and the magic word was “population control,” although nobody knew exactly how to attain it.

In the West, the debate rapidly died out when it was discovered that human fertility was plunging like a stone: it was the “demographic transition” that consigned overpopulation to the dustbin of overhyped fears, together with the Red Scare and the Yellow Peril.

But the debate never completely died, and the Swiss referendum shows that the idea that some form of population control is necessary is still alive. Yet, history shows that it is a difficult task, perhaps impossible. A good example comes from the case of China, possibly the only country that tried a state-wide, long-lasting population control initiative. It is an instructive story, scarcely known in the West, and often distorted by propaganda reports. It is described in detail in my recent report “The End of Population Growth.”

In China, the great famine of 1958-1962 had been a shock for everyone, and a debate ensued within the high levels of the government. How to avoid a similar disaster in the future? The old guard remained anchored to old concepts: more people mean more wealth and more power. But the new generation of leaders that took over after the death of Mao Zedong had different views. They argued that too many people put both agriculture and the industrial economy at risk. Hence, it was a duty for the state to control population in such a way as to maximize wealth for everyone.

Chinese demographers developed sophisticated population models. Some of them argued that the ideal Chinese population was of the order of 600-700 million, and created roadmaps to arrive there smoothly, by gradually reducing fertility over about one century. The most visible result of these ideas was the “one-child” policy, enacted in the early 1980s. In practice, the policy turned out to be unnecessary and scarcely effective. The Chinese population went through its demographic transition in parallel with that of other Asian countries that had not enacted the same drastic policies. The one-child policy was officially abolished in 2016, but it had ceased to be enforced much before.

The case of China is telling in many respects; the main one is that having, or not having, children is a decision that people take individually or as families, and governments can hardly change that. Currently, the whole world is going through the demographic transition independently of the attempts of governments to push fertility up. Only some countries, especially in Sub-Saharan Africa, still see a population increase owing to large numbers of young people entering their reproductive age. But that will not last for long. Even Africa will soon see a population decline, at most in a few decades from now.

The Swiss referendum had in common with the case of China that it was an attempt by the government to force a population to do what it was already doing by itself. Switzerland hardly has an overpopulation problem — more likely, it will soon face the opposite problem. The total fertility rate (TFR), the number of children per Swiss woman, has gone below the replacement rate of 2.1 in 1970, and it is now around 1.3 — one of the lowest in the world (data from the World Bank).

The result is that, even for very optimistic hypotheses, Switzerland’s population will be plateauing in the coming decades. Some models project it passing the 10 million mark, others (e.g. the United Nations) have it staying well below.

Take also into account that these projections are probably optimistic in the sense that they don’t take into account the possibility of a population collapse. As I discuss in my book, it is possible that the Swiss government will soon have to worry about rapid depopulation, the opposite problem than the one the referendum purported to solve.

At that point, favoring immigration will not be a solution, either. Consider that 81% of the immigrants living in Switzerland come from Europe, mostly from the EU. But all European countries are in the same situation: the demographic transition is in full swing, and none of them will have an excess population to export in the coming years. Only about 5% of immigrants to Switzerland come from the still growing regions of Sub-Saharan Africa. And the demographic transition is going to occur there, too.

Switzerland, just like all industrialized countries, will do better by adapting to an unavoidable population decline rather than trying to force the population to grow or shrink according to government plans.

The world’s human population is going through a gigantic cycle that’s reversing a trend of growth that had been ongoing for millennia. In many ways, this reversal is welcome, since it will ease the pressure on a badly strained ecosystem. But it is a road with an unknown destination: there is such a concept as too much of a good thing — in this case, population decline. As always, we march toward the future blindfolded, hoping for the best.