Read more

A clarification is worth making at the outset, because “the multiverse” is really an umbrella for more than one idea. One is Everettian many-worlds: branches of a single universal quantum state, which is where this essay begins. Another is the cosmological multiverse: potentially separate universes, or “bubbles,” thrown off in some models of the early cosmos, which is where it ends. These are not simply the same established mechanism seen at different scales. Part of what follows is an attempt to see how far one thread — chance, and being a single outcome among many — can be drawn through both, without pretending that physics has shown them to be the same thing.

What actually splits the world

The scientific backbone of the “many worlds” picture is an interpretation of quantum mechanics proposed by the physicist Hugh Everett in 1957. Ordinary quantum theory tells us that before we look, a particle can be described as a blend of possibilities — an electron spinning “up” and “down” at once. This superposition is not merely a convenient description of our ignorance: its interference effects are experimentally observable, in famous experiments where a single particle behaves as though it had travelled two paths at once. What the superposed alternatives actually represent before measurement, however, is precisely where interpretations of quantum mechanics begin to differ. The shared puzzle is that when we finally measure the particle, we find just one outcome. The blend seems to “collapse” to a single answer.

Everett’s radical proposal was to take the mathematics at face value and deny that anything collapses at all. Every possible outcome happens. The universe simply branches so that there is a version of events — and a version of you — for each result. Crucially, the branching is not triggered by human decisions. It is triggered by quantum events: a particle going one way or another, an atom decaying or not. Your everyday choices only enter the story insofar as they trace back to such quantum forks, which is itself a matter of debate.

So what, precisely, causes a fork? The answer has two ingredients. The first is superposition — the raw undecidedness described above. The second is a process called decoherence. When a quantum system in a blend interacts with its large-scale surroundings — a detector, the air, a stray photon — the “which outcome” information leaks out into countless particles almost instantly. Once that information has spread into the environment, the possibilities can no longer meet back up and interfere with one another. For all practical purposes, and in the many-worlds reading for real, they have become separate, self-contained branches. That splitting is the fork.

A common myth holds that a conscious observer is needed to make this happen — that the universe waits for someone to look. It does not. A dust grain in deep space triggers decoherence constantly, with no one watching. “Measurement” simply means any interaction that copies quantum information into enough surrounding matter that it cannot be undone. A human eye qualifies, but so does a photographic plate, a Geiger counter, or a single molecule of air.

Why the boulder rolls but the mutation matters

If everything is made of quantum fields at bottom — and we are, along with the geranium on the windowsill and the boulder on the hillside — then quantum events must touch everything. This is true. But the important question is not whether they touch big things; it is whether their effects amplify into something noticeable or wash out before they get there.

Consider the boulder. It is made of something like ten trillion trillion atoms, each with its own tiny quantum jitter. But those jitters point every which way and very nearly cancel, the same way a huge number of coin flips comes out close to fifty-fifty rather than wildly lopsided. The leftover uncertainty in the boulder’s position is smaller than the width of an atom — utterly swamped by ordinary things like friction and the shape of the ground. On top of that, the boulder is constantly “measured” by its surroundings, so it decoheres instantly and never gets to be in a genuine blend of two paths. Quantum mechanics is still true underneath, but for objects that large and warm it hands back plain, predictable, classical physics.

For most of the everyday world, then, the quantum layer is real but drowned. Yet there are special arrangements where a single quantum event is amplified all the way up to human scale. That is the entire point of Schrödinger’s famous cat: one radioactive atom either decays or does not — a genuine quantum fork — and it is wired to a device that either does or does not release a poison. Someone has deliberately built a machine to blow one microscopic coin-flip up into a life-or-death difference.

Nature builds such amplifiers on its own, no laboratory required. A single high-energy photon or radioactive decay can strike one base pair in a strand of DNA and cause a mutation — which might do nothing, or might seed a cancer, or might, over evolutionary time, supply the raw variation that natural selection acts upon. That is an unbroken path from one quantum event to a consequence the size of an organism, or a species. The human eye, remarkably, contains cells that can respond to a single photon. And chaotic systems like the weather magnify tiny differences relentlessly, so a quantum flicker that is negligible today can, in principle, be blown up into a different outcome weeks later.

Where the science genuinely runs out is the human brain. It is tempting to reason that because we are made of quantum stuff, our choices must have a quantum element. The mainstream view is sceptical: the brain is warm, wet and busy, an environment in which quantum blends would decohere almost instantly, so our decisions are effectively driven by ordinary — if fantastically intricate — neurochemistry, rather than by long-lived quantum superpositions within the brain. A minority of researchers speculate otherwise, but it remains unproven. Single quantum events certainly reach up into biology; whether they steer your thoughts is an open question.

The door left ajar

Here is a subtlety that rewards a second look. It is tempting to picture decoherence as a switch that usually flips to “decohere” but might occasionally fail. That is not quite right. Decoherence is not a separate event that fires or doesn’t. Rather, when branches entangle with their environment, their ability to interfere again does not drop cleanly to zero — it drops to something fantastically, absurdly small, but never exactly nothing. A residue always remains.

This means such possibilities are, strictly speaking, never forbidden — only suppressed. Quantum theory does not hand down a flat “impossible” for a room’s worth of air spontaneously crowding into one corner, or for a boulder tunnelling through a hillside rather than rolling down it. These sit in the far tails of the probability distribution, the outcomes we are right to refuse to round all the way down to zero.

The catch is how far down the tail they sit. For a single particle, quantum tunnelling and other small departures from the expected are routine and observed. For a macroscopic object — a boulder, a roomful of air — the suppression is so fantastically severe that such events are physically negligible on any realistic timescale; you could wait far longer than the present age of the universe and, overwhelmingly likely, never see one. The effect is real in principle and absent in practice.

In the many-worlds picture these vanishingly unlikely outcomes are not just abstractions. They would correspond to branches that genuinely exist — a branch where the air gathered in the corner, where the boulder tunnelled through the hill. Many-worlds does not make such outcomes more likely; it makes them concrete, as branches of almost unimaginably little “weight.” Exactly how ordinary quantum probability — the Born rule — should emerge from branch-weight in a theory where every outcome occurs remains one of the most debated questions in the many-worlds programme; proposed derivations exist, but whether any of them truly succeeds is itself contested.

One of many, not an accident

Follow this far enough and a personal question surfaces. If we sit at the end of a long chain of quantum forks, are we simply the product of an accident?

In one sense, yes: our existence is contingent to a staggering degree. The layers stack up. A particular universe with life-permitting constants; a supernova to seed our region with heavy elements; a specific cloud collapsing into this Sun; a Mars-sized body striking the early Earth to give us a stabilizing Moon. Then biology, where the contingency is starkest. The evolutionary biologist Stephen Jay Gould captured it with an image: “replay the tape of life.” Rewind Earth to the beginning and let it run again, he argued, and you would get a living world utterly unlike ours, almost certainly with no humans in it, because so much hinged on accidents of survival and timing. The asteroid that ended the dinosaurs some sixty-six million years ago is the emblem: alter the timing or location of that impact and the aftermath might have unfolded very differently, and the small mammals that eventually became us might never have got their opening.

But “accident” is the wrong word, and a better phrase is “chance as one of many.” It drops the flavor of error and keeps only the truth — that ours is one draw among many realized outcomes, no more mistaken than any other, simply the one we happen to inhabit. The phrase even travels across every scale: one branch among the many the wavefunction contains; one replay of the evolutionary tape among those that could have run; one habitable universe, perhaps, among a vast ensemble of mostly empty ones.

The many-worlds framing sharpens the point. It is not that one lucky fork went our way while we might have ended up elsewhere. Every fork went every way; all the branches are real. So a being asking “are we an accident?” can only exist in the branches that produced it. The improbability is genuine, but it is like a lottery winner marveling at their luck: from the outside the odds were astronomical, yet the only people around to be astonished are the winners. Conditioned on the plain fact that we are here to ask, our branch is not a fluke we happened to land in — it is the only kind of branch from which the question can be asked at all.

It is worth noting that not everyone agrees the tape would replay so differently. The paleontologist Simon Conway Morris argues nearly the opposite of Gould: that evolution has powerful attractors — eyes, wings and intelligence have each arisen independently many times — so that even a replayed Earth might produce something clever and eye-bearing, if not us. On that view we are less an accident than a recurring theme reality keeps rediscovering. Which of them is closer to right remains unsettled.

All the way down

There is one last place to point the question, the place where every thread frays into honest not-knowing. Was the Big Bang itself a fork?

Take it in two senses. Was the birth of our universe a quantum event, the kind of thing that branches? Here the serious answer is a cautious “quite possibly.” A field called quantum cosmology explores the idea that the origin of the cosmos was a quantum process rather than a classical one. Proposals such as the “no-boundary” model of James Hartle and Stephen Hawking, or Alexander Vilenkin’s picture of a universe arising by quantum tunnelling “from nothing,” treat the birth of a cosmos as the same kind of happening as a particle tunnelling or decaying — with all the branching, probabilistic character that implies. These are serious, actively studied proposals rather than established accounts of how the universe began, and each remains far beyond the reach of any direct test.

Was it, more precisely, one fork among many? One influential extension of inflationary cosmology, known as eternal inflation, suggests something like this. Inflation itself — a brief burst of extraordinarily rapid expansion in the first instant — is a central and well-supported part of modern cosmology; the further idea that inflation, once begun, never entirely stops, and endlessly buds off separate “bubble” universes, is a more speculative proposition. On that picture the Big Bang was not the beginning of everything but the moment one small region stopped inflating and settled into a universe — a bubble in a vast, perhaps eternal, inflating background. Each bubble would be its own Big Bang, seeded by a quantum event, and ours would then be, in the most literal way, one outcome among an unimaginable many, with different bubbles possibly freezing in different physical constants. That is where the fine-tuning question finds a possible home: ours is habitable because the barren ones have no one in them to notice.

It is worth being explicit here, because the essay has now reached its two endpoints. This cosmological “many” is a different idea from the Everettian branches we began with — not the same quantum mechanism seen at a larger scale. The phrase one of many fits both cases, and that resonance is part of what makes the multiverse such a seductive subject; but it is an analogy between two distinct theoretical pictures, not a demonstration that quantum branching gives rise to bubble universes, or the reverse. The unity is in the idea of chance and contingency, not yet in the physics.

And yet all of this quietly assumes there was already something for the fork to happen in — a quantum background, a sense in which “before” and “could-have-been-otherwise” apply. The Big Bang may be exactly where those assumptions run out. If it was the origin not merely of matter but of time itself, then asking what caused it, or whether it was a fork, may be like asking what lies north of the North Pole: a grammatically fine question that points nowhere. A fork needs a before and an after to fork between; if time begins at the Big Bang, there is no “before” for the forking to occupy. Even “from nothing,” in those theories, does not mean the philosopher’s nothing but a particular quantum state with no classical space or time — a stranger thing than the word suggests. The theories that seem to answer the question really push it back a step.

So the honest final word is a divided one. It is genuinely plausible — in the eternal-inflation picture, almost expected — that our Big Bang was one chance outcome among countless others, a fork in the fullest sense. But whether the deepest beginning of all was itself a fork, or whether “fork” and “cause” and “before” simply cease to mean anything down there, may be a question we are not built to answer from inside the very thing we are trying to explain. That may not be a temporary gap. It may be a permanent horizon.

There is something fitting in that. A chain of reasoning that begins with a single particle’s undecidedness ends at the edge of what can be asked — and the most honest thing to do is to look at that edge clearly, without pretending to see past it, and without pretending it is not there.

Hold the single most important number in this field, because everything follows from it. The human brain performs the most sophisticated computation known to us — perception, language, planning, the reading of this sentence — on roughly twenty watts. A dim bulb. The machines we build to approximate a fraction of that draw megawatts, and the frontier models now under construction are reckoned in gigawatts. We are feeding the output of power stations into silicon to chase a result that three pounds of tissue achieves on the budget of a nightlight.

Why is biology so far ahead? Partly because it never separates memory from processing — the brain computes where it remembers, sparing itself the constant shuttling of data that wastes most of a chip’s energy. Partly because it is event-driven: neurons stay silent and fire only when they have something to say, rather than a clock ticking billions of times a second whether or not anything is happening. Partly because it is analogue and massively parallel, computing with physics rather than forcing the world through binary switches. And partly because it is the product of hundreds of millions of years of ruthless editing, in which every surplus watt was a survival cost. Efficiency was never a feature. It was the price of staying alive.

I. The thing is already real

This is the part that surprises people: the field has left the laboratory and entered the price list.

In 2022 an Australian company, Cortical Labs, showed a culture of living neurons — grown across an electrode array, wired into a simulated world — teaching itself to play Pong. In March 2025 they turned the experiment into a product, the CL1, billed as the world’s first commercially available biological computer. It integrates hundreds of thousands of lab-grown human neurons on a silicon chip, keeps the cells alive for up to six months in a self-contained life-support system, sells for around thirty-five thousand dollars, and draws under a kilowatt. A Swiss startup, FinalSpark, rents time on dishes of living neurons over the internet — a model the industry has started calling wetware as a service. The academic programme behind it, organoid intelligence, was set out in a 2023 manifesto in Frontiers in Science: scale up clusters of human brain tissue grown from stem cells into durable, learning substrates, and wire them to machines.

So the literal question — can computation and learning run on living tissue rather than in a data centre? — is no longer speculative. It is shipping. You can buy it.

II. The line the dream keeps crossing

But there is a distinction the dream blurs, and getting it right is the whole discipline of thinking clearly here. It is the difference between substrate and transfer.

Running computation on living cells is real, and described above. Taking an intelligence that already exists — a large language model, the kind of system whose output you may be reading — and moving it into a body, the way you copy a file between drives, is something else. It runs into a wall that is not about funding or daring but about what these things are. A modern model is not a program in the ordinary sense. It is a lattice of hundreds of billions of numbers, discovered through months of training. A clump of neurons has no slot to receive them. You cannot pour a trained matrix into living tissue any more than you can email a symphony into a violin. Biology grows its intelligence. It does not accept a transfer.

The obvious objection does not rescue the dream. We can now store staggering quantities of information in DNA, at densities millions of times beyond conventional media — and not only in test tubes, but inside living cells. Surely, the thought goes, that solves it: encode the mind in the cells. But storage is not computation, and neither is running. DNA is the greatest archive nature ever produced and a hopeless processor. Writing crawls along at megabits per second. Reading means physically sequencing the molecules — hours of chemistry, where a thinking mind needs its memory in nanoseconds. You could store a model in a body. You could not run it there. The body would be the world’s most elegant cold-storage vault — a library where every book exists, perfectly preserved, and reading one means dissolving it first. The books don’t think.

What DNA storage does offer is quieter and more profound: it completes the picture of a fully biological machine. Life already divides this labour exactly. It computes with neurons — fast, hungry, live — and remembers with nucleic acid, slow and dense and near-immortal. A future biological computer that thinks with tissue and archives with DNA would not be bolting two gimmicks together. It would be copying the architecture of every cell you own.

III. You cannot strip the brain for parts

If you cannot pour a mind into a brain, perhaps you can borrow the brain’s power. Here a second instinct misfires, and the correction matters. The brain is not fast. Component for component it is embarrassingly slow — a neuron fires a few hundred times a second, a signal crawls down an axon at metres per second, a synapse takes a millisecond, while a transistor switches a billion times a second. At a single operation, silicon beats tissue by a million to one. What the brain has is not speed but the efficient parallelism described above: eighty-six billion neurons, all working at once, for almost nothing.

Frame it that way and the brain’s abilities are already being harnessed for computation — in proportion to how much of the person you leave behind. Borrow the architecture and you get neuromorphic chips, like Intel’s Loihi and IBM’s NorthPole: silicon that spikes and fires only when signalled, achieving dramatic efficiency gains on certain workloads for the same reason your head does. Rent the tissue and you get the organoid computers above. Piggyback on living people’s spare cognition and you get something most of us have done without noticing — every warped word you typed to prove you weren’t a robot donated two seconds of your visual cortex to digitising books and training vision systems. reCAPTCHA was a planetary-scale hijack of human pattern-recognition, hiding behind a security check.

But the strong version — wiring a conscious brain and running your computation through it, the way a server runs a model — fails twice. It fails technically, because you cannot read thoughts in or out fast enough; the brain has no address bus, and its memory is not separable from the wiring that is the person. And it fails morally, because at that point you are no longer borrowing an ability. You are conscripting a mind. The brain’s genius is integration. The reason it computes so beautifully is the reason it cannot be cleanly hijacked: the ability and the owner are the same object.

IV. The hybrid that works is addition, not fusion

So the most defensible idea in this whole territory is not uploading or hijacking, but augmentation — leaving the person whole and adding to them. And the right beachhead is memory, not as a guess but because it is where the most advanced work has already succeeded. Funded by DARPA’s Restoring Active Memory programme, Theodore Berger and Dong Song built a hippocampal prosthesis. They modelled the transformation a memory undergoes as it passes through the hippocampus, distilled the firing pattern of a successful memory into code, and played it back into the brains of epilepsy patients during recall. Memory improved by roughly a third. The tell that it was real rather than a placebo of electricity: random stimulation at the same strength did nothing. Only the patient’s own memory code worked.

Memory yielded first because the hippocampus is a legible funnel with a readable code at a defined bottleneck. Move toward general cognition — language, reasoning, the contents of abstract thought — and two things collapse. We do not know the code; we can decode motor intention and crude imagery, but not the neural representation of a concept. And the bandwidth is a drinking straw: a thousand electrodes against eighty-six billion neurons. So the realistic near-term hybrid is not a blended mind but a prosthesis the brain learns to wield as its own. The brain already absorbs tools into its self-model — a blind person’s cane, a musician’s instrument. An AI that watches what you encode, stores it perfectly, and feeds it back through a channel your brain learns to read as native recall is not the AI thinking your thoughts. It is the AI becoming the part of you that never forgets.

V. Do not draw the ceiling too low

It would be a mistake to read these walls as permanent. The history of computing is the warning: in 1984, Elite fit an entire 3D galaxy of thousands of star systems into thirty-two kilobytes, by refusing to store the universe and instead generating it from a tiny seed. “Impossible on this hardware” is a verdict that keeps being overturned.

But there are two kinds of limit, and they behave differently. Efficiency limits — “we know what to do, we just can’t fit it” — are soft; cleverness demolishes them. Bandwidth, power, model size, the cost of mapping tissue: all crumbling on steep curves. Knowledge limits — “we don’t yet know how the thing works at all” — are different. You cannot write clever code for an algorithm nobody has discovered. The neural code for cognition is a knowledge gap, and cleverness compresses the known; it does not conjure the unknown.

And yet even that wall is softer than pure reason suggests, for a reason the field’s own history supplies. DeepMind set out brain-shaped — its mission to “solve intelligence” and then use it to solve everything else, its founder a neuroscientist mining the brain for ideas. But look at what actually worked. AlphaGo, AlphaZero, AlphaFold — none of them replicate the brain. They reached superhuman capability by abandoning biological fidelity. “Build a powerful intelligence” and “copy the brain” turned out to be separable goals, and the field took the first. The lesson for any hybrid is liberating: you may not need to solve the brain at all. The hard road is to decode human cognition and then merge with it. The soft road is to build a capable AI, build a good-enough interface, and let the brain’s own plasticity do the integrating — machine and mind meeting halfway, without anyone ever fully cracking the code.

The cautionary counterweight is the worm. The 302-neuron nervous system of C. elegans has been fully mapped since 1986, and the open effort to turn that map into a faithful living simulation has run since 2011 — and still isn’t finished. Three hundred and two neurons, four decades of wiring diagram, and the simulation remains incomplete. Full bottom-up replication is real and advancing, but the people closest to it give cautious dates while the company-builders give certain ones. The good news is that the hybrid probably does not depend on it.

VI. The clearest good, and the turn it conceals

Set the speculation aside, because the least speculative part of this is also the most humane. Aids for disability are not the distant frontier of brain–machine work. They are its most mature corner — the place the field has already delivered. A man with ALS who could no longer speak had arrays placed in his speech cortex, and a neural network now decodes his attempted speech and voices it aloud in a synthesis of his own pre-illness voice — accurately enough that he uses it daily with his family. A stroke survivor silent for eighteen years was given a streaming brain-to-voice system that synthesises speech in real time. A “digital bridge” of wireless implants, reconnecting brain to spinal cord across the site of injury, let a paralysed man walk again. And much of it needs no surgery at all: apps and glasses that describe the visual world aloud are already in millions of blind users’ hands.

Restoration raced ahead because it is mostly an efficiency problem of the soft kind. The person still generates the command, the substrate is intact, the target is known; the job is to rebuild a bridge to a capability that is still there.

But restoration shades into enhancement, and there one project quietly becomes two. Restoring a lost function leans on existing machinery. Adding a function the brain never had — perfect recall, a new sense, knowledge poured straight in — has no intact code to rebuild toward, and inherits the hard knowledge problem entire. The ethics part ways even more sharply. Giving someone back their voice commands near-universal assent. Enhancement opens questions restoration does not: fairness, where a cognitive advantage bought with money would be a new and ugly inequality; coercion, the moment enhancement becomes a hiring edge and “optional” quietly erodes; and consent about altering what a person fundamentally is.

VII. Wanting better is the engine

And once augmentation is possible, the discussion stops being about machines. Every enhancement technology eventually becomes a reproductive technology, because parents inevitably ask whether an advantage can be passed on rather than merely purchased. The question moves from the clinic to the cradle — and the moment it does, it meets the oldest force in human nature.

Because enhancement-seeking is not a possible future direction for our species. It is the through-line of our entire relationship with the living world. We took a scraggly Mexican grass and made maize. We took one wild mustard and, by selecting different parts of it, produced cabbage, broccoli, kale, cauliflower and brussels sprouts — several vegetables from one plant. We turned wolves into everything from lapdogs to mastiffs, and a Soviet fox experiment showed we could pull a wild animal toward tameness in a few generations by selecting a single trait. The instinct is total, and it is already turned on ourselves: polygenic embryo screening exists now, offered to IVF patients in some countries, and its own ethicists concede that use will spread simply because parents want desirable traits. Even partner choice is the instinct woven into intimacy — we pair non-randomly on a dozen heritable traits, reading “good genes” in everything from height to wit. Our defining characteristic may not be intelligence at all. It may be dissatisfaction with what we were given.

So the honest question is not whether we will enhance. We will. It is what form the enhancing takes — and here the precedent contains its own map of how it goes right and how it goes wrong.

It imports the instinct but not the moral structure. Breeding a cow or a cob is morally easy precisely because the bred thing has no life of its own to lead, no claim against you, no future self who must live as the result. When the “improved” entity is a person — a future child — you act on someone who will inhabit the outcome, cannot consent in advance, and whose interests can diverge from the chooser’s. The vegetable gets no vote. The child does, eventually, and cannot yet.

The breeding precedent is also a cautionary tale. Hard selection for a narrow visible target produced the pug that cannot breathe, the dysplastic hip, whole lineages bottlenecked into fragility — “improvement” defined more narrowly than the creature’s wellbeing, every time. With cognition we understand the system far less than a Victorian dog-breeder understood shape, while the tools are weak and the predicted gains small. The precedent cuts both ways. It proves we can and will. It also shows what happens when we optimise hard for something we do not understand.

Then there is the fact no one writing about this gets to skip. The one time humans applied breeding logic to humans as deliberate, centralised policy, it produced the eugenics catastrophe — the forced sterilisation of tens of thousands, and worse, all of it conducted in the explicit language of “improvement.” Eugenics did not arrive as a fringe evil. It was mainstream, respectable, cross-spectrum science, endorsed across the political divide, because the improvement framing looked advantageous and the moral brakes were treated as soft. That is the danger your own argument predicts, and it shows exactly where it lives: not in the instinct, but in the structure through which it runs.

VIII. The race, and the brake

There is a darker mechanism still. Morality may be contested, and once a few defectors take an advantage from an unethical practice, the rest are dragged after them. The doper who wins forces the clean athlete to dope or lose. “If we don’t, someone else will” has dissolved more ethical lines than open malice ever has. When a scientist gene-edited two babies in 2018, the fear was never that the world applauded — he was condemned and imprisoned. The fear was that a single defector moves the baseline, because now the thing has been done.

The race to the bottom is real. But it is a tendency, not a law, and the difference matters more than any reassurance. Human reproductive cloning sat within reach for two decades, and the near-universal prohibition has largely held. The Montreal Protocol pulled the entire world off the chemicals eating the ozone layer, against every short-term incentive to cheat. Slavery went from a near-universal institution to near-universally abolished. What separates the collapses from the holds is fairly legible: how detectable defection is, how concentrated the payoff, whether the players can coordinate, how deep the taboo runs. Enhancement could go either way — which is the same place every step of this has landed. The instinct does not fix the outcome. The governance does.

Nor does “morality is subjective” rescue the pessimism, though it is often smuggled in to do exactly that. Whatever one thinks of the old metaphysical question, the leap from “subjective” to “therefore powerless” simply does not follow. The abolition of slavery was “merely” a shift in human attitudes, and it was among the most durable things our species has ever done. Subjective does not mean weak, and it does not mean arbitrary.

One last thing the record forces us to admit, and it is the hinge of the matter. The instinct to favour your own — in-group warmth, out-group suspicion — is as universal as the instinct to improve, present in every population ever studied, with no group holding a monopoly. That pairing is exactly why “better” is so dangerous in human hands. We have always racialised and tribalised our definitions of it, and eugenics is what that looks like when the brakes come off: improvement defined by the powerful, for the powerful, imposed on everyone else.

IX. The two heresies

So return, at the end, to where we began — because the word in the title has quietly changed its meaning along the way.

The first heresy was thinking that minds belong in bodies rather than machines. That one turns out not to be a heresy at all, but the truth we had forgotten. Intelligence’s native home is warm and wet and runs on a nightlight’s worth of power; the data centre is the deviation. Biological computing is not intelligence venturing somewhere alien. It is intelligence being coaxed back toward where it always lived, this time by design rather than by accident.

The deeper heresy is older and entirely ours: the conviction that we can improve on what we are given. And that one is not going away, because it is not a technology or a policy. It is us — the same reach that turned a grass into maize and a wolf into a companion, now turning, inevitably, on ourselves. The capability is given. The destination is chosen. Wanting things to be better is the engine, as it has always been. Deciding what “better” means, and who defines it for whom — whether enhancement takes the decentralised, consensual shape of a billion private choices, or the centralised, imposed shape of a breeding programme — that is the steering. And no instinct, no market, no inevitability supplies the steering. We do, or no one does.

The twenty watts were always the easy part. The heresy is the rest.

Grounding for the factual claims: Cortical Labs’ DishBrain (2022) and CL1 (2025); FinalSpark’s neuron platform; the organoid-intelligence framework in Frontiers in Science (2023); the Berger–Song hippocampal prosthesis under DARPA’s RAM programme; neuromorphic work at Intel and IBM; DeepMind’s stated mission and the neuroscience-inspired-but-not-replicated path of its systems; the long-running C. elegans / OpenWorm effort; the UC Davis and UCSF speech neuroprostheses and the brain–spinal “digital bridge”; polygenic embryo-screening provision and current professional guidance against trait selection. Specific figures — the memory-improvement result, the neuromorphic efficiency multiples, any connectome behavioural-prediction percentages — should be checked against the primary papers before publication, and described in ranges rather than single numbers where the benchmark varies.

Intermittent fasting arrived into that exhausted market with an unusual proposition. Most diets tell you what you may not eat. Fasting tells you, for long stretches, simply not to eat at all — and then, within a window, to eat more or less as you please. To someone worn down by the accountancy of calories and the moral weather of “good” and “bad” foods, this is a relief close to liberation. One rule. Watch the clock and forget the plate. The real promise is not about food at all. It is about the surrender of decision — about being handed, at last, a boundary that does the deciding for you.

This essay tries to look at that promise without the evangelism of its converts or the reflexive sneer of its critics. It asks a narrow, practical set of questions. Does it matter whether you eat once a day, twice or three times? What does the discipline actually cost? Why is the sugar so hard to let go of? And when the weight does come off — especially when it comes off fast — what comes off with it, and what gets stirred up in the process?

What fasting actually is

“Intermittent fasting” is a loose umbrella, and the looseness matters because the studies underneath it are not all measuring the same thing. Time-restricted eating compresses the day’s food into a window — sixteen hours fasting, eight hours eating, is the popular version, usually achieved by dropping breakfast or dinner. Alternate-day fasting and the 5:2 pattern keep some days normal and make others very lean. One meal a day, the regime its adherents call OMAD, takes time-restriction to its logical extreme: a single eating window, often an hour or less.

Beneath all of them sits one mechanism its champions love and one its sceptics keep returning to. The champions point to metabolic switching: after roughly twelve hours without food the body exhausts its easy sugar stores and begins burning fat, producing ketone bodies that the brain can use for fuel, and switching on housekeeping processes — including autophagy, the cell’s recycling of its own worn-out parts. Mark Mattson and Rafael de Cabo set out this optimistic case in a 2019 review for the New England Journal of Medicine, and it is a real and interesting biology. The sceptics reply, more drily, that for weight loss none of this may matter very much — because the simplest thing a fasting window does is shorten the hours in which you can eat, and so, for most people, reduce how much they eat. Strip away the romance and fasting may be, at bottom, a convenient delivery mechanism for eating less.

One meal, two meals, three

So does the number of meals matter? Here the honest answer is less exciting than either side would like.

Start with the oldest myth, the one your grandmother and several decades of cereal advertising installed: that frequent small meals “stoke the metabolism” and that breakfast is the most important meal of the day. The grazing hypothesis — six little meals keeping the metabolic fire burning — has largely collapsed under scrutiny. The energy your body spends digesting food scales with how much you eat, not with how many sittings you divide it into. A 2019 meta-analysis in the BMJ by Katherine Sievert and colleagues went further on breakfast specifically, finding that eating it was not an effective strategy for losing weight and that breakfast-skippers did not, on average, compensate by overeating later. The most important meal of the day, it turns out, was a slogan, not a finding.

At the other extreme sits OMAD. The cleanest look at eating once a day remains a small, tightly controlled study led by Kim Stote at the United States National Institutes of Health, published in the American Journal of Clinical Nutrition in 2007. Healthy adults ate the same number of calories either as three meals or squeezed into one. On a single meal a day they lost a little body fat — but they also showed rises in blood pressure and in total and LDL cholesterol, and impaired morning blood-sugar handling. The sample was tiny and the period short, so the findings should be held lightly. But they puncture the fantasy that the most aggressive window is automatically the most virtuous. One meal a day asks the body to absorb an enormous load in a single sitting, makes hitting your protein and nutrient needs genuinely hard, and tends to fray under the ordinary pressures of family meals and social life.

The popular middle ground — two meals inside an eight-hour window — is where most of the modern research actually lives, and where the most sobering recent result came from. The TREAT trial, run by Ethan Weiss and Dylan Lowe at the University of California, San Francisco, and published in JAMA Internal Medicine in 2020, compared sixteen-eight time-restricted eating against three structured meals a day. The fasting group lost a modest amount of weight — but so, more or less, did the control group, and the difference between them did not reach significance. More striking, a notable portion of what the fasting group lost appeared to come from lean mass rather than fat — a result that drew considerable attention, and some methodological challenge, and is best read against the trial’s modest size rather than treated as settled. Eat in a shorter window and you may simply eat a little less; you do not unlock a metabolic shortcut, and you may pay for careless fasting in muscle.

Three meals a day, the pattern your body and your culture already know, works perfectly well for weight loss if the total intake is contained. This is the conclusion that keeps surfacing whenever diets are compared head to head. Krista Varady at the University of Illinois Chicago, who has spent her career running fasting trials, finds again and again that alternate-day and time-restricted approaches produce weight loss roughly in line with plain daily calorie restriction — not because fasting is inert, but because the fasting is doing its work by trimming calories, which three sensible meals can do as well. The number of meals is not the lever. It is a packaging choice. The lever is how much you eat across the week, and whether you can keep doing it.

The sugar at the centre of it

Which brings us to the part everyone underestimates: the sugar. The first weeks of cutting back — on any regime, but especially one that drops a sweet breakfast or an evening treat — are genuinely, physically hard, and it is unkind to pretend otherwise. Refined sugar and the fast carbohydrates that behave like it provoke a sharp rise in blood glucose and a matching surge of insulin; the trough that follows can itself feel like hunger, so that the biscuit at eleven is, in part, the biscuit at nine still echoing. Flatten those swings and appetite tends to quieten. This is much of why low-carbohydrate and fasting approaches leave people reporting less hunger than the calorie arithmetic alone would predict.

But there is more to it than blood chemistry. Sweetness is wired to reward, learned from infancy, and woven through almost every social occasion we have. Letting it go can feel less like a dietary adjustment than a small bereavement, complete with the irritability and fog that the fasting crowd cheerfully call the “keto flu.” None of this means sugar is poison; the wellness industry’s habit of speaking about it in the vocabulary of addiction and toxicity does more harm than good. It means, more modestly, that sugar keeps appetite loud, and that a regime which quietens appetite has made the discipline easier to sustain. That is a real advantage. It is not magic, and it does not survive being oversold.

The discipline nobody sells you

Here is the finding the diet industry would rather you did not dwell on: in well-run trials, most diets produce broadly similar, broadly modest results, and most of them fail in the long run for the same unglamorous reason. People stop doing them. The variable that predicts success is not the cleverness of the mechanism but adherence — whether a person can actually keep going. Christopher Gardner’s DIETFITS study at Stanford, published in JAMA in 2018, set healthy low-fat against healthy low-carbohydrate eating over a year and found no significant difference in weight lost between them. What mattered was sticking to the plan at all. The right question, then, is not “which diet is best?” but “which one can you live with for years rather than weeks?”

Judged on that question, fasting’s great strength is its simplicity. A single rule about time is easier to hold in the head than a running tally of points. Its weakness is the mirror image of that strength: rigidity. A window that ignores the office lunch, the children’s tea, the dinner with friends, the shift that runs late, becomes a source of friction and quiet shame. And for some people — particularly anyone with a history of disordered eating — a sanctioned framework for not eating can shade from discipline into something darker, dressed up as health. The discipline that fasting demands is real and, for the right person, genuinely freeing. But it is a temperament, not a virtue, and it does not suit everyone who is desperate enough to try it.

What you lose when you lose it fast

But first, a contemporary qualification, because it would be incomplete in 2026 to weigh any of this without the drugs. The arrival of the GLP-1 medicines — semaglutide as Ozempic and Wegovy, tirzepatide as Mounjaro — has altered the landscape more than any diet ever did, and altered it at precisely the point this essay keeps circling. They work by quietening appetite, which is to say they make adherence easy in exactly the way fasting strains to. They are not an escape from the argument. They demand the same care over protein and muscle — arguably more, since rapid drug-driven loss can involve substantial lean-tissue loss if protein intake and resistance training are neglected — and they carry their own unsettled question of what happens when the prescription stops. They may prove less a rival to fasting than its pharmaceutical confession: an admission, written in prescription, that appetite was the true battleground all along.

That qualification made, the older warning still stands. When you lose weight, you do not lose pure fat. You lose a mixture of fat and lean tissue — muscle, and the water and connective material that go with it — and the faster and harsher the deficit, the larger the lean share tends to be. The most haunting demonstration of this is also the oldest: the Minnesota Starvation Experiment, run by Ancel Keys in 1944 and 1945, in which conscientious objectors were placed on prolonged severe restriction. They lost muscle, strength and warmth, and were altered psychologically in ways that lasted long after the food returned — obsessed with eating, irritable, hollowed out. It is an extreme case, but it draws the principle in heavy lines: starve the body quickly and it does not politely surrender fat alone.

In ordinary crash dieting the effect is milder but the direction is the same, and it carries a sting in its tail. Muscle is metabolically active tissue; it is part of what determines how many calories you burn at rest. Lose a chunk of it and your resting expenditure falls, which means the weight comes back more easily and the next attempt starts from a worse position. This is the engine of the serial dieter’s despair — each rapid loss leaving the body a little more efficient at storing, each regain favouring fat over the muscle that was sacrificed, so that someone can end a cycle heavier in fat than they began. In older people the stakes are higher still, because lost muscle in later life is not easily rebuilt and shades into frailty.

The protections are unglamorous and well established: eat enough protein, do resistance exercise so the body has a reason to keep the muscle it has, and lose the weight more slowly than impatience demands. And here fasting cuts both ways. A long fast funnelled into a single small eating window can make it genuinely difficult to take in enough protein to defend muscle — which may be exactly what the TREAT trial’s lean-mass signal was showing. Fasting can be done well, with attention to protein and a barbell or a set of press-ups in the week. Done carelessly, in pursuit of the fastest possible number on the scales, it spends muscle the dieter cannot afford to lose.

The poison in the larder

There is one more phenomenon worth a brief and honest hearing, because it is real and it has been thoroughly hijacked. Over a lifetime the body stores a class of stubborn pollutants in its fat — organochlorine pesticides such as DDT, polychlorinated biphenyls, dioxins, certain flame retardants. When that fat is broken down for fuel, the stored compounds are released back into the blood, and measured levels do rise as people lose weight; Duk-Hee Lee and others have documented it across dieting and bariatric-surgery cohorts. This much is not folklore.

What it means for any individual, though, is far less settled. The associations drawn in the literature — with thyroid function, insulin sensitivity, even a blunting of metabolism — are live lines of research, not established fact, and two things have to be held together. First, this is no reason to remain obese: for almost everyone the gains of shedding excess weight comfortably outweigh a transient rise in circulating pollutants. Second, it is the modest, technical finding from which the entire “detox” industry has spun its fairy tale of teas, cleanses and juice fasts. The evidence allows two conclusions at once: a real toxin-release phenomenon exists, and the detox industry built on top of it is, very largely, nonsense. If there is a practical lesson, it is the one the muscle already taught — lose the weight slowly, and give the body time to clear what the melting fat lets go.

The honest prescription

So what is left for the tired person with the drawer of three-sized clothes? Something unsatisfying, and for that reason trustworthy. There is no window and no meal count that dissolves fat by magic. The lever, every time the evidence is examined closely, is sustained and tolerable restriction — eating less, across years, in a way you can actually keep up. Fasting helps some people because it turns that hard instruction into a simple one: not what to eat, but merely when. For others its rigidity becomes one more rule to fail, one more private verdict of weakness.

And this is the reframing that matters most, the one hiding inside every honest answer above: a diet, properly understood, is not something you go on and then come off. The very word does the damage — it implies a temporary state, a sentence to be served until the weight is gone and ordinary life resumes. But ordinary life is precisely what put the weight on, and returning to it returns the weight, which is why the drawer fills with clothes in three sizes. What actually works is not a diet at all. It is a change of life: a way of eating, moving and living that you adopt not until the clothes fit but for as long as you have. A diet has an end date, and on that date it fails. A lifestyle has none — and that, unromantic as it sounds, is the whole of why it can succeed. The goal is not to lose the weight. It is to become, quietly and permanently, the kind of person for whom the weight does not come back.

If there is advice in any of this, it is modest. Choose the pattern you could imagine still following next winter, not the most punishing one you can survive for a fortnight. Defend your muscle with protein and a little resistance work, and lose the weight more slowly than your impatience wants — both your metabolism and the pollutants in your fat will thank you for the patience. Treat the discipline as a temperament to be matched, not a moral test to be passed. And if you carry a history of disordered eating, or live with diabetes, or take medication, take none of this as a programme but as a prompt to speak to a doctor first, because aggressive fasting is not benign for everyone who is desperate enough to reach for it.

The deepest dishonesty in diet culture is not that it overpromises results. It is that it sells the failure back to the person as a failure of character, when so much of it is a failure of method and of biology — bodies built to defend their fat, regimes built to be abandoned. The person who has lost the same two stone four times is not weak. They have, more likely, been given the wrong instruction, repeated louder each time it failed. A clearer instruction will not make the work easy. But it might, at last, make it honest — and an honest, dull, survivable plan has quietly outlasted every miracle ever printed on a magazine cover.

Sources & notes

de Cabo R, Mattson MP. ‘Effects of Intermittent Fasting on Health, Aging, and Disease.’ New England Journal of Medicine, 2019.

Stote KS et al. ‘A controlled trial of reduced meal frequency without caloric restriction in healthy, normal-weight, middle-aged adults.’ American Journal of Clinical Nutrition, 2007. (One-meal-a-day study.)

Lowe DA, Weiss EJ et al. ‘Effects of Time-Restricted Eating on Weight Loss and Other Metabolic Parameters’ (the TREAT trial). JAMA Internal Medicine, 2020.

Sievert K et al. ‘Effect of breakfast on weight and energy intake: systematic review and meta-analysis.’ BMJ, 2019.

Gardner CD et al. ‘Effect of Low-Fat vs Low-Carbohydrate Diet on Weight Change’ (DIETFITS). JAMA, 2018.

Varady KA, University of Illinois Chicago — alternate-day and time-restricted fasting trials and reviews (various, 2009–2021).

Keys A et al. The Biology of Human Starvation (the Minnesota Starvation Experiment), 1950; conducted 1944–45.

Lee D-H et al. — research on persistent organic pollutants (POPs) released from adipose tissue during weight loss (various).

GLP-1 receptor agonists — semaglutide (STEP trials) and tirzepatide (SURMOUNT trials); see also emerging literature on lean-mass loss and post-cessation weight regain.

Before the advice starts, one thing needs to be said directly: you are not in this position because you failed. The job that is disappearing is disappearing because the economy is undergoing a structural shift that has nothing to do with your competence or your commitment. AI is displacing entire categories of work that intelligent, capable people have done well for decades. You are not a casualty of personal failure. You are in the path of an industrial transition whose scale has not yet been honestly communicated to the people it affects most.

That said: the transition is real, the timeline is short, and the people who navigate it best are those who act early rather than wait for certainty. This guide is about acting. Each piece of advice comes with a reason. The reason matters — because at your stage of life, with your level of experience and judgment, you deserve to understand the basis of what you are being asked to do, not just be told to do it.

Section 1: The First Thirty Days

Before strategy, before retraining — the foundations

1. Get the financial picture clear, completely and honestly, within the next two weeks.

Write down every outgoing: mortgage, utilities, insurance, subscriptions, food, car, children’s costs, care contributions for parents, debt repayments. Write down every income source and its reliability. Then calculate the true number — how long can you sustain the current position without a change in income, and what is the minimum income that covers the non-negotiable commitments. This number is your operating baseline. Every decision from this point is made with reference to it. Not knowing this number is more frightening than knowing it, even if the number is uncomfortable.

2. Do not make any large financial decisions for thirty days.

The period immediately after a job loss or serious threat to income is not the period in which good long-term financial decisions are made. The psychological research on decision-making under stress is consistent: cognitive bandwidth narrows, time horizons shorten, and the temptation to take decisive action — any action — in order to feel less helpless produces decisions that are regretted later. Give yourself thirty days to get the picture clear before you act on it.

3. Tell the people who need to know — and no one else yet.

Your partner or spouse needs to know the full picture, now. Your children need to know something appropriate to their age — not the detail, but that things are changing and that you are managing it. Your parents, if their care depends partly on your financial contribution, need a realistic conversation. Beyond these, the circle stays small for now. You are not hiding failure. You are protecting your decision-making space while you work out what to do.

4. Find one person who has navigated something comparable and talk to them.

Not for answers — for the experience of hearing that this is survivable. The isolation of early job loss or serious career disruption is one of its most damaging features. The person who has been through a significant career transition in their forties and come out the other side is more useful to you right now than any careers adviser, because they have done the actual thing rather than studied it. You almost certainly know someone. Contact them this week.

This guide does not tell you not to feel the weight of what you are carrying. You should feel it — it is real and it is heavy. The question is whether you carry it in a way that stops you functioning or in a way that informs your decisions without paralyzing them. The distinction is not about being strong. It is about being honest about what you can and cannot control.

Section 2: The Honest Audit

What you actually have — it is more than you think

5. Audit your skills as they are — not as they were described in your job title.

Your job title described the role you were hired to fill. Your actual skills are the things you could do in service of that role: analysis, communication, managing complex relationships, making decisions under pressure, running projects, understanding a sector, explaining things to people who do not understand them, spotting problems before they become crises. Write these down specifically, in plain language, without job-title vocabulary. This list is often considerably longer than people expect, and it is transferable in ways the job title is not.

6. Identify the sector knowledge you carry that took years to build.

Domain expertise — the deep, specific knowledge of how a particular industry or function works, who the players are, what the real problems are, what has been tried and failed — is extremely difficult to acquire quickly and is not replaceable by AI in the way that process-based work is. If you have spent fifteen years in financial services, healthcare administration, insurance, logistics, or any other complex sector, you understand things about that sector that a generalist does not and cannot. This knowledge has market value that you may be underestimating because it is familiar to you.

7. Write down every relationship in your professional life that still has value.

Former colleagues, clients, suppliers, managers, people you have helped over the years — these relationships are assets. Not in a transactional sense but in the sense that the people who know your work, your reliability, and your character are the most likely source of your next opportunity. This is not networking in the uncomfortable, self-promotional sense. It is the recognition that most career transitions in mid-life happen through people, not through job boards, and that the groundwork for those connections has already been laid over decades.

8. Be honest about what you do not want to do again.

A forced career transition is also an opportunity to refuse the parts of the previous career that were genuinely wrong for you. If you spent fifteen years in a role that depleted you in specific ways — the commute, the politics, the type of work — this is the moment to name that clearly. The temptation is to replicate the previous role as closely as possible because it is familiar and because the speed of the transition feels urgent. But replicating the wrong thing quickly is not better than taking slightly longer to find something that is actually right.

You have something that no 22-year-old has and no AI can replicate: two decades of hard-won judgment about how things actually work, how people actually behave, and what actually matters when the situation is complicated. That is genuinely valuable. The task is finding where it is most valuable and getting it in front of the right people.

Section 3: The Financial Bridge

Buying yourself time without destroying your future

9. Contact your mortgage lender before you miss a payment, not after.

Lenders have legal obligations and commercial incentives to work with borrowers who are proactive about difficulty. Most major UK lenders offer mortgage payment holidays, term extensions, and switch-to-interest-only options for borrowers in genuine hardship. The terms available to you when you call before default are significantly better than the terms available after it. This call feels like an admission of failure. It is actually a competent financial decision that protects the asset at the centre of your family’s security.

10. Review every subscription and recurring cost in the next seven days.

The average UK household has between ten and fifteen recurring subscriptions whose combined monthly cost is higher than most people estimate when asked. Stream services, gym memberships, insurance policies that have rolled over at non-competitive rates, software subscriptions, club memberships — none of these individually is significant, but their total is often several hundred pounds a month. Cancel or pause every non-essential. This does not solve the problem but it extends your runway, and every additional week of runway is a week in which your situation can change.

11. Do not touch pension savings unless there is no alternative.

Pension savings are protected assets with long-term compounding value that is very difficult to rebuild once broken. The tax penalties and lost growth from early pension access are severe and often irreversible. They should be the last resort, accessed only when all other options have been genuinely exhausted. The options that come before them include: reduced mortgage payments, consolidated debt, income from freelance or consulting work, spousal or partner income, family support, and benefit entitlements that may apply to your situation and that many people do not claim because they do not know they qualify.

12. Check every benefit entitlement you qualify for, without embarrassment.

Universal Credit, Council Tax Reduction, Working Tax Credit, Child Benefit, Carer’s Allowance (if you are contributing to elderly parent care) — these are entitlements that you have paid into through decades of National Insurance contributions. Claiming them is not charity. It is the system working as designed. The Citizens Advice Bureau can map your entitlements quickly and free of charge. Many people in your position leave significant monthly income on the table because they feel that claiming benefits is beneath them. It is not. It is prudent.

The financial pressure of this period will be real and sustained. The single most damaging response to sustained financial pressure is the decision made in desperation — the business investment made with borrowed money, the pension accessed early, the salary sacrifice accepted from the wrong employer because something felt better than nothing. Build the bridge carefully. The bridge only needs to last long enough for the next phase to begin.

Section 4: The Transition Itself

What to move toward and how to think about it

13. Think in terms of capability, not job title.

The job title that described your previous role may not exist in five years. The capabilities it required — analysis, communication, management, domain expertise, judgment — will still exist and will still have value, in roles and organisations that may be described very differently. The person who approaches their transition by looking for the same job title in a different organisation is looking in a narrower field than the person who asks: where do my capabilities have the most value, and what are the different forms that could take?

14. Consider the consulting or advisory route seriously.

Many mid-career professionals with deep domain expertise find that the consulting or advisory model — working with multiple clients rather than one employer — provides both more income flexibility and better use of the expertise they have spent decades building. The transition to consulting is not easy: it requires building a client pipeline from scratch and tolerating income variability that employment shields you from. But it is a route that many people in your position have taken successfully, and your existing professional relationships are your most important asset in making it work. Do not dismiss it because it feels uncertain — the employment alternative is also uncertain now.

15. Target organisations that are in Phase One of the AI transition, not Phase Two.

As discussed in this series: Phase One organisations are using AI to make existing workforces more efficient. They still need experienced humans to supervise, direct, and apply judgment to AI output. Phase Two organisations have built AI into their operational backbone and need very few humans. Your target market is Phase One — organisations that are adopting AI but have not yet figured out how to get the most from it, where your combination of domain expertise and human judgment has real value and is not yet being replaced. These organisations are the majority of the market right now.

16. Learn to talk about AI without fear and without overselling.

Many mid-career professionals feel threatened by AI and either avoid the topic or overcompensate with enthusiasm they do not feel. Neither position is helpful in a hiring context. What is helpful is the capacity to say: I understand what AI can and cannot do in my field, I have used these tools in these specific ways, and my value is in the judgment and domain expertise that those tools depend on to be applied well. This is honest, credible, and positions your experience as complementary to AI rather than in competition with it.

17. Do not retrain for a technical role you are not genuinely drawn to.

The retraining industry has made significant money selling coding bootcamps and data science programmes to mid-career professionals whose job is disappearing. Some of these transitions work. Most do not — not because the people are incapable but because spending six months learning a technical skill you have no deep interest in produces a weak practitioner competing against people for whom the field is a genuine vocation. Retraining is worth the investment when it deepens or extends something you already have, not when it attempts to replace it wholesale with something unfamiliar.

The transition period is not a waiting room. It is a period in which the decisions you make about where to direct your energy and attention will shape the next decade. The people who come through this well are not the people who panicked into the first available option. They are the people who gave themselves enough financial runway to make a considered move, and then made it with the full weight of their experience behind them.

Section 5: The People Around You

Your family is a resource, not just a responsibility

18. Talk to your partner about the full picture — once, properly, and then regularly.

The worst version of this situation is the one where the person carrying the professional uncertainty also carries the burden of managing the domestic picture alone, to protect their partner from worry. It does not work. The partner senses the withheld information, fills the gap with their own anxiety, and the household loses the collective problem-solving capacity of two adults who both understand the situation. Have the full conversation. Your partner is almost certainly more resilient and more resourceful than the version of them you are trying to protect.

19. Be appropriately honest with your children without making them carry adult weight.

Children pick up on household stress with great accuracy and fill in the blanks with their imagination, which is usually worse than the reality. A child who is told something is changing, that you are working on it, and that they are not at risk of losing their home or their security, is in a better position than a child who senses tension and has no explanation for it. The level of detail appropriate depends on the child’s age. The principle — that they deserve an honest, age-appropriate account — applies regardless.

20. Have an explicit conversation with elderly parents about what is and is not sustainable.

If you are currently contributing financially or practically to elderly parent care, a period of career disruption requires an honest conversation about what that contribution looks like going forward. This is one of the most uncomfortable conversations this guide asks you to have, because it involves acknowledging limitation to people you want to be able to provide for. But a conversation now, when there is still time to make arrangements, is far better than a crisis later when the options have narrowed. Local authority care assessments, attendance allowance, and carer support services may all be relevant. The Citizens Advice Bureau can help map these.

21. Let the people who want to help, help.

The social cost of appearing to need help when you have always been the reliable, capable, providing adult is real. It is also, right now, less important than the practical benefit of accepting the help that is genuinely being offered. A friend who can provide childcare cover during an interview. A sibling who can take a turn with a parent. A former colleague who can make an introduction. These offers are not charity. They are the ordinary reciprocity of relationships built over decades, and refusing them to protect your sense of self-sufficiency costs you more than it preserves.

Section 6: Your Own Wellbeing

This is not optional — it is operational

22. Protect the basics: sleep, exercise, and one relationship that is not about the problem.

The capacity to think clearly, make good decisions, and sustain the effort of a career transition is directly dependent on physical and psychological baseline. Sleep deprivation worsens every cognitive function relevant to this situation. The absence of physical exercise removes the single most effective non-pharmaceutical intervention for the anxiety and low mood that this period reliably produces. The absence of at least one relationship in which you are not the person managing a crisis removes the recovery space that makes sustained effort possible. These are not luxuries. They are the maintenance of the instrument you are relying on to solve the problem.

23. Notice the difference between productive worry and unproductive worry.

Productive worry asks: what can I do about this, and what is the next specific action? Unproductive worry revisits the same ground repeatedly, produces no new information, and generates anxiety without generating action. When you notice yourself in unproductive worry — and you will — the intervention is not to stop worrying but to ask the productive version of the question and identify one concrete next action. Then take it. This is not a technique for making the anxiety disappear. It is a technique for not letting it consume the cognitive resource you need for everything else.

24. Do not define yourself entirely by what you do for work.

The mid-career professional who has built their identity substantially around their professional role is particularly vulnerable to the psychological impact of that role’s disappearance. The identity does not disappear with the job — but it can feel that way, and the feeling is disorienting. The antidote is not denial but the active investment of attention in the other dimensions of who you are: as a parent, a partner, a friend, a person with interests and capabilities and relationships that exist independently of your employment. This is not distraction. It is the maintenance of the self that the professional role was always only a part of.

You have navigated hard things before. You do not know that from this vantage point, because this one is in front of you and the others are behind. But the same judgment, resilience, and capacity for sustained effort that built a twenty-year career will navigate this — if you direct it clearly, protect the foundations, and give yourself the time to move purposefully rather than desperately.

Section 7: What Nobody Tells You

The things that are true and rarely said

25. It will probably take longer than you think — and that is survivable.

Career transitions in mid-life take, on average, longer than early-career transitions. The market is less forgiving of salary expectations, the roles available at the right level are fewer, and the process of moving from one professional identity to another requires a period of genuine uncertainty that cannot be shortened by effort alone. This is not a reason for despair. It is a reason to build the financial runway, manage the family communication, and protect the personal foundations described in this guide — because the transition that takes twelve or eighteen months and is navigated well produces a better outcome than the one that produces a panicked decision at month three.

26. The next chapter will probably not look like the previous one — and that may be correct.

The desire to replicate the previous career as closely as possible is understandable and often counterproductive. Many people who navigate a mid-career transition well arrive at something that uses their capabilities in a different form — consulting rather than employed, a smaller organisation rather than a large one, a role with less status and more satisfaction, an industry adjacent to their previous one rather than identical. The criteria that mattered at thirty — salary level, job title, institutional prestige — are not always the criteria that matter most at forty-five. You have the experience to know what actually matters to you. Use it.

27. Your experience is not obsolete. Your previous role may be — but you are not.

The narrative that mid-career workers are economically obsolete in the age of AI is one of the most damaging myths of the current transition, and it is wrong. What AI is displacing is routine information processing at scale. What it cannot replace is the judgment, the domain expertise, the professional relationships, the ability to read a complex human situation, and the accumulated wisdom of two decades of navigating real-world complexity. These capabilities are in increasing demand precisely because the organisations adopting AI are discovering that the technology needs experienced human direction to be useful. You are not the past. Positioned correctly, you are exactly what the present needs.

THE LONG VIEW — A PRACTICAL GUIDE FOR THE MID-CAREER TRANSITION

Companion to: The Myth of Education | After the Degree | You Are 16

This guide does not constitute financial or legal advice. For mortgage difficulty, contact your lender directly. For benefit entitlements, visit Citizens Advice. For mental health support, speak to your GP. None of these steps is a sign of failure. All of them are signs of competent navigation of a genuinely difficult situation.

The real question is not whether AI helped — help is not the issue, and never has been — but where assistance ends and authorship begins. I want to answer that directly, including the hardest version of the objection, rather than defend only the easy cases.

The ideas still come from the writer

Let me start with the point that matters most, because everything else depends on it. An AI writing tool does not have a life, a career, a set of hard-won convictions, or a reason to say anything. The writer does. When I sit down to argue that a policy is misguided, or to explain what I learned running a small business, or to describe a place I have actually stood in, the argument, the experience and the judgement are mine. The tool has none of them until I supply them.

This is worth stating plainly because the fear about AI writing is really a fear about authorship — that the human is being quietly written out of their own work. But an article is not just an arrangement of sentences; it is a point of view. The angle, the claim, the thing the piece is for, the reason it is worth a reader’s time — these originate with the person, not the machine. The AI is closer to a very fast, very patient editor sitting beside you than to a ghostwriter working in your place. You decide what is true, what matters, and what you are willing to put your name to. The tool helps you get it down.

Anyone who has tried to use these tools without their own ideas learns this immediately: ask for an article about nothing in particular and you get exactly that — smooth, competent, and empty. The substance has to come from somewhere, and without a writer bringing purpose, experience and judgement to the process, smooth prose is not enough.

This is, quietly, how most writers who use AI already treat it: in a 2025 BookBub survey of 1,229 authors, the most common use — cited by 81 per cent of AI-using respondents — was research rather than generating prose. They are using these tools to inform their own work, not to hand over the writing of it.

Speed: getting from thought to draft

The most obvious benefit is speed, and it should not be dismissed as mere convenience. Much of the time between having something to say and having said it is spent not thinking but stalling — staring at a blank page, rewriting the same opening line, losing the thread while hunting for a word. That friction is where good ideas quietly die. Plenty of things worth writing never get written simply because the effort of starting is too high.

AI collapses that gap, and the effect is measurable. In a controlled study by MIT researchers Shakked Noy and Whitney Zhang, published in Science in 2023, 453 professionals — marketers, consultants, grant writers, analysts and others — completed realistic writing tasks with and without ChatGPT. Those using the tool finished 40 percent faster while their work was independently judged to be of higher quality. You can describe roughly what you want to say and have a working draft in front of you in minutes rather than hours. A draft is not a finished piece, but it is enormously easier to improve something than to conjure it from nothing. A writer with a rough draft in hand is already halfway to a good article; a writer facing an empty screen may never begin at all. Speed, here, is not about lowering standards. It is about producing more of the work that would otherwise never exist.

Putting the ideas together

Most of us think in fragments. We have a strong opening observation, a couple of supporting points scattered in our heads, a conclusion we feel but cannot yet phrase — and no clear sense of the order they should go in. Turning that scatter into a structure is one of the genuinely hard parts of writing, and it is where AI is unusually helpful.

You can hand it your raw thinking — bullet points, half-sentences, a voice memo transcribed, a messy brain-dump — and ask it to help you see the shape. Which point should lead? What is the natural progression? Where is the argument thin? This is not the AI having the ideas for you; it is the AI helping you arrange the ideas you already have, the way talking a problem through with a sharp colleague helps you understand your own thoughts. The writer remains the author and the responsible decision-maker. The tool simply makes the assembly faster and the structure clearer.

Let me describe what this actually looks like when I write, since that is what I am really defending. I normally know what I want to say before I know how I want to say it. I will have the argument, an example and a conclusion in my head, but not yet the structure that holds them together. AI helps me turn that loose collection of thoughts into something I can interrogate and improve. Sometimes I reject what it produces outright. Sometimes it exposes a weakness in my argument I would rather not have seen. Sometimes it finds a clearer way of expressing something I already believed. But at every step I am the one deciding what survives onto the page. The tool proposes; I dispose.

What if AI gives you an idea?

Here is the objection that a careful sceptic will press, and it is the one worth confronting head-on. Everything so far assumes AI is polishing ideas that are already yours. But what if it does more than that — what if it suggests an argument, an analogy or an implication you had genuinely not thought of? At that point, the sceptic says, the defence weakens: the machine is no longer editing your thought, it is contributing to it.

The answer is that writers have never developed their ideas in intellectual isolation, and no one has ever pretended otherwise. We read books and absorb their arguments. We change our minds mid-conversation with a friend. We hear something on the radio, or employ a researcher, or hand a draft to an editor who proposes a structure we had not seen and points out an implication we had missed. None of that has ever been thought to destroy authorship. The relevant question was never did anyone or anything help me think of this? It is did I understand it, examine it, make it my own, and decide that I genuinely believe it before putting my name to it?

That is the line between assistance and abdication. If AI proposes an argument I had not considered, it does not thereby become my argument. It becomes mine only after I have interrogated it, tested it against what I know, and consciously adopted it — or rejected it. Influence has always been part of writing. Responsibility is the part that cannot be outsourced.

Polishing

There is also the work of the final mile: tightening loose sentences, cutting repetition, smoothing an awkward transition, catching the word you have used four times in one paragraph. This is slow, unglamorous labour, and it is precisely the kind of thing AI does well. A draft that says what you mean but says it clumsily can be lifted, line by line, into something clean and readable — without changing what it means.

The same MIT study that measured the speed gain also found the polish was real: output quality, graded blind by experienced professionals, rose by roughly 18 percent — and the largest improvements went to the writers who had started out weakest, which suggests the tool lifts a floor rather than merely flattering those already fluent. Polishing with AI is closest to working with a good copy-editor, and no one accuses a writer who uses an editor of being a fraud. The judgement about whether the polished version still sounds like you, still says what you intended, remains yours. You accept some suggestions and reject others. The voice is something you protect, not something you surrender. But the sheer grind of refinement — the part that exhausts writers and often gets skipped under deadline — becomes manageable.

Checking statistics and claims

A more serious point, and one that cuts against a common worry. AI tools are sometimes accused of spreading errors, and it is true that they can produce confident nonsense if used carelessly. But used deliberately, an AI assistant is a strong prompt toward more rigour, not less — provided you are clear about what it can and cannot do.

Here a distinction matters. An AI working only from its internal model is not a fact-checker; left to itself it can state a wrong figure with complete confidence, and treating its unaided output as verification is exactly the misuse the sceptics warn about. What it is good at is the prior step: flagging which claims rest on a source, noticing where a number looks implausible or internally inconsistent, and reminding you which assertions you are stating as fact without evidence. A tool connected to live search can go further and point you toward a source — but pointing is not confirming.

So the division of labour is precise. The AI identifies what should be checked; the writer verifies it against the original — the report, the official dataset, the study itself — before it goes out under their name. No responsible author should publish a statistic on an AI’s say-so alone. Treated as a checklist rather than an oracle, the tool makes careful writing easier to sustain; treated as an oracle, it is a liability.

Where authorship begins

None of this means AI writing is without risk. It can flatten a distinctive voice into blandness if you let it. It can invent facts if you trust it blindly. It can tempt a writer to publish something they have not really thought through. These are real objections, and the answer to all of them is the same: the writer must stay in charge. The interesting question, in the end, was never whether people write with AI. It is where assistance ends and authorship begins.

The answer, I think, has been latent in every section above. Authorship was never a matter of typing each word unaided. It lies in exercising judgement and accepting responsibility for what finally appears under your name. AI can suggest, organise, polish and challenge. What it cannot do is understand the argument on your behalf, believe it for you, verify it for you, or answer for it once it is published. That work — and the responsibility it carries — stays with the person whose name is at the top.

So the older question, is using AI cheating?, turns out to be the wrong one. The ideas may be prompted, the structure suggested, the prose tightened by a machine. But the responsibility was always yours. That, in the end, is what authorship means.

A note on this article: I used AI in preparing and editing this piece. The argument, the judgements and the final decisions are mine — which is, of course, rather the point.

Sources

Shakked Noy and Whitney Zhang, “Experimental evidence on the productivity effects of generative artificial intelligence,” Science, 14 July 2023 — summarised in MIT News. 453 professionals; ChatGPT users completed writing tasks ~40% faster with ~18% higher graded quality, and the largest gains went to lower-performing writers.

BookBub Partners, How Authors Are Thinking About AI, survey of 1,229 authors, May 2025. 45% of authors use AI tools; among those users, 81% use it for research.

When Morgan Stanley published analysis in 2023 projecting that GLP-1 receptor agonists could reduce caloric consumption in the United States by as much as 20 to 30 percent among the treated population, the investment community took notice. When Walmart’s chief financial officer noted in early 2024 that they were beginning to see a measurable shift in the food purchasing patterns of customers they believed to be on GLP-1 medications, the food industry started paying very close attention. When McDonald’s, Conagra, General Mills, and Kraft Heinz all mentioned GLP-1 adoption in their investor communications as a material consideration for future demand, it became clear that a pharmaceutical intervention developed to manage the metabolic consequences of the food system had acquired the scale to feed back into that system’s economics in ways that nobody had fully modelled.

This article examines what those consequences look like, where they are most acutely felt, how the industries most exposed are responding, and what the second and third-order effects of GLP-1 adoption at global scale might be for the agricultural supply chain, the labour market of the food service sector, and the governments whose fiscal positions depend on the tax revenues those industries generate.

The framing of the companion article — that the food industry creates the metabolic disruption and the pharmaceutical industry manages the symptoms — acquires a further dimension here. The drug that manages the symptoms is now large enough to disrupt the industry that created the need for it. This is, in the history of industrial capitalism, a fairly unusual situation.

The Scale of the GLP-1 Market

To understand the macroeconomic consequences of GLP-1 adoption, the starting point is the scale of the pharmaceutical market itself and the speed at which it is expanding.

Novo Nordisk’s semaglutide products — Ozempic for type 2 diabetes management and Wegovy for chronic weight management — generated combined revenues of approximately 232 billion Danish kroner (roughly £17 billion) in 2023, making semaglutide the fastest-growing pharmaceutical product in history by revenue ramp. Eli Lilly’s tirzepatide products (Mounjaro and Zepbound) are on a comparable trajectory. By early 2025, Novo Nordisk had briefly overtaken LVMH to become the most valuable company in Europe by market capitalisation — a pharmaceutical company whose primary product is a weekly injection to suppress appetite outvaluing the entire luxury goods sector of the continent.

Analyst projections for the GLP-1 market by 2030 range from 130 billion US dollars to over 200 billion US dollars annually, depending on assumptions about pricing, insurance coverage, treatment persistence, and the approval of oral formulations that would substantially expand the addressable population. These projections assume continued strong adoption in the United States, expanding uptake in Europe as insurance coverage broadens, and emerging market penetration in the Gulf states and East Asia where obesity prevalence and private healthcare spending are both rising.

Market context: In the United States as of mid-2025, approximately 6 percent of the adult population had been prescribed a GLP-1 medication. Persistence rates — the proportion of patients still on the medication after twelve months — are approximately 40 to 50 percent at current pricing, reflecting a significant dropout rate driven by cost, side effects, and the return of appetite and weight upon discontinuation. If oral formulations with improved tolerability and lower manufacturing costs reach market at accessible price points, the persistence rate and addressable population both expand materially. At 10 percent adoption with 60 percent persistence, the caloric impact on the US food market becomes a structural rather than marginal consideration.

The global pharmaceutical supply chain consequence of this market is itself significant. Novo Nordisk has committed tens of billions of euros to expanding manufacturing capacity for semaglutide — constructing new facilities in Denmark, the United States, and France, creating skilled manufacturing employment and driving demand for the specialised equipment and precursor chemicals used in peptide synthesis. The pharmaceutical supply chain is a net beneficiary of GLP-1 growth. The food supply chain is not.

What GLP-1 Users Actually Eat Less Of

The appetite suppression produced by GLP-1 receptor agonists is not uniform across food categories, and the non-uniformity is commercially significant.

GLP-1 medications reduce appetite most dramatically for high-calorie, high-fat, high-sugar foods — precisely the categories that constitute the core revenue of the fast food and ultra-processed food industries. This is not coincidental. The GLP-1 hormone is released naturally in response to eating, as part of the satiety signalling system. Its pharmaceutical mimicry replicates and amplifies the signal that ultra-processed food engineering was specifically designed to override. When the override is removed, the preference pattern shifts toward lower-calorie, higher-protein, higher-fibre foods — and away from the caloric density that makes fast food economically efficient to produce and margin-accretive to sell.

Survey data from GLP-1 users consistently shows reduced consumption of fast food, carbonated soft drinks, alcohol, confectionery, and snack foods, with relative maintenance or increase in consumption of lean protein, vegetables, and whole foods. The pattern is sufficiently consistent that it has been named in the industry as the GLP-1 dietary shift — and its commercial implications for specific food categories are not subtle.

Research context: A 2024 analysis published in Obesity by Almandoz et al. surveyed GLP-1 medication users on dietary changes and found statistically significant reductions in consumption of fast food (reported by 67% of respondents), sugary beverages (63%), alcohol (50%), and processed snacks (71%), with concurrent increases in vegetable (48%) and lean protein (52%) consumption. The directional consistency of these findings across multiple surveys is more significant than the precise percentages, which vary by population and methodology.

The soft drink category warrants specific attention. Carbonated soft drink consumption is one of the highest-margin products in the food and beverage supply chain, and its revenue concentration in a small number of dominant brands — Coca-Cola, PepsiCo, and their bottling networks — makes the sector unusually exposed to demand shifts that are consistent across a large population. Both companies have acknowledged GLP-1 as a demand consideration in investor communications. The mechanism is direct: GLP-1 reduces the palatability of very sweet beverages for many users, and the habitual consumption pattern that drives soft drink volumes is precisely the kind of behaviourally reinforced consumption that appetite normalisation disrupts.

Alcohol is a less obvious but analytically important category. Multiple observational studies and a growing number of clinical reports suggest that GLP-1 medications reduce alcohol consumption and craving — a finding consistent with the mechanism of GLP-1 receptors in the reward circuitry, which processes both food and alcohol reinforcement through overlapping pathways. If this effect is confirmed at scale, the implications extend beyond food into the alcoholic beverage industry, hospitality, and the public health economics of alcohol-related harm.

Fast Food: The Demand Shock

The fast food industry’s business model rests on a set of assumptions about consumer behaviour that the GLP-1 shift is beginning to challenge at the margin — and that, at scale, would challenge fundamentally.

Fast food economics depend on frequency of visit, average transaction value, and the specific product mix within each transaction. The highest-margin items in a fast food transaction are typically the drink and the side — the Coca-Cola and the fries — rather than the burger or sandwich that drives the visit. A customer whose appetite is suppressed, whose preference for sweet and fatty foods has been reduced, and whose portion size has decreased may still visit the same outlet but will, on average, spend less per visit and will purchase fewer of the high-margin accompaniments that account for a disproportionate share of the outlet’s profit.

McDonald’s disclosed in its 2024 annual report that it was monitoring GLP-1 adoption as a potential demand headwind. The company’s chief financial officer specifically referenced the possibility of reduced frequency and ticket size among GLP-1 users as a consideration for medium-term revenue planning — an unusual degree of candour from a company that has historically been reluctant to acknowledge external threats to its demand model.

The geographic concentration of fast food consumption compounds the sectoral exposure. The United States, where GLP-1 adoption is highest and where fast food penetration is also highest, is the market in which the interaction of these two trends is most acute. If GLP-1 adoption reaches 10 percent of the US adult population — a trajectory consistent with current prescription growth and projected oral formulation launches — the affected population represents a meaningful fraction of the core fast food customer base.

Revenue exposure modelling: Goldman Sachs research published in 2024 estimated that at 10 percent US adult GLP-1 adoption, the combined effect of reduced visit frequency and lower average transaction values could reduce same-store sales at major US quick service restaurant chains by 3 to 5 percent annually — a figure that, while not catastrophic in isolation, represents a structural headwind rather than a cyclical one, and one that compounds over time as adoption rates continue to grow. For an industry operating on typical EBITDA margins of 15 to 20 percent, a persistent 3 to 5 percent top-line reduction has significant implications for franchise economics and expansion capital.

The franchise model that dominates the fast food industry creates a specific vulnerability. The major fast food brands — McDonald’s, Yum! Brands, Restaurant Brands International — are primarily franchise operators rather than direct restaurant operators. Their revenue is largely royalty-based, insulating the parent company from the direct operating costs but creating a pass-through exposure: if franchise unit economics deteriorate, franchisees reduce capital investment, close underperforming units, and in stressed cases exit the system entirely. The franchise network, which represents the physical infrastructure of the fast food supply chain, is more exposed to a sustained demand reduction than the parent company’s headline financials suggest.

The Agricultural Supply Chain: Working Backwards

The consequences of reduced fast food demand do not stop at the restaurant counter. They propagate backwards through a supply chain of considerable complexity and geographic reach, affecting commodity markets, processing industries, packaging manufacturers, and the agricultural producers at the upstream end of the chain.

Beef is the highest-profile commodity exposure. The United States fast food industry is the largest single buyer of ground beef in the country, and its purchasing patterns set price signals that affect cattle markets nationally. A sustained reduction in fast food beef demand — driven by reduced burger consumption among GLP-1 users — would put downward pressure on feeder cattle prices, with consequences for the ranching sector and the meatpacking industry that processes beef for the quick service restaurant channel.

The potato industry has a specific and concentrated exposure. The frozen french fry supply chain — dominated by a small number of large processors including McCain Foods, Lamb Weston, and Simplot — is directly dependent on quick service restaurant volumes. French fry consumption is one of the most consistently reported reductions in GLP-1 user surveys. Lamb Weston, whose revenues are approximately 70 percent derived from the restaurant sector, cited GLP-1 adoption as a demand uncertainty in its investor communications in 2024 and announced production capacity reductions at several processing facilities.

The corn and soy complex — the foundational commodity layer of the ultra-processed food supply chain — faces a more diffuse but potentially larger exposure. High-fructose corn syrup, refined corn starch, soybean oil, and the corn-derived inputs to animal feed are embedded throughout the fast food and food manufacturing supply chains. A structural reduction in ultra-processed food consumption, whether driven by GLP-1 adoption or by regulatory changes to food marketing and agricultural subsidy, would reduce demand for these commodities through multiple channels simultaneously.

Supply chain concentration risk: The global french fry supply chain illustrates the concentration vulnerability of agricultural commodity processing. The top four frozen potato processors control approximately 80 percent of the North American market. Their capital-intensive processing facilities — each representing hundreds of millions of dollars of fixed investment — are configured for the specific potato varieties and volume throughputs of the quick service restaurant channel. Reconfiguring that capacity for different products or different markets is not a short-term option. A sustained 10 to 15 percent reduction in restaurant french fry volumes would leave significant processing capacity stranded.

The packaging industry faces analogous pressures. Fast food packaging — cups, wrappers, containers, bags — is a high-volume, thin-margin business whose economics depend on the consistent throughput volumes of the restaurant chains it serves. Reduced transaction counts reduce packaging demand in a sector where the customer concentration is high and the ability to diversify rapidly is limited.

Commodity agriculture, at the furthest upstream point of the supply chain, is the sector with both the largest potential exposure and the longest adjustment cycle. Agricultural investment decisions — in land, in crop varieties, in irrigation and storage infrastructure — are made on horizons of years to decades. The price signals from a GLP-1-driven reduction in fast food demand will reach commodity markets gradually, but the structural nature of the demand shift — unlike a recession, which is cyclical — means that the adjustment will need to be permanent rather than temporary. Farmers and agricultural investors who are positioned for a food system whose demand profile is shifting are facing a genuine strategic challenge that the current price signals do not yet fully reflect.

Food Manufacturing: The Reformulation Imperative

The large packaged food companies — Nestlé, Unilever, Kraft Heinz, Conagra, General Mills, Kellogg’s — face a more complex strategic challenge than the fast food chains, because their exposure is more diversified and their options for response are more varied.

These companies have been aware of the long-term trend toward healthier eating for at least two decades, and most of them have made acquisitions in the natural and organic food space as portfolio hedges against the gradual erosion of their legacy ultra-processed product lines. What the GLP-1 adoption curve adds to this pre-existing trend is speed and specificity. GLP-1 users are not gradually drifting toward healthier products. They are making abrupt changes to high-value, high-frequency purchase categories — snacks, confectionery, soft drinks, frozen convenience foods — that represent the core profit pools of the major food manufacturers.

The reformulation response — reducing sugar, fat, and caloric density in existing products while maintaining palatability — is the least structurally disruptive adaptation available. Several major manufacturers have accelerated reformulation programmes specifically citing GLP-1 as a demand consideration. The challenge is that reformulation is expensive, technically difficult, and frequently produces products that consumers prefer to the original in health terms but not in taste terms — which is precisely the appeal dimension that drives repeat purchase.

The more structural response is portfolio reorientation toward the product categories that GLP-1 users are increasing: lean protein, high-fibre foods, vegetables, whole grains. This is a longer-term strategic shift that requires investment in different supply chains, different processing capabilities, and different distribution relationships. It is also the direction in which the broader secular trend in consumer preferences — independent of GLP-1 — has been pointing for some time.

The Kraft Heinz strategic situation illustrates the exposure of highly leveraged food companies to a structural demand shift. The company carries approximately 19 billion US dollars of debt from the 2015 merger, secured against the future cash flows of brands — Heinz ketchup, Kraft macaroni and cheese, Oscar Mayer processed meats — that are among the most concentrated in the ultra-processed categories most affected by GLP-1 dietary shifts. A sustained reduction in these products’ volumes would impair the debt service capacity of a company that has limited financial flexibility to respond. Kraft Heinz is not unique in this position; it is an extreme example of a dynamic that affects the leveraged balance sheets of multiple major food manufacturers.

Labour Market Consequences

The food service sector is one of the largest employers in most developed economies, and its labour market exposure to a structural reduction in fast food volumes deserves explicit analysis.

In the United States, the restaurant industry employs approximately 15 million people — roughly 10 percent of the private sector workforce. The quick service restaurant segment, which is most directly exposed to GLP-1 demand effects, accounts for the largest share of this employment. A sustained reduction in fast food volumes that results in unit closures, reduced operating hours, or headcount reductions in individual outlets would have meaningful labour market consequences concentrated in the demographics — young workers, lower-income workers, workers in suburban and exurban markets — who are most represented in quick service restaurant employment.

The food manufacturing sector employs a further several million workers in the United States alone, and the upstream agricultural and processing industries add millions more globally. The labour market consequences of a structural shift in food demand are therefore not confined to the visible workforce of the restaurant counter. They extend through the processing plant worker, the delivery driver, the packaging operative, and ultimately the farmhand and agricultural contractor at the end of the chain.

The geographic concentration of this exposure matters. Fast food employment and food processing employment are both concentrated in specific labour markets — the suburban commercial strips where quick service restaurant clusters are located, and the rural communities where food processing plants are frequently the dominant employer. A structural reduction in these industries’ demand would not be distributed evenly across the economy. It would be concentrated in specific places, affecting specific communities, in ways that have political and social consequences beyond the economic ones.

The automation dynamic interacts with GLP-1 exposure in a specific way. The quick service restaurant industry was already investing heavily in automation — AI-driven ordering kiosks, robotic food preparation, automated delivery — as a response to rising minimum wages and labour scarcity. A reduction in demand volumes that makes existing labour investment less productive accelerates the economic case for automation. The workers most at risk from a GLP-1-driven demand reduction are therefore also the workers most at risk from the automation investment that the reduction incentivises. The two trends compound rather than offset each other.

Government Revenue and Healthcare Fiscal Dynamics

The fiscal consequences of GLP-1 adoption at scale operate on two channels simultaneously: the revenue side, through reduced tax receipts from the food and beverage industries affected by demand shifts, and the expenditure side, through potentially significant reductions in healthcare costs associated with obesity, type 2 diabetes, cardiovascular disease, and their comorbidities.

The revenue reduction channel is relatively straightforward to model, though the precise numbers are uncertain. Value-added tax and sales tax on food and beverage products, corporation tax on the profits of fast food chains and food manufacturers, payroll taxes on the workforce of the food service sector, and the agricultural subsidy expenditure that supports commodity production — all of these are affected by a structural reduction in ultra-processed food consumption. The aggregate fiscal exposure of governments to the revenue side of the GLP-1 disruption is non-trivial, though it is partially offset by increased pharmaceutical revenues and the employment and tax contributions of the pharmaceutical sector’s expansion.

The expenditure reduction channel is potentially much larger and much more significant for long-term fiscal sustainability. The healthcare cost of obesity, type 2 diabetes, and their associated conditions in the United States is estimated at approximately 1.7 trillion US dollars annually, representing a very significant fraction of total US healthcare expenditure. If GLP-1 medications substantially reduce the prevalence and severity of these conditions across the treated population, the downstream healthcare cost reduction would be one of the largest fiscal improvements in modern public finance history.

Healthcare fiscal modelling: A 2024 analysis by IQVIA estimated that if GLP-1 adoption reached 15 percent of the eligible US adult population over ten years, the cumulative reduction in obesity-related healthcare costs would exceed 200 billion US dollars over the modelling period, offset by approximately 150 billion US dollars in GLP-1 medication costs, producing a net fiscal positive. The model is sensitive to assumptions about treatment persistence, comorbidity prevention rates, and long-term price trends for GLP-1 medications. It also does not account for the healthcare system capacity freed by reduced obesity-related admissions, which itself has significant economic value.

The tension between these two fiscal channels — reduced food industry revenues in the near term, reduced healthcare costs over the medium term — creates a policy environment in which governments have a financial interest in both restraining GLP-1 adoption (to protect near-term tax revenues) and accelerating it (to reduce long-term healthcare liabilities). The way this tension resolves in different countries will depend on the structure of their healthcare systems, the composition of their tax base, and the political economy of the food and pharmaceutical industries within their jurisdictions.

For the United Kingdom, where the NHS bears the full cost of obesity-related healthcare and where GLP-1 prescribing through the NHS is capacity-constrained by manufacturing supply and budget allocation, the fiscal calculus is particularly interesting. The NHS is simultaneously the entity that would benefit most from a large-scale reduction in obesity prevalence and the entity whose current budget is under the most pressure from the cost of treating obesity’s consequences. The GLP-1 investment decision for the NHS is, in effect, a large-scale public health infrastructure decision dressed in the form of a pharmaceutical prescribing question.

The Food Industry’s Strategic Response

The food industry’s response to GLP-1 disruption is visible in investment decisions, acquisition activity, product development pipelines, and the messaging directed at both investors and regulators, and it broadly follows the pattern of a mature industry confronting a structural demand shift that it cannot prevent but can adapt to.

The acquisition of protein-rich food brands — a trend visible across multiple major food conglomerates — reflects the alignment of portfolio strategy with the specific dietary shift of GLP-1 users. Lean protein is the category most consistently increased by GLP-1 users, and the brands positioned in this space — premium canned fish, high-protein dairy, legume-based products — are being acquired or developed at valuations that reflect their positioning in the post-GLP-1 demand landscape.

Portion size reduction is being pursued in some categories as both a nutritional response and a cost management response. Smaller portions at the same or similar price points improve margin per unit sold while aligning with the reduced appetite capacity of GLP-1 users. The product development language around this shift — ‘right-sized portions,’ ‘snack formats,’ ‘mindful eating’ — reflects the marketing challenge of positioning a reduction in quantity as a benefit rather than a diminution.

The engagement with GLP-1 pharmaceutical companies directly, as commercial partners rather than competitive threats, is an emerging strategic posture that several major food companies are exploring. The logic is that GLP-1 users require specific nutritional support — high protein, adequate micronutrients, foods that manage the gastrointestinal side effects of the medication — and that food companies positioned as the nutritional partners of the GLP-1 journey have a commercial opportunity that is not available to those who remain associated with the dietary patterns the medication is designed to reduce.

Strategic speculation: The possibility of a convergence between the pharmaceutical and food industries — in which food products are developed and marketed in specific clinical partnership with GLP-1 therapies — is not yet an established market but is being actively explored. The commercial logic is coherent: a patient on a GLP-1 medication who is losing weight needs to maintain muscle mass and bone density, which requires specific protein and micronutrient intake; a food company that positions its products as the clinically recommended nutritional support for GLP-1 therapy has a defensible market position that is not replicable by competitors without the same clinical partnership. Whether this represents a new commercial model or an extension of the existing pharmaceutical-food industry relationship described in the companion article is a question worth watching.

Emerging Markets: The Second Wave

The GLP-1 disruption described in this article is, to date, primarily a US and to a lesser extent European phenomenon. The manufacturing constraints, pricing levels, and insurance coverage requirements that limit GLP-1 access in high-income countries are multiplied in emerging markets. But the trajectory is clear, and the second wave of GLP-1 adoption — in the Gulf states, in East Asia, in Latin America, and eventually in South and Southeast Asia — will interact with the food system disruptions of those markets in ways that the current analysis only partially captures.

The Gulf states are the most immediately relevant emerging market exposure. Obesity prevalence in Saudi Arabia, Kuwait, and the UAE is among the highest in the world, wealthy populations with private healthcare access are already pursuing GLP-1 medications through private channels, and the governments of these countries have explicit national health strategies that identify obesity reduction as a priority. The fast food penetration of Gulf markets is also very high — the drive-through fast food culture of these countries closely mirrors the US market that GLP-1 is already affecting, and the demographic profile of Gulf fast food consumers overlaps substantially with the profile of the population most likely to seek GLP-1 medication.

China is the most significant long-term market consideration. With over 250 million adults with obesity and the world’s largest and fastest-growing pharmaceutical market, China’s GLP-1 adoption trajectory — constrained currently by domestic manufacturing capacity and regulatory approval timelines — will eventually represent a demand shift for the global food supply chain comparable in scale to China’s entry into commodity markets as a buyer over the past thirty years. Chinese domestic fast food consumption, and the agricultural supply chains that serve it, face a structural demand headwind that is currently in its earliest stages.

The GLP-1 revolution is not a pharmaceutical story. It is an economic disruption of the food system that produced the conditions for the pharmaceutical intervention. The industries whose products created the obesity epidemic are now facing the macroeconomic consequences of its pharmaceutical management. The irony is precise and the disruption is real.

The Long View

The macroeconomic impact of the GLP-1 boom on the global fast food and food manufacturing supply chain is, at this stage, in its early chapters. The current evidence — reduced footfall signals, commodity demand adjustments, manufacturing capacity reductions, investor communications acknowledging demand headwinds — represents the leading indicators of a structural shift whose full dimensions will only be visible over the next five to ten years as adoption rates, treatment persistence, and oral formulation accessibility evolve.

The direction of the shift is not in doubt. The industries most exposed — quick service restaurants, carbonated soft drink manufacturers, processed snack companies, and the agricultural commodity processors that supply them — are facing a structural reduction in the demand for their highest-margin products, driven by a pharmaceutical intervention whose adoption is accelerating and whose underlying demand is determined by the prevalence of the metabolic conditions those products helped create.

The industries best positioned to benefit — lean protein producers, high-fibre food manufacturers, the pharmaceutical companies themselves, and the healthcare systems that will eventually capture the fiscal dividend of reduced obesity-related expenditure — are entering a period of structural demand growth whose scale the current market is still underestimating.

The workers and communities most exposed to the adjustment — in the quick service restaurant sector, in food processing plants, in commodity agriculture — are the least financially equipped to manage the transition and the furthest from the policy conversations in which its consequences are being discussed.

The regulatory and policy framework that will govern how the adjustment proceeds — who bears its costs, how rapidly it occurs, what investment is directed toward the communities most affected, and whether the fiscal dividend of reduced healthcare costs is captured and deployed to cushion the transition — does not yet exist in any coherent form.

The GLP-1 boom has introduced into the global food system an exogenous shock whose origins are pharmaceutical but whose consequences are agricultural, commercial, fiscal, and social simultaneously. It is, in the precise sense of the term, a disruption: an intervention from outside the system that is changing the rules of the system faster than the system’s institutions can adapt.

Whether it produces a food system that is better for human health depends not on the drug itself — the drug is a symptom manager, not a system reformer — but on whether the disruption it creates is used by those with the standing to do so as an opportunity to address the structural conditions that made a drug of this kind necessary in the first place.

The evidence, as this series has repeatedly found, suggests that the institutions with the standing to make that choice are also the institutions with the strongest incentives not to.

THE LONG VIEW — MACROECONOMICS, FOOD & PHARMACEUTICAL DISRUPTION

Companion to: The Obesity Conundrum

Market data and projections referenced in this article are drawn from publicly available analyst research and company disclosures. They represent directional estimates and should not be relied upon as investment advice. The Long View does not hold positions in any of the companies or sectors discussed.

Read more

You have two compulsory years of education ahead of you. After that, the path is yours to choose. The guide below is honest about what is coming, practical about what you can do right now, and short enough to actually read. Each piece of advice comes with the reason behind it, because being told what to do without understanding why is exactly the kind of education this guide is trying to fix.

The world you are entering is different from the one your parents entered, different from the one your teachers were trained for, and different from the one that most career advice was written for. AI is already doing much of the work that graduates used to do. Companies built around AI are beginning to replace companies that use AI as a tool. The degree that was a reliable ticket to a good job twenty years ago is no longer automatically that. None of this is a reason to panic. It is a reason to prepare differently.

Section 1: The Next Two Years

While you are still in compulsory education

1. Pass your exams — but do not confuse passing exams with being educated.

You still need qualifications. A-levels, BTECs, T-levels, or their equivalents open doors and close off fewer options. Get them. But understand that what you remember in five years will not be the content you revised — it will be the skills you used to think, write, and work things out. Do both: meet the requirement and use the time to develop something real alongside it.

2. Pick at least one subject you find genuinely interesting and go deeper than the syllabus.

The exam syllabus is the floor, not the ceiling. If you are studying history, read a book about a period not on the syllabus. If you are studying biology, watch lectures online about whatever aspect genuinely interests you. Going beyond the curriculum is how you discover what you are actually capable of — and it is how you develop the ability to learn independently, which is the single most valuable thing you can do with the next two years.

3. Develop one practical skill that produces something visible.

Code a small project. Build something. Design something. Write something and publish it. Make a short video. The specific skill matters less than the fact that you completed something real, outside of school, with no one telling you exactly what to do. This is the beginning of your portfolio — evidence of capability that is yours, not the school’s, and that does not disappear when you lose the exam certificate.

4. Learn to manage money at a basic level now.

Understand what a budget is. Know how tax works in outline. Know the difference between an asset and a debt. Know what compound interest does to borrowing. These are not taught reliably in school and the cost of not knowing them lands on you personally within a few years of leaving. An hour with a parent or a reliable online resource on each of these topics is a better investment than most of the revision you will do this year.

5. Learn to cook six meals properly from ingredients.

This sounds trivial. It is not. The ability to feed yourself well on a limited budget is practical independence. It is also the entry point to understanding nutrition, planning, and the difference between what you need and what is being sold to you. Start simple. Cook it, eat it, repeat.

Section 2: Your Mind

The thing you will carry into every room you ever enter

6. Learn to ask why — and then ask it again.

The single most valuable intellectual habit you can develop is the refusal to accept a conclusion without understanding the reason behind it. Why is this the answer? What evidence supports it? What would change it? This is called critical thinking and it is not taught as a subject anywhere on the standard curriculum. You develop it by practising it — by not accepting the first answer, including yours.

7. Get comfortable being wrong.

The people who learn fastest are not the people who are right most often. They are the people who are wrong and notice it and update. The school system rewards being right and punishes being wrong, which trains you to avoid being wrong rather than to use it. In the real world, the person who can say I was wrong, here is why, and here is what I think now — this person is trusted and valuable. Practice it. It takes years to undo the school habit.

8. Know the difference between what you know, what you think, and what you have been told.

In an age when AI can produce a confident, well-written, plausible answer to almost any question, the most important cognitive skill is the ability to evaluate whether what you are reading is reliable. Ask: what is the source? What is their interest in this conclusion? What would the other side say? You do not need to be suspicious of everything — you need to be honest about what you actually know versus what you are assuming.

9. Develop a tolerance for genuine uncertainty.

Most real decisions have to be made without enough information. The person who freezes until they have certainty, or who grabs the nearest confident-sounding answer to escape the discomfort of not knowing, is not equipped for complex situations. The person who can hold I do not yet know without anxiety, and keep working toward a reasonable position, is much more useful — and much less manipulable.

You are going to spend your life making decisions with incomplete information. The capacity to do that well is not taught anywhere. You have to build it yourself, deliberately, starting now.

Section 3: Understanding AI — Not Just Using It

The thing that is changing everything and that most people misunderstand

10. Use AI as a thinking partner, not an answer machine.

AI tools can give you an answer to almost any question in seconds. If you use them only as answer machines, you outsource your thinking and your capability shrinks. If you use them as thinking partners — to challenge your reasoning, to show you angles you have not considered, to help you refine a question — your capability grows. The difference is whether you engage with the output or just consume it. Engage with it.

11. Understand roughly how AI works, not just what it does.

You do not need a computer science degree. You need to know that AI systems produce outputs based on patterns in their training data, that they produce confident-sounding answers regardless of whether the underlying question has a definitive answer, and that they fail in specific and predictable ways. A person who understands this can use AI well and know when not to trust it. A person who does not understand this gets confidently wrong answers and does not notice.

12. Build skills that AI cannot yet replicate.

AI is very good at pattern recognition, information synthesis, and producing fluent, structured text. It is poor at genuine novelty, at complex social judgment, at knowing which question to ask in a genuinely ambiguous situation, and at taking responsibility for a decision. The skills most resistant to AI substitution in the next ten years are judgment under uncertainty, interpersonal intelligence, creative problem-framing, and the ability to evaluate AI output critically. These are the skills worth investing in.

Section 4: Building Something Real

The evidence that will matter more than your grades

13. Start building a portfolio before you leave school.

A portfolio is a collection of things you have actually done — projects, work, writing, designs, code, whatever is relevant to what you want to do. The companies that are growing fastest are already hiring on portfolio evidence rather than degree certificates. The people who arrive at eighteen or twenty-one with a portfolio of real work have a significant advantage over those who arrive with qualifications alone. You do not need it to be perfect. You need it to be real and to show what you can do.

14. Do something outside school that other people can see.

A blog, a YouTube channel, an open-source project, a local business, a community initiative — anything that is visible, that you built, and that other people have engaged with. This is not about becoming an influencer. It is about creating evidence that you can initiate something, sustain it, and produce an output without being told exactly what to do. In a hiring process, this evidence is worth more than most qualifications.

15. Treat every project as practice for the next one, not as an end in itself.

Your first project will not be good. That is fine. Its purpose is not to be impressive — its purpose is to teach you what the second project needs. The people who produce good work in their twenties are almost always people who produced mediocre work in their teens and kept going. The willingness to start, to finish imperfectly, and to learn from it is more valuable than waiting until you are confident enough to do it well.

Section 5: People

The part of the future that AI cannot automate

16. Learn to communicate clearly in writing and in person.

The ability to explain a complex idea simply, to write an email that is understood the first time, to speak in a room of people without losing the thread — these are capabilities that the workplace needs and that school does not reliably develop. The way to develop them is practice: write more, speak more, ask for feedback, notice when you are not understood and work out why. There is no shortcut.

17. Build relationships with adults who are doing things you find interesting.

Most people get their first opportunities not through job applications but through people they know. Not in a manipulative sense — in the sense that adults who see a young person who is curious, capable, and reliable will remember them when an opportunity arises. Find the adults in your life and your community whose work interests you. Ask them about it. Show genuine interest. Offer to help with something concrete. This is not networking in the corporate sense. It is relationship-building, and it is one of the most reliable paths from where you are to where you want to be.

18. Be the person who does what they say they will do.

Reliability is rarer than talent. The person who says they will do something and does it — on time, to the agreed standard, without needing to be chased — is someone that others want to work with and will go out of their way to help. This sounds simple. In practice it requires discipline, honesty about what you can actually commit to, and the courage to say I cannot do that rather than agreeing and then disappointing. Start developing this habit now. It compounds over years into one of the most valuable professional reputations you can have.

Section 6: The Decision at 18

University, apprenticeship, work — how to think about it

19. Do not go to university because you do not know what else to do.

University is the right choice in specific circumstances: you want to enter a profession that requires a degree (medicine, law, engineering), you have a genuine intellectual interest in a subject that you want to pursue at depth, or you need the specific network and experience that a particular institution provides. It is not the right choice simply because it is the default. Three years and significant debt to avoid a decision is an expensive form of delay, and the labour market for non-vocational graduates from lower-ranking institutions is tighter than the universities’ marketing suggests.

20. Consider degree apprenticeships seriously.

A degree apprenticeship gives you a degree, a salary, no tuition debt, and three to four years of real work experience in an organisation. The employers offering them — in accountancy, technology, engineering, law, and finance — are doing so because they found that people who combine study with real work are more capable than graduates who have spent three years in full-time education. For a significant proportion of people in a significant proportion of fields, this is the better route. It is undersubscribed relative to its value because schools and parents are less familiar with it.

21. If you go to university, use it for more than the degree.

The degree is the minimum requirement for attending university. The valuable things are: the people you meet, the ideas you encounter outside your subject, the organisations you join, the projects you build, the experience of living independently and solving problems without anyone handing you the answer. The graduate who leaves with a 2:1 and nothing else has underused the three years significantly. The graduate who leaves with a 2:1, a portfolio of real projects, relationships across multiple fields, and a clear sense of what they can do is a different proposition entirely.

22. Learn a skill that earns money alongside everything else.

Whether or not you go to university, developing a freelance or part-time income from a practical skill — writing, coding, design, tutoring, video editing, photography — is valuable for three reasons. It earns money. It builds portfolio evidence. And it teaches you what it actually takes to exchange value with another person, which is a form of education that no classroom has yet replicated. You do not need to build a business. You need to understand what it means to be paid for something you can do.

Section 7: The Honest Bit

What nobody tells you

23. The world does not owe you a job — but you are not owed a bad one either.

The labour market is genuinely uncertain. AI is displacing roles that your parents’ generation found reliable. Degrees are worth less than they were. The competition for good roles is real and it is global. None of this means you cannot build a good career — it means building one requires more intentionality than it used to. The people who will do well are not the people who were told the right thing to study. They are the people who developed genuine capabilities, built evidence of those capabilities, and kept adapting.

24. Know what you want before you ask for it.

AI can give you any information you ask for. The difficult question is what to ask. A significant number of people get the career, the salary, the lifestyle they thought they wanted and discover it does not feel the way they expected. The capacity to know yourself — to understand what actually gives you energy and what drains it, what kind of work produces the feeling of a day well spent, what you are willing to sacrifice and what you are not — is not developed by answering exam questions. It is developed by paying attention to your own experience and being honest about what it tells you.

25. Find something you can be genuinely good at and go as deep as you can.

The person with genuine mastery in one area — who can do something at a high level that other people find difficult — is always employable. The person with shallow familiarity across many areas is much more vulnerable to substitution, including by AI. You do not need to have decided what that thing is at sixteen. You need to be looking for it, actively and honestly, and to keep going deeper when you find something that holds your attention and rewards your effort.

The world is changing faster than any institution can keep up with. That is actually good news for you — because it means the people who figure out how to navigate it themselves, rather than waiting for the system to tell them what to do, have a significant advantage. You are one of those people. Start now.

This article attempts that in a deliberately unheated tone. It sets aside the question of fault. It treats the warming trend as a measured fact rather than a moral argument, and it ends with the only lever most of us actually control: our own household choices.

1. The weather in front of us

With only a few weeks of the season left, summer 2026 was on course to become the UK’s hottest on record. Through to mid-August the country was running around 1.9°C above its long-term average, with England some 2.5°C above and already logging its warmest June on record. It became the first year ever to reach 35°C in May, June and July; by mid-July it had recorded more days above 30°C than the whole of the famous summer of 1976; and by mid-August a fifth heatwave had carried the temperature to 38.1°C at Kew Gardens in London — the UK’s fifth-hottest day in records going back to the nineteenth century. Two separate days reached or exceeded 38°C, equalling the record set in 2022.

More days above 30°C by mid-July 2026 than during the whole summer of 1976

This was not an isolated British event. Western Europe had its hottest June on record — more than 3°C above the 1991–2020 baseline — during what was, globally, the second-warmest June ever measured. The UK’s summer, in other words, is one corner of a pattern stretching across the continent.

A single hot summer, however, proves very little on its own. Hot summers have always happened. The question that matters is whether they are becoming more common — and there the record is clearer than any one heatwave could ever be.

2. One hot year, or a warming line?

Here it helps to borrow an idea from statistics: regression to the mean. Any extreme reading — a blistering July, a freakishly mild January — tends to be followed by something closer to average, simply because extremes are, by definition, unusual. If the climate were stable, a record-breaking summer would most likely be followed by cooler ones, and the long-run average would sit still. Records would be broken only occasionally, and roughly as often at the cold end as the warm end.

That is not what the UK data shows. The average itself is moving. 2025 became the UK’s warmest year on record at 10.09°C — only the second year ever above 10°C — and the tenth warmest year in a row to fall within the last two decades. All ten of the UK’s warmest years on record have now occurred since 2005. When records fall almost only in one direction, and the baseline keeps stepping up, we are no longer watching noise regress toward a fixed mean. We are watching the mean itself climb.

10.09°C the UK’s warmest year on record (2025) — all ten warmest have come since 2005

The Met Office puts the underlying rate at roughly 0.25°C per decade since the 1980s. An attribution study estimated that human-influenced climate change made 2025’s record roughly 260 times more likely — and that temperatures at that level, once genuinely rare, should now be expected about once every three years. In other words, what used to be an outlier is becoming the ordinary. This is the sense in which the phrase “the new normal” is not rhetoric but arithmetic.

There is a wider regional pattern too. Europe is the fastest-warming continent, heating at close to twice the global average; in the fifty years since 1976 the continent as a whole has warmed by around 2°C. The UK’s experience is a local instance of something happening across the whole of Europe.

3. Hotter, drier summers — and the missing snow

Many people’s impression that summers feel hotter and drier, while winters seem to have lost much of their snow, is broadly consistent with the measurements — which is worth stating plainly, because personal recollection and instrument records do not always agree. On this, they do.

The dry side of the heat

Summer 2026 was not only hot but noticeably dry: by mid-season the UK had seen only about 42% of its average seasonal rainfall, with England and Wales down at 34% and parts of eastern England at just 26%. This followed a spring 2025 that was England’s driest in over a century. Warmth and dryness tend to reinforce one another — dry ground heats faster, and hot air draws more moisture from soils and reservoirs — which is why “hotter and drier” so often arrive together.

Warmer winters, vanishing snow

The winter signal is just as real, if quieter. Recent decades’ winters have been both milder and wetter; the winter half-year in 2016–2025 was about 13% wetter than in 1961–1990, with more of the season’s precipitation falling as rain rather than snow. Snow-lying days have fallen by up to five days per decade since 1979 across much of the UK, and by as much as ten in the higher North Pennines. Roughly a fifth of the country has seen a marked drop in days with snow on the ground. The disappearance of reliable snow from many lowland areas is one of the most tangible fingerprints of the trend.

The mechanism is simple enough to state without drama: cold snaps still happen, but they now sit on top of a warmer background, so fewer of them are cold enough, for long enough, to bring widespread lying snow. The extremes have not vanished; the baseline beneath them has risen.

4. Can it be reversed, or must we adapt?

This is the question that most divides people, and it deserves a straight answer: realistically, both are true at once.

The warming already in the system is unlikely to reverse on the timescale relevant to household and infrastructure planning. Carbon dioxide already emitted persists for a very long time, and the climate system — the oceans above all — responds only slowly, over decades and longer, so temperatures continue to adjust well after emissions occur. A good deal of what we are now seeing — the hotter summers, the milder winters, the retreat of reliable snow — is best treated, for planning purposes, as the ground we now stand on rather than a phase to wait out. Adaptation is not defeatism here; it is prudence. Homes, gardens, water use and daily routines built around the climate of the twentieth century are increasingly built for a place that no longer exists.

At the same time, the future is not fixed. The difference between a world that warms a little more and one that warms a great deal is enormous, and that difference depends on how much greenhouse gas is emitted from here on. Reducing emissions does not undo the past, but it does bend the line — it decides whether summers like 2026 become the typical case or the mild end of a much harsher range. So the honest position is neither “we can switch it back” nor “nothing can be done,” but rather: adapt to what is already locked in, and reduce emissions to limit how much more gets locked in.

5. What does this mean for ordinary life?

Statistics about decadal averages can feel remote until they arrive at the front door. A warmer Britain is, in practice, a set of small, concrete shifts. Bedrooms that once cooled overnight now stay uncomfortable until morning, and sustained warm nights — not just hot afternoons — are what most strain older bodies. Gardens face longer dry spells, hosepipe restrictions become more routine, and the plants and lawns that suited a cooler, wetter country begin to struggle. Homes designed to hold heat in now need, increasingly, to keep it out. Wetter winters push water toward drains and floodplains rather than snow onto hills, which is steadily reshaping flood risk and, with it, the cost and availability of insurance. Even the economics of the home are turning: the long-standing British assumption that homes need heating but never cooling no longer entirely holds. None of this is catastrophic, but taken together it changes what a sensible household plans for — which is where the practical response begins.

6. What one household can reasonably do

For the individual, the more useful question is what can reasonably be done at household level. Two kinds of action are worth separating: adapting to the heat that is coming, and trimming the emissions that determine how much more heat arrives.

Adapting to hotter summers and milder winters

● Keep heat out before trying to cool: external shading, closing curtains and windows by day, and opening up at night make a large difference to how a home copes with a heatwave, at little cost.

● Treat water as the scarce summer resource it is becoming — water butts, drought-tolerant planting, and mulched beds hold up far better through dry spells than thirsty lawns.

● Plan for the health side of heat: know which rooms stay coolest, keep a cool space for sleeping, and check on elderly relatives and neighbours during prolonged hot spells.

● Expect milder, wetter winters rather than snowy ones — which shifts the sensible investments toward drainage, damp-proofing and flood resilience more than toward cold-weather kit.

Reducing your own contribution to warming

Analyses of personal climate choices tend to agree on which levers are large and which are small. The largest, for most households, cluster around transport, home energy and diet; recycling and similar habits, while worthwhile, are comparatively minor. Roughly in order of impact:

● Reduce car and air travel: living with less car dependence is consistently found to be the single most impactful personal choice, and cutting back on flights — especially long-haul ones — is close behind. A single avoided long flight can outweigh a year of smaller efforts.

● Decarbonise the home: switching to a renewable electricity tariff or generating your own, improving insulation, and — when a boiler reaches the end of its life — moving to a heat pump are among the most durable reductions available to a household.

● Shift the diet: moving toward more plant-based meals is one of the higher-impact food choices — by one estimate several times more effective than cutting food waste and far more than composting, though every step helps.

● Buy less and keep it longer: the emissions embedded in manufacturing mean that repairing, buying second-hand and simply replacing things less often quietly avoids a good deal of carbon.

It is worth being honest about proportion. One household’s emissions are a vanishingly small share of the global total, and no personal choice will move the thermometer on its own. But the case for most of these measures does not rest on saving the world. A home that is shaded and well insulated, a garden planted for drier summers, water used sparingly, durable things bought once rather than often — these leave the household more comfortable, more resilient and frequently better off, whatever governments or other countries choose to do. That is the quiet strength of adaptation: sensible preparation need not wait on everybody agreeing about everything.

A measured conclusion

The data tells a coherent and undramatic story. The UK and Europe are warming; the warming is steady rather than random; summers are becoming hotter and drier, winters milder and less snowy, and the extremes of recent years are better read as the consequences of a rising average rather than simply flukes destined to regress away. Much of that change is unlikely to reverse within the planning horizon of an ordinary household, and calls for level-headed adaptation. How much further it goes, though, is still unwritten. Neither panic nor denial is warranted. Attention, and a few sensible adjustments, are.

References

Figures for summer 2026 reflect Met Office data published up to 13 August 2026 (checked 16 August 2026). The season closes on 31 August; refresh against the Met Office’s end-of-season summary before republication.

Met Office, Summer 2026 so far: a look at the season at the halfway point (blog, 2026).

Met Office, Warmest UK summer on record “increasingly likely” as temperatures stay well above average (blog, 11 Aug 2026).

Met Office, Latest heatwave peaks with 38.1°C recorded and rain on the horizon in the south (13 Aug 2026).

Met Office, 2025 is a double record-breaker: UK’s warmest and sunniest year on record (2026).

Met Office, Climate extremes are becoming the new ‘normal’ for the UK — State of the UK Climate (2026).

Copernicus Climate Change Service, Record heatwave brings hottest June for western Europe during second-warmest June globally (2026).

Copernicus / European State of the Climate, Why is Europe warming so quickly? — Europe warming at nearly twice the global average.

WMO / UNRIC, Europe Heatwave 2026: WMO warns of rising extreme heat.

The Conversation, Why snow days are becoming increasingly rare in the UK.

World Resources Institute, 19 Ways to Help the Climate, Ranked.

For most of human history, paternity was a matter of trust, inference, and the private arithmetic of conception. A child was presumed to belong to the man who was present at its birth, who had raised it, who had provided for it — and the biological truth beneath that presumption was, in the overwhelming majority of cases, unknowable. This unknowability was, in its way, a form of social stability. Families were built on commitment as much as biology, and the two were, by practical necessity, allowed to coexist without interrogation.

A cardboard box containing a saliva collection tube, a prepaid envelope, and a QR code has changed this. The consumer genetic testing industry — 23andMe, AncestryDNA, MyHeritage, and dozens of competitors — has placed in the hands of ordinary individuals a biological instrument whose implications for family structure were not considered in any policy process and were not anticipated by most of the people who used it.

The testing was sold as ancestry. What it revealed, with clinical precision and no prior warning, was parentage.

This article examines what the data says about non-paternity rates, what genetic testing has done to the discovery and consequences of that data, how those consequences distribute asymmetrically across men, women, and children, what the biology of human attraction and mate selection tells us about why this pattern exists at all, and what a society that has suddenly acquired the capacity to know what it previously could not might do with that knowledge wisely rather than destructively.

The Numbers

The question of how frequently children are raised by men who are not their biological fathers has been studied with varying methodologies and has produced a range of estimates whose imprecision reflects the difficulty of the measurement rather than the irrelevance of the question.

The most rigorous large-scale study of non-paternity rates in the modern era is the analysis by Larmuseau et al. published in Current Biology in 2016, which examined patrilineal Y-chromosome transmission across seventeen generations of Belgian family records. The study found an extra-pair paternity rate — children whose biological father was not the man recorded as their father — of approximately one to two percent per generation in the historical population, rising to approximately four percent in periods of social disruption and economic stress.

Studies using direct paternity testing data, rather than historical genetic inference, tend to produce higher estimates. Research reviewing paternity testing laboratory data in the UK, the United States, and Australia consistently finds non-paternity confirmation rates of between 10 and 30 percent among men who submit samples for testing — but this population is not representative, because men who submit for paternity testing have typically already formed a suspicion. The tested population is not a random sample of fathers.

The most methodologically careful population-level estimates, adjusting for the selection bias in voluntary testing data and drawing on genealogical databases from multiple countries, converge on a figure in the range of three to eight percent for Western populations — meaning that somewhere between one in thirty and one in twelve children in the modern Western world are being raised by a man who is not their biological father, without that man’s knowledge.

Research context: A 2021 meta-analysis by Anderson (Perspectives on Psychological Science) examined 32 studies across multiple methodologies and populations and found mean non-paternity rates of 3.3 percent in low-suspicion populations and 29.8 percent in high-suspicion populations, confirming the selection bias effect. The population-level figure most consistently cited in peer-reviewed literature for contemporary Western societies is in the range of 2 to 4 percent, though some studies of specific demographic groups — younger mothers, lower socioeconomic status populations, populations with higher rates of relationship instability — produce figures considerably above this range. At 3 percent, applied to a UK population of approximately 12 million dependent children, the number of children whose paternity status may differ from their registered father’s is not trivial.

What consumer genetic testing has changed is not the underlying rate — these patterns were present long before the testing existed. What it has changed is the visibility. Prior to affordable direct-to-consumer DNA testing, a man who raised a child had essentially no practical mechanism to confirm paternity without commissioning a specialist test that carried significant social cost — the act of seeking it was itself an accusation. Consumer ancestry testing, by contrast, is socially normative, family-positive in its marketing, and produces biological information as a by-product of genealogical curiosity. The discovery that follows is rarely sought. It is stumbled upon.

How Discoveries Are Made

The mechanism by which genetic testing reveals non-paternity is most commonly not a direct paternity test but the triangulation of DNA matches across shared databases. A man who purchases an ancestry test and uploads his results may find that the genetic relatives appearing in his match list do not include the cousins, aunts, uncles, and grandparents he expects on his father’s side — while an entirely unfamiliar set of matches appears, consistently, on a genetic family tree he has no social connection to.

The discovery is frequently made not by the man himself but by the child. A woman or man in their thirties or forties, curious about ancestry, uploads their DNA and finds that their genetic father’s family is not the family they grew up in. The social father may be dead. The biological father may be unknown or difficult to locate. The mother may have carried this knowledge privately for decades, or may have genuinely not known which relationship produced the pregnancy.

The clinical and counselling literature on these discoveries — termed NPEs, Non-Paternity Events, or more recently Misattributed Parentage Experiences — has expanded rapidly alongside the testing industry. Support communities such as the DNA Detectives group and the NPE Friends Fellowship document thousands of cases annually and report consistent patterns: the discovery is almost universally destabilising regardless of the age at which it occurs, the grief is multilayered and does not follow the recognised stages of bereavement in predictable ways, and the institutional support available to people navigating these discoveries is minimal.

The language of the field is worth noting. The term MPE — Misattributed Parentage Experience — was developed by support organisations specifically to avoid the loaded moral connotations of ‘non-paternity event,’ which frames the situation from the perspective of the man whose paternity is in question. The shift reflects the reality that the experience is most frequently and most devastatingly a child’s experience rather than a father’s — a discovery about one’s own origins, not only about another person’s behaviour.

The use of genetic testing evidence in divorce and family law proceedings has created a second pathway of disclosure that is specifically adversarial rather than coincidental. A man initiating or responding to divorce proceedings may commission a paternity test on a child conceived during the marriage, the results of which, if they confirm non-paternity, become evidence in proceedings that affect financial settlement, maintenance obligations, and the emotional context in which the family dissolution occurs. This use of genetic evidence is legal in most jurisdictions, is increasingly common, and produces disclosures in a setting that is already characterised by conflict and whose consequences for the child are entirely unmanaged by the legal process.

The asymmetry of this disclosure mechanism deserves explicit acknowledgement. Genetic testing can reveal maternal infidelity with certainty — the child is either the genetic offspring of the man testing or it is not. It cannot reveal paternal infidelity, because the man’s genetic material does not appear in the child’s DNA in a way that confirms or excludes his fidelity, only his parentage. A woman who suspects her partner has fathered a child by another woman cannot resolve that suspicion through a DNA test of children within her own family. The evidentiary asymmetry is complete: genetic testing is a tool that exposes one form of infidelity and is blind to the other.

The Effect on Children

The child is the third party in every non-paternity disclosure, and in most cases the least considered one. The legal frameworks governing paternity testing, the counselling infrastructure available to families navigating these discoveries, and the social scripts for managing the aftermath are all oriented primarily toward the adults involved. The child’s experience — which is qualitatively different from both the man’s and the woman’s — is addressed, when it is addressed at all, as a secondary consequence of adult decisions.

The developmental and clinical literature on children who discover their parentage has been significantly disrupted is consistent on several points. The age at which discovery occurs does not reliably predict the severity of disruption — discoveries in adulthood can be as destabilising as discoveries in childhood, and the compound grief of discovering that one’s entire family narrative has a different structure than believed is not proportionate to the practical changes the discovery requires. The loss of paternal identity — the sense of biological connection to a lineage, a heritage, a set of traits and predispositions — is experienced as a real loss even when the social relationship with the non-biological father remains intact.

When non-paternity is disclosed in the context of divorce proceedings, the child is in the worst possible position to absorb it. The social father’s relationship with the child may already be under strain. The legal process that follows — the potential revision of maintenance obligations, the father’s possible emotional withdrawal, the mother’s defensive positioning — all occur in a period when the child’s attachment security is already disrupted. The child is not a party to the proceedings. No one is formally required to consider how the disclosure is communicated to them, when, or in what context.

The legal position varies significantly by jurisdiction. In the UK, a man who has acted as a child’s father — who is registered as the father, who has lived with the child, who has contributed to its upbringing — retains parental responsibility regardless of biological paternity, and maintenance obligations are not automatically extinguished by a non-paternity finding. In other jurisdictions, the position is less settled, and there have been cases in which non-paternity findings have been used to argue for the reduction or elimination of financial obligations to children whose only fault is the biology they were born into.

The child support consequences of non-paternity findings are among the most practically damaging outcomes of disclosure for children, because they affect the economic stability of the household in which the child lives at a time when that household is already under stress. A child whose social father reduces or eliminates financial contributions following a non-paternity finding may face simultaneous disruption to their paternal attachment relationship, their family structure, their economic circumstances, and their identity narrative — four separate dimensions of security destabilised simultaneously.

The Genetics of Attraction: Why This Happens

Any honest examination of non-paternity patterns requires engagement with the biology of human mate selection and infidelity — not as a moral framework, but as an explanatory one. The evolutionary biology of human reproductive strategy has produced a substantial and rigorous scientific literature that illuminates why the patterns genetic testing is now revealing are not aberrations or moral failures but the expression of biological drives that operated in an environment very different from the one modern humans inhabit.

The human mating system is neither strictly monogamous nor strictly polygamous. It is, in the terminology of evolutionary biology, a mixed strategy — one in which the dominant social pattern is pair bonding with significant investment by both parents, and in which extra-pair mating occurs at rates that are not trivial. This mixed strategy is not unique to humans. It is found, with variations, across most species that combine significant parental investment with complex social structures.

The specific genetic mechanism that the research on human extra-pair attraction has focused most extensively on is the Major Histocompatibility Complex — the MHC, a set of genes that plays a central role in immune function and whose expression is detectable, at least partially, through body odour. The research of Claus Wedekind and colleagues at the University of Bern, published in the 1990s and subsequently replicated and extended across multiple populations, demonstrated that women tend to find the body odour of men with dissimilar MHC profiles more attractive than men with similar ones.

Research context: The MHC and mate choice literature has developed substantially since Wedekind’s original sweaty T-shirt studies. A 2022 meta-analysis by Havlicek and Roberts examined 34 studies and found consistent directional evidence for MHC-dissimilar mate preference, with the effect strongest for natural odour assessment and weakest in visual assessment contexts. The evolutionary logic is immunological: offspring of MHC-dissimilar parents have broader immune repertoires and greater pathogen resistance. The mechanism operates largely below conscious awareness — people describe finding a partner’s scent attractive without being able to articulate why, and typically without any knowledge of MHC genetics.

The MHC research has a specific and troubling implication for pair bonding and infidelity. Hormonal contraception — the oral contraceptive pill and its equivalents — substantially alters the MHC preference pattern. Women on hormonal contraception tend to prefer MHC-similar scent profiles, the reverse of the unmedicated preference. Women who begin a relationship while on hormonal contraception and later discontinue it — commonly when they intend to conceive — may find that their partner’s attractiveness, as registered through scent, has changed. The mechanism is biological, the effect is real, and neither partner has done anything to produce it.

A second evolutionary mechanism relevant to understanding extra-pair mating patterns is the Dual Mating Strategy hypothesis, developed by David Buss, Randy Thornhill, and colleagues across a body of research spanning three decades. The hypothesis proposes that women’s mate preferences are not fixed but shift across the menstrual cycle — that during the fertile phase of the cycle, preferences shift toward men with higher indicators of genetic quality (facial symmetry, testosterone markers, immune diversity indicated by MHC dissimilarity) and away from the indicators of investment capacity and long-term reliability that are preferred across most of the cycle.

The hypothesis generates significant debate within evolutionary psychology — its application to human behaviour is contested and the evidence base, while substantial, is not without methodological criticism. But the directional finding — that short-term and long-term mate preferences in women are not identical, and that the shift between them is hormonally mediated rather than consciously chosen — is among the more robust findings in the relevant literature.

The biology does not excuse behaviour. It contextualises it. Human beings are not the straightforwardly monogamous creatures that their social institutions assume. The gap between the institution and the biology is not a modern development. It is as old as the institution.

The cuckoldry literature — the evolutionary and historical study of men who unwittingly raise children fathered by other men — is extensive and pre-dates genetic testing by decades. Studies of birds, which share the pair-bonding with extra-pair mating pattern with humans, established the mechanisms and their frequency before the equivalent human data was accessible. The term itself, borrowed from the cuckoo’s brood parasitism, is ancient — its presence in Chaucer and Shakespeare reflects that the phenomenon was culturally recognised long before it was biologically measurable.

Historical and anthropological context: Evolutionary anthropologist Robin Baker’s work Sperm Wars (1996) synthesised evidence from multiple sources — sperm morphology studies, paternity assessments, behavioural research — to argue that human reproductive competition extends beyond visible mate competition to include physiological adaptations to sperm competition. The work is controversial and some of its specific claims have been challenged, but the broader evidence for extra-pair mating as a feature of human reproductive strategy rather than an exception to it has strengthened in subsequent decades, not weakened.

The Asymmetry of Discovery and Consequence

The asymmetry that genetic testing has introduced into the landscape of infidelity disclosure is not merely evidentiary. It is deeply structural, and its consequences distribute very differently across the parties involved.

For the man who discovers non-paternity, the experience is typically a compound loss: the loss of the paternal relationship as he understood it, the loss of trust in a partner and in a relationship history, and sometimes the loss of financial certainty if legal proceedings follow. The psychological literature on men’s responses to non-paternity discovery documents high rates of acute distress, relationship dissolution, and in some cases significant mental health consequences. These are real harms and they deserve acknowledgement.

For the woman, the consequences of discovery vary enormously depending on circumstances. A woman who has known about the non-paternity throughout the child’s life is in a different position from a woman for whom the discovery is mutual — who genuinely did not know, or who made a private decision decades ago in circumstances that have since changed beyond recognition. The moral calculus is not uniform, and the legal and social consequences apply to a range of situations that the blunt instrument of a DNA result does not distinguish.

The woman who deliberately maintained a false paternity through a long relationship faces genuine moral accountability for a deception that has affected multiple lives. The woman who had a brief relationship at a time of instability, who made a private decision in a state of fear or uncertainty, who has raised the child within a subsequent relationship and built a family that functions — her situation is different, and the consequence of disclosure in that circumstance can be the destruction of a household that was, by every measure that matters for the child, working.

The woman who is not the primary focus of these discoveries at all is the one whose child commissions an ancestry test out of curiosity, decades after her death, and whose private history is retrospectively exposed without her capacity to provide context, explanation, or apology. This is now common. The dead have no privacy in the age of consumer genetics.

The single most significant asymmetry, however, is the one already noted: genetic testing reveals maternal infidelity with certainty. It cannot reveal paternal infidelity. A man who has fathered children in other relationships — which is a form of infidelity with equivalent moral weight — is not exposed by any genetic test conducted within his primary family. The technology illuminates one form of betrayal and is entirely blind to the other. If society were designing an infidelity detection system from first principles, it would not design this one. But it did not design it. It bought it in a cardboard box marketed as a Christmas gift.

Was This Always Happening?

The question of whether the non-paternity rates genetic testing is revealing represent a new pattern or an ancient one is important, because the answer shapes what conclusions it is reasonable to draw from the data.

The historical genetic evidence — from the Larmuseau patrilineal study and from comparable work in other countries — suggests that extra-pair paternity at rates of two to four percent per generation has been a consistent feature of human family structure across centuries, predating modern contraception, modern feminism, modern working patterns, and all the other contemporary conditions to which infidelity is sometimes attributed. The rates observed today are not meaningfully higher than the historical baseline. They are simply, for the first time, visible.

This matters enormously for how we interpret what we are seeing. If the non-paternity rates revealed by genetic testing reflected a new social pathology — a breakdown of commitment and family values unique to contemporary life — then the appropriate response would be a social and cultural one, addressing the conditions that produced the change. The evidence does not support this interpretation. What the evidence supports is that human beings have always had a reproductive biology that does not map cleanly onto the institutions of monogamous pair bonding, and that those institutions have, for most of history, been maintained by the absence of the means to detect the gap.

The gap between the institution and the biology is not a failure of individuals. It is a feature of the species. The institution of monogamous pair bonding was not designed by biologists. It was designed by social systems whose primary objectives were the stable transmission of property, the maintenance of clear lines of inheritance, the management of paternal investment, and the organisation of childcare responsibilities in a way that survived the absence of one parent. These are valuable social objectives. They do not require that the biology perfectly comply with them to be worth maintaining.

The concept of cuckolding — in its evolutionary rather than its contemporary cultural sense — has been present in human awareness for as long as records exist. The anxiety about paternity certainty is documented in legal codes from ancient Mesopotamia, in Roman law, in the common law of England. The social and legal responses to this anxiety — the regulation of women’s movement and sexuality, the mechanisms of public shame for infidelity — were, in significant part, mechanisms for managing the gap between monogamous institution and human biology in the absence of the biological tools to close it. Genetic testing has provided those tools. The social frameworks for using them wisely do not yet exist.

What Society Now Faces

The widespread availability of consumer genetic testing has placed a tool of profound social consequence in the hands of individuals who did not ask for it, were not warned about it, and have no institutional support for managing its consequences. The companies that sell these tests bear some responsibility for this — the informed consent processes for consumer ancestry testing do not adequately prepare purchasers for the possibility of discovering information about family structure that they did not seek and may not be equipped to manage.

The scale of the impending disclosure wave is not trivial. As consumer genetic databases grow — 23andMe alone has processed over 14 million tests, Ancestry’s database exceeds 20 million — the probability that a non-paternity situation within a family will be discovered by one of its members, even without anyone actively seeking the information, increases year on year. The database does not need to contain the man’s own test. It needs to contain a sufficient density of the biological father’s relatives to allow triangulation. In large databases, that threshold is already crossed for much of the Western population.

The legal framework has not kept pace. The admissibility of DNA evidence in family proceedings, the obligations of testing companies to flag potential non-paternity discoveries in their results communications, the rights of children to know their biological parentage, the obligations of parents to disclose — these questions are being resolved piecemeal, inconsistently across jurisdictions, and almost always reactively rather than by design.

The counselling infrastructure is embryonic. The NPE support community is primarily peer-led rather than professionally resourced. General practitioners, who are often the first point of contact for people navigating these discoveries, are not trained in the specific psychological profile of misattributed parentage experience, which is distinct from bereavement, from relationship breakdown, and from other forms of identity disruption that do have established therapeutic frameworks.

The policy question that no jurisdiction has adequately addressed is whether the consumer genetic testing industry should be required to provide pre-test counselling that explicitly addresses the possibility of discovering non-paternity or unexpected parentage, comparable to the genetic counselling required before clinical genetic testing for heritable conditions. The testing companies have resisted this requirement on commercial grounds — the friction of a counselling requirement would reduce uptake. The public health case for it is, on the evidence of the disclosed NPE population, substantial.

Frailty, Not Failure

This article has been written without blame because the evidence does not support a blaming framework. What the genetics of attraction, the evolutionary biology of human mating strategy, and the historical non-paternity data together suggest is not that the people at the centre of these discoveries are uniquely or exceptionally flawed. They are human.

They are human in the sense that their choices were made under biological pressures that operate largely below conscious awareness, in social circumstances that were specific and sometimes extreme, at moments in their lives when the distance between aspiration and action was, as it is for all of us at some points, larger than they would wish. The woman who had an affair twenty years ago and built a family in the decades that followed is not reducible to that choice. The man who raised a child he did not know was not biologically his is not a fool. The child caught in the middle of their story did not choose it and does not deserve its consequences.

The appropriate social response to what genetic testing is revealing is not a moral campaign against infidelity — rates which, as the historical data suggests, would be unaffected by such a campaign, because they pre-date every moral campaign that has previously been mounted. It is the construction of the support structures, legal frameworks, and cultural narratives that allow individuals and families to navigate these discoveries with the minimum of additional harm.

It is also, perhaps, the beginning of a more honest cultural conversation about the distance between the institutions we have built for managing human relationships and the biology those relationships are built on. That conversation does not require abandoning monogamy, or excusing betrayal, or reducing human beings to their evolutionary drives. It requires acknowledging that the drives exist, that the institutions do not perfectly contain them, and that the gap between the two has always been managed — by privacy, by discretion, by the social agreement not to look too closely — and is now, for the first time, being illuminated by technology that has no interest in maintaining the agreement.

The truth in the saliva is not a verdict. It is data. What we do with it — how we build the frameworks that allow it to be known without destroying what it reveals — is the question that the science has posed and that only wisdom can answer.

The wisdom required is not technological. It is the older kind: the capacity to hold complexity without resolving it too quickly, to understand that human beings are simultaneously capable of profound love and biological complexity, and that families are built not only on genetics but on the accumulated weight of ordinary days — the meals, the school runs, the illnesses sat through, the arguments resolved, the particular history of two people and the children between them that belongs to no laboratory and cannot be measured by any test.

The technology will not stop. The discoveries will continue to accumulate. The question is whether the society receiving them will respond with the punitive reflexes that the adversarial legal system encourages, or whether it will find the more difficult and more humane path: building frameworks that acknowledge what is now known, protect the children who bear the consequences of what adults have done, and extend to all the parties involved the recognition that they are, each of them, doing their best with the biology and the circumstances they were given.

THE LONG VIEW — GENETICS, FAMILY & HUMAN NATURE

Scientific references in this article are drawn from peer-reviewed literature. The article is not intended as legal or genetic counselling advice. Individuals who have been affected by non-paternity discoveries are encouraged to seek support from qualified genetic counsellors and the NPE peer support community.

There is a particular quality to the memory of a summer afternoon spent lying on grass, watching clouds move. No agenda. No schedule. No screen. Just the slow, unhurried business of being a child in the world — learning, without knowing you were learning, what it felt like to be bored and then unbored, to invent and then abandon, to exist without being monitored or managed or measured.

That quality is not nostalgia. It is a description of something specific and developmentally important that is disappearing from the lives of children in ways that the research is beginning to document and that common observation has been recording for some time.

The question this article asks is not simply what has been lost. It is who permitted the loss, under what pressures it occurred, and who — if anyone — is now responsible for protecting what remains.

The Geography of Childhood Past

A child growing up in Britain in the 1970s or 1980s had, by contemporary standards, an extraordinary degree of unsupervised freedom. They left the house after breakfast on a Saturday and returned for tea, and in the intervening hours they negotiated the social and physical world largely without adult intervention. They fell off bicycles and got back on. They argued over rules they invented themselves for games that had no referee. They found places — a particular tree, a stretch of canal bank, a gap in a fence — that belonged to them and to their friends and to no one else.

The developmental literature is clear about what was happening in those hours. Children were building the executive function, emotional regulation, conflict resolution, and risk assessment capacities that come specifically from unstructured, self-directed play. They were learning to tolerate frustration without an adult mediating it. They were learning to be bored, and to generate their own solutions to boredom. They were building, incrementally and through accumulated small experiences, the architecture of a self.

None of this was understood in those terms at the time. It did not need to be. It happened because the conditions for it to happen existed: sufficient physical space, sufficient parental trust in the environment, sufficient absence of competing demands on children’s time, and a social norm that regarded a certain amount of unsupervised roaming as ordinary and healthy rather than negligent.

Childhood was not perfect then. But it had, in the unscheduled afternoon, something that the current version has largely surrendered: time that belonged to the child, not to the adult agenda.

The Architecture of Fear

The transformation did not happen overnight, and it did not happen for a single reason. But the primary driver of the change in children’s physical freedom is not difficult to identify: parental fear, amplified by a media environment with a structural interest in maintaining it.

The objective data on child safety in public spaces in the United Kingdom does not support the level of fear that now governs children’s mobility. Rates of stranger abduction — the specific risk that animates most parental anxiety about unsupervised outdoor play — are not significantly higher than they were forty years ago. They were low then. They remain low now. The crimes that do occur are horrific, and their coverage is exhaustive and repeated across every available platform. The coverage is not proportionate to the statistical risk. It is proportionate to the audience engagement the stories generate.

The result is a population of parents who have made a rational calculation based on a distorted information environment. They are not irrational to be afraid. They are responding accurately to the signals they have been given. The signals are wrong. But the fear is real, and its consequences are measurable: children who do not walk to school alone, who do not play in parks without supervision, who do not develop the navigational and social independence that previous generations acquired as a matter of course.

Into that vacuum of unsupervised time, the smartphone and the tablet arrived with extraordinary precision. A child who is not allowed outside is a child who needs to be somewhere. A screen that provides unlimited stimulation, social connection, and entertainment is a solution to a problem that fear created. The technology did not manufacture the anxiety. It colonised the space the anxiety emptied.

The monitoring technology now deployed by parents — location sharing on phones, GPS-enabled watches marketed to families, tracking applications that show a child’s movements in real time — represents the logical endpoint of a trajectory that began with a reasonable desire to know that a child is safe. What it produces is a generation of children who have never experienced the particular freedom of being, for an afternoon, genuinely unobserved. The developmental consequences of that absence are not yet fully documented. They are unlikely to be trivial.

The Scheduled Life

For children who are not permitted to use unstructured time independently, that time must be filled by adults. The consequence is the scheduled childhood — the Monday football, the Tuesday piano, the Wednesday drama club, the Thursday swimming lesson, the Friday coding workshop — in which every hour of a child’s week outside school is allocated to an organised activity with an adult in charge.

The motivations are mixed and, individually, not unreasonable. Some parents genuinely want to give their children access to experiences and skills they did not have themselves. Some are responding to a competitive educational environment in which extracurricular achievement has become part of the credentialling system. Some are covering the practical reality that they cannot be at home during the after-school hours and an organised activity is safer than an empty house. Some, if they are honest, are filling their children’s time because they do not know what else to do with it.

The aggregate effect, whatever the individual motivation, is a childhood in which the experience of boredom — of having nothing organised, nothing required, no adult supervising the next transition — has become rare. This matters more than it might appear. Boredom is not an absence. It is, developmentally, an active state. The child who is bored is the child who is being forced to generate their own stimulus, their own narrative, their own engagement with the world. The research on creativity, self-regulation, and intrinsic motivation consistently identifies unstructured time as the environment in which these capacities develop.

A generation that has not been bored in the developmental sense — that has moved from one organised activity to the next, with screen time filling the gaps — is a generation that has been systematically deprived of the conditions in which certain fundamental capacities are built. The academic literature on declining attention spans, rising anxiety, and difficulty with self-direction among young people is extensive. The connection to the disappearance of unstructured time is underexplored in the public conversation.

The child on the schedule is not experiencing childhood. They are experiencing a junior version of adult productivity culture — with all its demands and none of its compensations.

The Holiday That Is Not a Holiday

The school year in England runs for approximately thirty-nine weeks. It was designed around an agricultural calendar that has not been relevant for over a century, and it has not been seriously reformed since. For a workforce in which the majority of parents are employed full-time — a workforce that the economic changes of the past four decades have both produced and required — it creates a structural impossibility.

No employer offers thirteen weeks of annual leave. The gap between the school holiday calendar and the working year is covered by a combination of annual leave consumed entirely on childcare, grandparental availability where it exists, holiday clubs, paid childcare, and an informal economy of mutual arrangement among parents that has become, in many households, a significant logistical undertaking.

The consequence for children is that the unstructured downtime that holidays once represented — the long, formless weeks of summer in which children self-organised, grew bored, invented things, and gradually learned to occupy themselves — has been replaced by a structured programme that mirrors the school term in its density of organised activity, if not in its academic content. The child at holiday club is supervised, scheduled, and entertained. They are not, in any meaningful sense, on holiday.

The child who is not at holiday club is, in many cases, at home with a screen. The parent who is working cannot supervise. The alternative to organised activity is not the free-range summer of forty years ago. It is eight hours in front of a device with no one in the house. These are not equivalent options, and the choice between them is not a free one.

The structural misalignment between the school calendar and the working year is a policy failure of considerable vintage that no government has found it expedient to address. The costs of that failure are distributed across individual households in ways that are largely invisible to the institutions that produced it.

The Screen That Never Sleeps

A child who is two years old and has learned to navigate a touchscreen is not demonstrating advanced development. They are demonstrating that the human infant brain, shaped by millions of years of evolution to learn from its environment with extraordinary efficiency, will learn whatever the environment most consistently presents to it.

The tablet and the smartphone are not neutral educational tools. They are delivery mechanisms for content optimised by some of the most sophisticated behavioural engineering ever applied to a consumer product, designed by teams of psychologists and product designers to maximise engagement — which in practice means maximising the difficulty of stopping. The algorithms that govern what a child sees next on YouTube or a social platform are not designed with any consideration for developmental appropriateness. They are designed to keep the session going.

The replacement of the television — which was, for all its limitations, a finite and scheduled medium — by the on-demand platform represents a qualitative shift that has not been adequately absorbed into the public understanding of what children are now exposed to. A generation ago, a child watching television was watching a programme that ended, after which something else came on that was not necessarily relevant or appealing. The natural break existed. The on-demand platform has no natural break. The next piece of content begins before the current one has finished. The architecture is explicitly designed to eliminate the moment of decision.

The question of what children are watching on those platforms is separate from the question of how much they are watching, and both are serious. The pathway from age-appropriate content to inappropriate content on major platforms is short and, for a curious child, not difficult to follow. The research on children’s accidental and deliberate exposure to pornographic and violent content online is consistent: it is widespread, it begins younger than most parents believe, and its effects on the developing understanding of relationships, bodies, consent, and emotional intimacy are not benign.

A child of eleven or twelve who has had sustained exposure to online pornography has not had their innocence gently extended. They have had it replaced with a set of images and narratives about human sexuality that bear no relationship to the reality of adult intimacy, and that will require, if addressed at all, a considerable amount of careful unlearning. In many cases it is not addressed. The parents do not know. The child does not say. The images persist.

The child knows what adults do in private. They do not know how adults feel. They have access to the most explicit content human sexuality has ever produced, and no framework within which to place it. They are, in the truest sense of the word, lost.

Knowing Everything and Understanding Nothing

There is a particular type of child visible in schools and in social settings today who presents as precociously mature. They use adult vocabulary. They reference adult cultural products — films, music, online figures — that their parents may not have encountered. They discuss topics that, a generation ago, would not have entered a child’s conversational register until considerably later. They appear, on the surface, to be growing up faster.

The appearance is misleading. What has accelerated is not emotional or psychological maturity. It is exposure. The child has access, through the device in their hand, to a volume and range of information about the adult world that no previous generation of children has had. But access to information is not the same as the capacity to process, contextualise, or emotionally integrate it. The developmental sequence cannot be bypassed by providing the content ahead of schedule.

What results is a dissonance that the educational and psychological literature is increasingly documenting: children who know, in cognitive terms, far more than their years suggest is appropriate, and who are emotionally and psychologically still children, with all the fragility and developmental needs that entails. They are fluent in the language of adult complexity without having the internal resources to manage it. The anxiety, the emotional dysregulation, the difficulty with genuine intimacy that characterises so much of the current generation of young people is, in part, the consequence of this dissonance.

The adults in their lives — parents, teachers, the broader social environment — have largely failed to provide the interpretive framework that would allow the information to be contextualised. That failure is not always deliberate. It is partly the product of adults who are themselves overwhelmed, who were not given that framework by their own parents, and who do not have confident access to the language needed for conversations that the current environment makes urgent.

There is a cruelty in the current situation that deserves to be named. The child is not protected from the content of the adult world. They are denied the structures — stable households, consistent adult presence, unhurried time — that would allow them to encounter that content with some resilience. They receive the exposure without the preparation. That is not an accident of technology. It is the consequence of a series of social and economic choices made by adults, for adult reasons, whose costs are being carried by children.

Who Is Responsible?

The temptation, in any analysis of what has happened to childhood, is to assign blame cleanly. To the technology companies, whose products are designed without reference to child development and whose age restrictions are unenforceable and largely unenforced. To the media, whose coverage of child safety risks bears no relationship to their statistical distribution and has produced a generation of parents incapacitated by fear. To employers, whose working patterns treat the school calendar as someone else’s problem. To governments, whose childcare and education policies have consistently prioritised economic participation over the conditions necessary for healthy child development.

All of these are legitimate targets. None of them is sufficient as a single explanation.

The technology companies have built products that they know are harmful to children, and have responded to the evidence with a combination of token age verification measures and lobbying against regulatory intervention. This is documented. It is not a matter of interpretation. The internal research at several major platforms has demonstrated awareness of harm to young users, and the public position of those platforms has not reflected that awareness. This is a failure of corporate ethics of a serious and specific kind.

The media environment has maintained a level of coverage of child safety threats that serves its commercial interest in audience retention rather than the public interest in accurate risk assessment. The parents who will not allow their children to walk to school are not stupid. They are responding to a real emotional signal generated by a distorted information environment. The organisations generating that information bear responsibility for the consequences.

Employers and the economic model they operate within have made the dual-income household the financial norm in ways that have transferred the practical problem of childcare entirely to individual families, without corresponding investment in the infrastructure — affordable childcare, flexible working arrangements, a school calendar aligned with working life — that would allow it to be managed without the costs being borne by children.

Governments have known about these dynamics for decades and have consistently chosen short-term economic metrics over the long-term investment in childhood that the evidence suggests is necessary. The political return on early years investment is too distant for the electoral cycle to capture. The cost of not making it will be carried by a generation that cannot yet vote.

But the question that sits beneath all of this is more uncomfortable than any institutional critique. It is the question of what adults — individually, collectively, as a culture — have decided that childhood is for.

What Childhood Is For

The scheduled childhood, the monitored childhood, the screen-occupied childhood — these are not simply failures of policy or corporate ethics. They reflect a set of assumptions about what childhood is and what it is supposed to produce that have shifted, quietly and largely without examination, over the past forty years.

The older assumption — still operative within living memory — was that childhood was a protected period of development in which a human being was allowed to grow at the pace that growth required, in conditions of relative safety and relative freedom, acquiring through experience rather than instruction the capacities that adult life would demand. The role of adults was to provide the conditions, not to fill the time.

The current assumption, rarely stated but visible in the structure of children’s lives, is that childhood is a period of preparation — for academic achievement, for competitive entry to the next stage of education, for a labour market that will reward the child who arrives with the most credentials, the broadest skill set, and the longest head start. The role of adults is to ensure that no developmental opportunity is missed, which in practice means ensuring that no time is unscheduled.

These two assumptions produce radically different childhoods. The first produces adults who can tolerate uncertainty, manage boredom, regulate their own emotions, and generate their own motivation. The second produces adults who are highly competent at executing tasks set by others, deeply uncomfortable with unstructured time, and vulnerable to the removal of the external scaffolding that has always told them what to do next.

The mental health data for young adults in Western economies — the anxiety, the difficulty with independent decision-making, the reliance on pharmaceutical support for basic emotional regulation — is consistent with the second model having been widely applied. This is not a coincidence.

The question of who is protecting children is, in the end, a question about which model adults have chosen to operate with. The protection that children need is not primarily from physical danger — the risk of which, as noted, has not increased in the way that fear suggests. The protection they need is from the adult world’s tendency to colonise their time, compress their development, and expose them prematurely to content and complexity they do not have the internal resources to manage.

That protection requires something more than policy and regulation, though both are necessary. It requires a cultural shift in what adults believe childhood is for — a willingness to defend unscheduled time as a developmental necessity rather than a gap to be filled, to accept that a child watching clouds from a patch of grass is not wasting time but doing some of the most important work of their life.

It requires, in short, the recovery of a kind of patience with childhood that the current economic and cultural environment has made very difficult to sustain.

Whether it will happen is another question. But the cost of it not happening is already visible, in the anxiety of a generation that grew up knowing everything and understanding very little, that reached adulthood technically capable and emotionally unprepared, that was never, in any real sense, allowed to be a child.

THE LONG VIEW — CHILDHOOD & SOCIETY

This article presents analytical commentary on social trends affecting child development. References to research are directional; readers seeking primary sources are encouraged to consult the academic literature

In 1975, approximately 3 percent of the world’s adult population was clinically obese. By 2024, the World Health Organisation placed that figure at over 16 percent — more than one billion people. Overweight prevalence, which includes those not yet at the clinical obesity threshold, now accounts for over 40 percent of the global adult population. In some high-income countries the figures are considerably higher. The United States has an adult obesity rate above 40 percent. The United Kingdom is above 28 percent and rising. These are not marginal shifts in a population’s dietary habits. They are a transformation of the human body at civilisational scale within fifty years.

The standard explanation is also the most convenient one: people eat too much and move too little. This explanation has the virtue of placing the cause of the epidemic in the behaviour of individuals, which in turn places the responsibility for its resolution with those same individuals, and leaves the systems that produce, market, and distribute food largely outside the frame.

The science does not support that framing. Not entirely, and increasingly not at all.

What the research of the past thirty years has produced is a considerably more complex and considerably more uncomfortable picture — one in which the food system, the agricultural chemistry underpinning it, the neurobiological engineering of the products it sells, and the pharmaceutical and supplement industries that manage the consequences are operating as an integrated system whose overall effect on human health is poorly understood, insufficiently regulated, and in many of its key mechanisms operating contrary to the good intentions of many of the individuals within it.

The Scale of the Problem

The global obesity figures are not equally distributed, which is itself an important signal about causation. Obesity tracks closely with the penetration of the ultra-processed food system — not simply with wealth, and not simply with caloric availability. There are food-secure populations with high caloric intake that maintain low obesity rates. There are food-insecure populations with lower average caloric intake that have high obesity rates. The correlation with ultra-processed food consumption is, across multiple large epidemiological studies, considerably stronger than the correlation with total caloric intake.

The Global Burden of Disease study, one of the most comprehensive systematic analyses of health data assembled, identifies high body mass index as the third leading risk factor for death and disability globally, after high blood pressure and high fasting plasma glucose — both of which are themselves significantly associated with the same dietary patterns that produce obesity. The conditions are not separate. They are different expressions of the same upstream causes.

Type 2 diabetes, which was genuinely rare fifty years ago and is now a global epidemic affecting over 500 million people, is in the majority of cases either directly caused by or massively exacerbated by the same metabolic disruption that drives obesity. Non-alcoholic fatty liver disease, which did not exist as a clinical entity worth describing fifty years ago because it was so uncommon, now affects approximately one quarter of the global adult population. Cardiovascular disease, certain cancers, polycystic ovarian syndrome, gout, sleep apnoea — the list of conditions with strong mechanistic links to the current dietary environment is long and its total burden on healthcare systems is the largest single driver of healthcare cost in every country in which it has been measured.

Research context: The NOVA food classification system, developed by Carlos Monteiro at the University of São Paulo, provides the most rigorous framework for distinguishing ultra-processed foods from minimally processed and processed foods. Meta-analyses of studies using the NOVA classification consistently find dose-response relationships between ultra-processed food consumption and obesity, type 2 diabetes, cardiovascular disease, depression, and all-cause mortality. The associations persist after adjustment for total caloric intake, confirming that the effect is not simply about eating more — it is about what is being eaten.

Good Intentions, Unpredictable Consequences: The Agricultural System

The transformation of the global food supply over the past seventy years was not conceived as a project to make people ill. It was conceived, in significant part, as a project to prevent mass starvation. The Green Revolution of the mid-twentieth century — the development of high-yield crop varieties, synthetic fertilisers, and the agricultural chemistry that dramatically increased food production — is estimated to have prevented the deaths of hundreds of millions of people from famine. This is not a trivial achievement, and any honest assessment of the agricultural system must acknowledge it.

But the agricultural chemistry that made the Green Revolution possible has had consequences for the food supply and for human health that were not anticipated and are still not fully characterised.

Glyphosate, the active ingredient in the herbicide marketed under the name Roundup and now the most widely applied agricultural chemical in history, was approved for agricultural use in 1974 and has been applied to food crops at increasing scale ever since. By 2019, an estimated 9.4 million tonnes had been applied globally since its introduction. Its primary mechanism — the inhibition of the EPSPS enzyme in the shikimate pathway — was initially described as safe for mammals because mammals do not have a shikimate pathway.

The problem with this reasoning, which was not adequately considered at the time of approval, is that the gut microbiome does have a shikimate pathway. The trillions of microorganisms that constitute the human gut microbiome — and which are now understood to play a central role in metabolic function, immune regulation, mental health, and appetite signalling — are vulnerable to glyphosate through exactly the mechanism that was described as evidence of its safety.

Published research: Shehata et al. (2013) in Current Microbiology demonstrated differential susceptibility of gut bacteria to glyphosate, with beneficial species showing greater sensitivity than pathogenic ones. Subsequent research by Samsel and Seneff (2013, 2015) in Entropy and the Journal of Biological Physics proposed mechanistic pathways by which glyphosate disruption of gut microbiome composition could contribute to multiple chronic conditions including obesity, diabetes, and inflammatory bowel disease. While these papers remain contested in some quarters, the broader literature on glyphosate and gut microbiome disruption has grown substantially, and the regulatory approval framework that was applied when glyphosate was introduced did not require microbiome assessment.

Pesticide residues more broadly are present in the food supply at levels that individual regulatory agencies regard as safe on a compound-by-compound basis, but which have not been adequately assessed in combination — the mixture effects of dozens of compounds simultaneously present in a standard Western diet remain poorly characterised. The assumption that the safety of individual compounds at threshold levels implies the safety of their combined presence is an assumption, not a demonstrated fact.

The nutritional composition of the food supply has also changed in ways that are not captured by calorie or macronutrient labelling. The mineral content of fruits and vegetables grown in intensively farmed, chemically maintained soils has declined measurably over the past fifty years — a phenomenon documented in the work of Donald Davis at the University of Texas, whose analysis of USDA nutritional data between 1950 and 1999 found significant declines in protein, calcium, phosphorus, iron, riboflavin, and vitamin C in forty-three crops, attributable to the prioritisation of yield over nutritional density in breeding programmes. The food supply contains more calories and fewer micronutrients per unit than it did a generation ago. This matters metabolically: micronutrient deficiency is itself a driver of appetite dysregulation.

What is In the Food: The Engineering of Overconsumption

Ultra-processed foods are not simply convenient versions of whole foods. They are, in many cases, the products of deliberate and sophisticated engineering whose explicit commercial objective is to maximise consumption. The internal industry terminology — ‘bliss point,’ ‘crave-ability,’ ‘mouth-feel,’ ‘vanishing caloric density’ — describes a research and development process whose goal is to identify the precise combination of sugar, fat, salt, and texture that suppresses the normal satiety signalling that would otherwise tell the human organism to stop eating.

The science of this engineering is now reasonably well understood. The combination of rapidly digested refined carbohydrates and fats that characterises pizza, ice cream, crisps, and the majority of packaged snack foods produces a dopaminergic response in the brain’s reward circuitry that is, in its neurobiological mechanism, structurally similar to the response produced by addictive substances. The work of Ashley Gearhardt at the University of Michigan on the Yale Food Addiction Scale has established, across multiple studies, that a significant proportion of the population meets diagnostic criteria for food addiction, and that the foods most associated with addictive eating patterns are consistently the ultra-processed, high-fat, high-sugar products that constitute an increasing proportion of the global diet.

Children are a specific and particularly significant target of this engineering. The developmental neurobiology of children makes them both more susceptible to the dopaminergic reward signalling of ultra-processed foods and less capable of the executive function required to override it. The food marketing industry has known this for decades — the research on children’s food preferences, food advertising, and the mechanisms by which brand loyalty is established in early childhood is extensive, and significant portions of it have been conducted by or on behalf of the food industry itself.

Children love pizza and ice cream not because evolution designed them to eat pizza and ice cream. They love them because those products were specifically engineered to be loved by children — to hit the precise neurobiological targets that evolution did design, in service of a commercial objective that has nothing to do with the child’s nutritional needs.

The sugar question has its own specific history. The work of Cristin Kearns and colleagues at the University of California San Francisco, published in JAMA Internal Medicine in 2016, documented internal Sugar Research Foundation documents from the 1960s demonstrating that the sugar industry funded research specifically designed to shift the scientific consensus on dietary fat — implicating fat rather than sugar in cardiovascular disease and obesity — in ways that shaped dietary guidelines and public understanding for decades. The low-fat dietary guidelines that dominated nutritional advice from the 1970s onward produced a food industry response that reduced fat in processed foods and replaced it with sugar, producing products that were marketed as healthier and were, in terms of their metabolic effects, frequently worse.

The scientific literature on sugar and obesity is extensive. Robert Lustig’s mechanistic analysis of fructose metabolism, published in Nature Reviews Endocrinology (2012) and subsequently expanded, provides a detailed biochemical pathway by which the fructose component of high-fructose corn syrup — widely used as a sweetener in the American food supply since the 1970s and increasingly globally — produces hepatic fat accumulation, insulin resistance, and leptin resistance through mechanisms distinct from glucose metabolism. Leptin resistance is specifically relevant to obesity: leptin is the hormone that signals satiety to the hypothalamus, and its disruption produces the experience of hunger in the presence of adequate caloric intake — a mechanistic explanation for why people can be simultaneously over-nourished calorically and genuinely, neurobiologically hungry.

The Slaughterhouse and What Lies Behind It

The meat production system that supplies the majority of the protein in the Western diet operates at a scale and under conditions that are almost entirely invisible to the consumer purchasing its products. Paul McCartney’s observation is not merely rhetorical: if slaughterhouses had glass walls, the rational response of a significant portion of the population would be to reconsider their dietary choices. The observation is correct not only as a statement about the experience of animals within the system, but as a statement about what the system does to the food it produces.

The routine use of antibiotics in industrial livestock farming — not for the treatment of illness but for growth promotion and prophylactic disease prevention in animals kept at densities that would otherwise make disease endemic — is the largest single driver of antibiotic resistance globally. The WHO has identified antimicrobial resistance as one of the primary public health threats of the coming decades. The connection between the current antibiotic use pattern in food animal production and the diminishing effectiveness of the antibiotic arsenal available for human medicine is direct, documented, and the subject of a regulatory response that remains, by most assessments, inadequate to the scale of the problem.

The growth hormones used in beef production in the United States, and banned in the European Union precisely because of their hormonal activity, are endocrine disruptors — compounds that interact with the human hormonal system in ways that have been linked to early puberty onset, hormone-sensitive cancers, and metabolic disruption. The fat-soluble nature of these compounds means that their concentrations are highest in the adipose tissue and organs of the animals treated with them — the parts of the animal consumed as food.

The feeding practices of industrial animal agriculture have also changed the nutritional profile of the meat itself. Grass-fed beef contains omega-6 to omega-3 fatty acid ratios of approximately 2:1. Grain-fed feedlot beef contains ratios of 20:1 or higher. The shift in this ratio is metabolically significant: the balance between omega-6 and omega-3 fatty acids influences inflammatory signalling throughout the body, and the high omega-6 ratios characteristic of the modern diet are associated with chronic low-grade inflammation — a mechanism linking diet to cardiovascular disease, insulin resistance, and depression.

The processing of meat at scale introduces further chemical compounds not present in the original animal product. Sodium nitrite, used to preserve processed meats and maintain their colour, produces nitrosamines during digestion — compounds classified as probable carcinogens by the International Agency for Research on Cancer. The IARC classification of processed meat as a Group 1 carcinogen — the same category as tobacco and asbestos — is based on sufficient evidence from epidemiological studies of its association with colorectal cancer. This classification generated significant media coverage and significant pushback from the processed meat industry. The science has not changed.

Why Do People Eat Too Much? The Neurobiological Answer

The question of why obesity is increasing cannot be answered by willpower. It can be answered by neurobiology.

The human appetite regulation system evolved in an environment characterised by caloric scarcity and food variety. The hypothalamus, which integrates signals from leptin, ghrelin, insulin, peptide YY, and numerous other hormones to regulate hunger and satiety, is calibrated for a nutritional environment that no longer exists for most people in high-income countries. It is exquisitely sensitive to caloric density, to the presence of fat and sugar in combination, and to the reward signals generated by the dopaminergic system — because in the ancestral environment, foods that were high in calories and generated strong reward responses were rare, valuable, and worth consuming to excess when available.

Ultra-processed foods are designed to exploit precisely this calibration. They are calorically dense, nutritionally sparse, engineered for maximum reward response, and produced at a cost that makes them the most accessible calories in the modern food environment for the populations most vulnerable to their effects. The association between poverty and obesity in high-income countries — which is the inverse of the historical association between wealth and corpulence — is a direct consequence of the food economics of the ultra-processed system. The cheapest calories are the most engineered ones.

The gut-brain axis is a second and increasingly important mechanism. The gut microbiome — whose composition, as noted, is affected by agricultural chemicals, antibiotic use in both food production and medical practice, and the fibre poverty of ultra-processed diets — produces neurotransmitters including serotonin, dopamine precursors, and GABA, and communicates with the brain via the vagus nerve in ways that influence mood, appetite, and food preference. A microbiome disrupted by antibiotics, glyphosate exposure, and low dietary fibre is a microbiome producing different appetite signals than a healthy one. The appetite disruption of the modern dietary environment is not purely a matter of the foods being consumed. It is partly a matter of what the disrupted microbiome is telling the brain to want.

Research context: The work of Tim Spector at King’s College London, including the PREDICT studies examining individual metabolic responses to identical foods, has demonstrated that two people eating the same meal can produce dramatically different blood sugar and fat responses depending on their gut microbiome composition, sleep, stress levels, and time of eating. The implication is that population-level dietary guidelines based on average responses are of limited utility for individuals whose metabolic responses are highly variable — and that microbiome composition, which is malleable and influenced by the food environment, is a significant determinant of individual metabolic health.

The Pharmaceutical and Supplement Response

The pharmaceutical industry’s engagement with the obesity epidemic has followed a pattern recognisable from other chronic disease areas: the identification of a large and growing patient population, the development of pharmacological interventions that manage symptoms without addressing causes, and the establishment of treatment regimens that, in many cases, require indefinite continuation.

The GLP-1 receptor agonists — semaglutide (marketed as Ozempic and Wegovy), tirzepatide (Mounjaro), and their successors — represent the current frontier of pharmaceutical obesity intervention. They are genuinely effective in producing weight loss in the majority of patients who take them, and for patients with severe obesity and associated comorbidities they represent a real clinical advance. The mechanism — mimicking the gut hormone GLP-1 to suppress appetite and slow gastric emptying — works precisely because it operates on the same appetite signalling systems that ultra-processed food engineering has disrupted.

The commercial opportunity this represents has not been lost on the pharmaceutical industry. The global market for GLP-1 agonists is projected to reach several hundred billion dollars annually within the current decade. This is, by any measure, one of the largest pharmaceutical market expansions in history — driven entirely by a condition whose primary cause is the food environment, and whose treatment does not address that food environment in any way.

The supplement industry occupies a different but structurally related position. The decline in micronutrient density of the food supply, the disruption of gut microbiome function, the chronic inflammation associated with the modern dietary pattern, and the appetite dysregulation produced by ultra-processed food engineering have created a large and growing market for products that address each of these deficits individually — probiotics, omega-3 supplements, vitamin and mineral complexes, protein powders, appetite management products, and the extensive category of weight management supplements whose efficacy is, in the majority of cases, not supported by evidence of the quality required for pharmaceutical approval.

The food industry creates the metabolic disruption. The pharmaceutical industry manages the symptoms. The supplement industry fills the gaps in between. Each is a legitimate business operating within the law. Together they constitute a system in which illness is more commercially valuable than health.

The question of whether this constitutes a conspiracy requires precision. A conspiracy requires coordination, shared intent, and deliberate concealment. What the evidence supports is something less intentional and more systemic: a set of industries whose commercial incentives are structured in ways that make the perpetuation of the problem more profitable than its resolution, operating in a regulatory environment that has consistently been shaped by the lobbying power of those industries, producing outcomes that no individual actor necessarily intended but that the system as a whole reliably generates.

The sugar industry’s documented funding of research designed to exonerate sugar and implicate fat, described earlier in this article, is the clearest example of intentional manipulation of the evidence base. The pharmaceutical industry’s consistent underinvestment in prevention relative to treatment is a rational response to commercial incentives that does not require coordination to produce. The food industry’s deployment of addiction science in product development is legal, profitable, and has been conducted with full knowledge of the health consequences of the products being engineered.

The structural observation that deserves examination: the three largest growth markets in global consumer healthcare — obesity pharmaceuticals, gut microbiome products, and metabolic health supplements — are each responses to conditions whose primary driver is the ultra-processed food system. The same investment capital that funds the largest food companies is, in many cases, invested in the pharmaceutical and supplement companies that treat the conditions those foods produce. The financial system does not distinguish between industries. It distinguishes between returns. The returns in treating the obesity epidemic are, currently, substantially larger than the returns in preventing it.

Is There a Conspiracy, or Is This Simply Capitalism?

The honest answer is that the distinction matters less than it might appear. A conspiracy requires intent. What the food, pharmaceutical, and supplement industries have produced does not require intent to explain. It requires only the reliable operation of incentive structures that reward market expansion, penalise disruption of existing revenue streams, and externalise the health costs of commercial decisions onto individuals and healthcare systems.

The regulatory framework that should constrain these incentives has been consistently inadequate. In the United States, the Generally Recognised As Safe (GRAS) designation — which allows food manufacturers to self-certify the safety of new food additives without independent review — has resulted in the introduction of thousands of compounds into the food supply without systematic evaluation of their individual or combined effects. In Europe, the precautionary principle has produced a somewhat more restrictive regulatory environment, but lobbying against the restriction of specific compounds has been a consistent feature of the regulatory process.

The revolving door between regulatory agencies and the industries they regulate is documented. Senior officials at the FDA have moved to senior positions at food and pharmaceutical companies. Senior figures from those companies have occupied regulatory positions. This does not prove that regulations are corrupt. It does create a structural environment in which the regulatory framework is developed in close proximity to the interests of the industries it is supposed to constrain.

The dietary guidelines that governments issue, and that shape food education, school meals, and public health messaging, are produced through processes in which industry influence is well documented. The composition of the committees that produce these guidelines, the funding of the research on which they draw, and the political pressure applied to specific recommendations — notably the recommendation to reduce ultra-processed food consumption, which has been consistently weakened in final guidelines relative to what the independent scientific evidence supports — are matters of public record.

The specific case of dietary fat and cardiovascular disease guidance illustrates the mechanism. The low-fat dietary guidelines introduced from the 1970s onward were based on epidemiological evidence that has subsequently been substantially revised. The food industry response — replacing fat with sugar in processed foods and marketing the resulting products as heart-healthy — produced a dietary environment that increased metabolic disease burden. The guidelines that enabled this response were shaped by a process in which the sugar industry’s financial influence on the research literature is now documented. The people who died of the conditions that resulted did not die as the result of a conspiracy. They died as the result of a system whose incentives were misaligned with their health, operating in a regulatory environment that was inadequate to correct those misalignments.

The Children

The framing of childhood obesity as a parental failure — too much screen time, not enough exercise, poor food choices — is both partially accurate and substantially misleading.

Children in every culture, exposed to ultra-processed foods, prefer them to whole foods. This is not a failure of parenting. It is the predictable output of an engineered food product operating on a developing nervous system that has not had fifty thousand years of evolutionary pressure to resist it, because the product did not exist until fifty years ago.

The marketing of ultra-processed foods to children — through cartoon characters, toy tie-ins, digital advertising calibrated by age and platform behaviour, and the placement of these products at child eye level in retail environments — is a sophisticated and heavily resourced industry whose entire purpose is to establish brand preference and consumption habits during the developmental period when those habits are most durable. The food industry spends more on marketing to children than it spends on research and development.

The school food environment in most high-income countries reflects the same commercial dynamics. The contracts between food companies and school districts, the vending machine arrangements, the branded food programmes — these are not features of a system designed to optimise child nutrition. They are features of a system in which the commercial interest of food companies and the budget constraints of educational institutions have produced an accommodation that serves neither the children’s health nor the public interest.

The long-term consequences for this generation of children — who have grown up in a food environment more thoroughly engineered for overconsumption than any previous generation, whose gut microbiomes have been shaped by the dietary patterns of early childhood in ways that have long-term metabolic consequences, and who are arriving at adulthood with rates of insulin resistance, fatty liver disease, and metabolic syndrome that would have been considered extraordinary in any previous generation — are already visible in the clinical data. They will be more visible in the healthcare data of the next two decades.

What the Evidence Actually Supports

The evidence supports the following conclusions, stated as precisely as the current state of the research allows.

Obesity is not primarily a failure of individual willpower. It is primarily the consequence of a food environment that has been engineered to produce overconsumption, in combination with agricultural and food processing practices that have altered the nutritional and chemical composition of the food supply in ways that disrupt normal metabolic and appetite regulation.

The agricultural chemicals applied to the food supply — herbicides, pesticides, growth promoters, preservatives, and the chemical residues of industrial livestock production — have effects on human gut microbiome composition, endocrine function, and metabolic regulation that were not adequately assessed at the time of their introduction and remain incompletely characterised. The assumption that these compounds are safe individually at threshold concentrations does not address their combined effects or their effects on the gut microbiome, which is a metabolically active system involved in appetite regulation, immune function, and mental health.

The ultra-processed food industry has applied neurobiological research to the engineering of products specifically designed to override normal satiety signalling. The commercial incentives for this engineering are straightforward and have not been adequately constrained by the regulatory framework. The consequences for population health are measurable and large.

The pharmaceutical and supplement industries have developed significant commercial interests in the management of the conditions produced by the food environment. The financial returns from treating obesity and its comorbidities are currently larger than the financial returns from preventing them. This creates a systemic incentive against the fundamental changes to the food environment that would reduce the burden of metabolic disease.

The regulatory, political, and scientific environments in which food policy is developed have been materially shaped by the commercial interests of the industries whose products are driving the epidemic. The mechanism does not require intent at the individual level. It requires only that the incentive structures operate as they demonstrably do.

The world is not getting heavier because people have lost the will to be thin. It is getting heavier because the system that feeds it has been optimised, with extraordinary precision and at very considerable expense, to make it so.

What Would Actually Help

The interventions that the evidence supports are not glamorous, are not commercially attractive, and are not the interventions that receive the largest share of investment.

Reforming the regulatory framework for food additives and agricultural chemicals to require assessment of microbiome effects, mixture toxicity, and endocrine activity — rather than individual compound safety at threshold concentrations — would produce a substantially different regulatory outcome for many compounds currently in widespread use. The scientific tools to conduct this assessment exist. The political will to require it has not yet materialised.

Redirecting agricultural subsidy from commodity crops — primarily corn and soy, the raw materials of ultra-processed food production — toward whole food production and regenerative farming practices that restore soil microbiome diversity and nutrient density would change the economics of the food system in ways that make nutritionally dense foods more price-competitive with ultra-processed alternatives. The subsidy structure that currently exists in the United States and to varying degrees in European agricultural policy is, in its effect, a subsidy for the food environment that produces obesity.

Restrictions on the marketing of ultra-processed foods to children — comparable to the restrictions on tobacco advertising that were introduced when the health evidence for tobacco became sufficiently clear — would reduce the commercial effectiveness of the engineering described in this article. The evidence for harm is now sufficiently clear to justify this intervention on the same grounds.

Investment in food education that is genuinely independent of industry funding, and that provides people with the practical and cognitive tools to navigate a food environment designed to exploit their neurobiology, would not solve the structural problem but would improve the individual capacity to resist it.

None of these interventions is radical. All of them are resisted, consistently and at considerable expense, by the industries whose commercial models depend on the current food environment remaining in place. The resistance is not evidence of conspiracy. It is evidence of the rational operation of commercial incentives in the absence of regulatory frameworks adequate to constrain them.

The question of whether those frameworks will be reformed before the healthcare costs of the current trajectory become politically unsustainable is, like many of the questions this series has asked, not one to which an honest answer is currently optimistic.

THE LONG VIEW — HEALTH, FOOD & THE INDUSTRIAL SYSTEM

Scientific references in this article are directional and drawn from published peer-reviewed literature. Readers are encouraged to consult primary sources. The article reflects the state of the published evidence and does not constitute medical advice.

It is a Sunday afternoon. The boy is seven. His grandfather has collected him from the house — Mum is working, Dad is away — and they are sitting in the garden with squash and a plate of biscuits that Granddad made sure were there. The boy has been quiet in the car. Now he is not.

About the Xbox

Granddad, do you want to know what level I’m on?

Tell me then.

Level forty-three. It took me ages. Nobody else in my class is on forty-three.

That’s a long time playing.

Yeah. I play it when I get home from school mostly. And at the weekend when Mum’s doing her calls.

Does she do many calls at the weekend?

Quite a lot. She has to. She says sorry about it. She always says sorry. I don’t mind though.

He picks at the edge of a biscuit.

It’s a bit boring though sometimes. The game I mean. Not boring exactly. I don’t know the word. You just keep doing the same bit again.

What would you do instead, if you could?

I don’t know. Go somewhere maybe. I used to go to Jack’s house but he moved. There’s the park but Mum says not by myself.

How far is the park?

Not far. Like five minutes. But she says it’s not safe. She showed me a thing on her phone once about why. I didn’t really understand it.

I used to walk to the park when I was your age. By myself.

Did you? Were you scared?

No. I just went.

That’s weird.

Editorial note: It is not that the child does not want to be outside. It is that the permission structure has made outside feel like a managed exception rather than a default. He has absorbed the anxiety without understanding its origin. The Xbox is not his preference. It is the available option.

About School

Granddad, am I clever?

What makes you ask that?

Mrs Patterson said I need to focus more. She said it to Mum at parents’ evening. I wasn’t supposed to hear but I did.

What do you think she meant?

I don’t know. I try to focus. It’s just sometimes I’m thinking about something else and then I’ve missed the bit she was explaining and then I don’t know what we’re doing.

Does that happen often?

Sometimes. Not always. Maths is alright. I like maths when I know what to do. When I don’t know it feels bad.

What does it feel like?

Like everyone else knows and I’m the only one that doesn’t. Even if that’s not true.

I expect it’s not true.

Probably. But it feels true.

He finishes the biscuit. Starts another one.

I’ve got tutoring on Thursdays now. For maths and reading. Mum sorted it. The lady is nice but I don’t really want to do more school when school’s finished. Is that bad?

No. That’s not bad.

Mum thinks I’m behind. I don’t know what I’m behind on. Behind who?

That’s a very good question.

She worries a lot. About my future. She talks about my future a lot and I don’t really know what my future is. It’s quite far away isn’t it.

It is. You’ve got plenty of time.

She doesn’t think so. She says year three is really important. Is year three really important?

Year three is important. So is sitting in a garden eating biscuits.

She’d say that’s wasting time.

Editorial note: The child is not failing. He is distracted, as seven-year-olds have always been distracted, because the developmental work of being seven includes a great deal of internal processing that looks, from the outside, like inattention. The tutoring is not a response to a deficit. It is a response to parental anxiety meeting an educational culture that has moved the baseline expectation forward in ways that the research does not support. He is not behind. The schedule is ahead of him.

About the Week

Granddad, what do you do on Mondays?

Not much these days. I go to the shops. I might walk by the river. Why?

I’ve got football on Mondays. I don’t really like football that much but Dad signed me up. He thinks I should do a sport.

Do you like any sports?

I quite like swimming. We do it at school. But there’s no swimming club near us that fits. Mum looked.

What else have you got in the week?

Football Monday. Coding club Tuesday — that’s at school so it’s okay. Thursday is tutoring. Mum was going to do drama on Wednesday but she said it was too much. So Wednesday is free. I like Wednesdays.

What do you do on Wednesdays?

Xbox mainly. Or sometimes I draw. I like drawing but nobody knows that really. It’s not one of my activities.

Why not?

I don’t know. I just do it in my room. I drew a whole city once. It took ages. With roads and everything.

I’d like to see that.

I’ve still got it. I could show you.

I’d like that very much.

He looks pleased, then a bit uncertain, like he’s checking whether that was the right thing to say.

It’s probably not that good.

I’ll be the judge of that.

Editorial note: He has a genuine interest and is already managing the expectation downward before showing it. He has learned, somewhere in the past two years, to present his own enthusiasm with a disclaimer. That is not a natural feature of a seven-year-old. It is an acquired one. The question of where he acquired it is worth sitting with.

About Mum and Dad

Granddad, do you and Grandma ever argue?

Sometimes. Everyone argues sometimes.

What do you argue about?

Oh, small things usually. Who forgot to do what. Nothing serious.

Mum and Dad argue about serious things.

Do they.

I hear them at night sometimes. Not always. Sometimes. I can’t always hear the words but I can tell it’s serious by the sound.

That can feel worrying.

I don’t get that worried. I did at first. Now I just put my headphones on. I’ve got headphones in my room for the Xbox. They’re quite good at blocking stuff out.

He says this matter-of-factly. He has solved the problem with the tools available.

Dad’s away a lot. For work. He has to be. Mum says he doesn’t have to be that much but he says he does. I don’t really know who’s right.

You don’t need to know who’s right.

When he is home he tries to do stuff with me. Like football in the garden. But he’s usually on his phone a bit. I don’t say anything because he’s trying. You can tell when someone’s trying.

That’s a perceptive thing to notice.

What does perceptive mean?

It means you see things clearly.

Oh. I see quite a lot of things I don’t say. Is that okay?

That’s more than okay.

Editorial note: He is doing what children in uncertain household environments consistently do: he is reading the emotional atmosphere with precision and managing his own responses to avoid disruption. This is a cognitive load that is not visible on any school assessment. It is, for some children in some households, the primary work of their day.

About Being By Yourself

Granddad, do you get lonely?

Everybody gets lonely sometimes. Do you?

A bit. Not loads. It’s weird to say because there’s always people around at school. But you can be with people and still be a bit lonely can’t you.

Yes. You can.

I’ve got friends. I mean I have friends. But they’re kind of different now from Year One. In Year One you were just friends with everyone. Now it’s more complicated.

In what way?

There’s like groups. And you have to know which group you’re in. And sometimes the group changes and you don’t know why. Liam was my best friend and then he wasn’t and I don’t know what happened. He’s not mean to me. He’s just with different people now.

That’s hard.

Yeah. I didn’t tell Mum because she’d have rung the school and made it worse.

So you sorted it yourself.

I didn’t really sort it. I just got used to it. Is that the same thing?

Sometimes it is.

He looks at the garden for a while. Not unhappy. Just looking.

I like it here, Granddad. It’s quiet. Not bad quiet. Good quiet.

I know what you mean.

At home it’s always either noise or screens. Even when nobody’s home there’s the TV on because Mum says it’s better than silence. I don’t think silence is that bad.

Neither do I.

Can we just sit here for a bit?

We can sit here as long as you like.

Editorial note: He is seven. He has correctly diagnosed the difference between the noise of an anxious household and the quiet of a secure one. He does not have the language for it. He has the perception.

About Growing Up

Granddad, what were you like when you were my age?

A bit of a nuisance, to be honest. Always outside. Always in some kind of trouble that wasn’t serious trouble.

What kind of trouble?

Coming home late. Getting muddy. Going further than I was supposed to.

Did you get told off?

Sometimes. Not very severely.

I don’t really get told off much. I don’t do anything to get told off for. There’s not much opportunity is there. I’m at school or at home or at an activity. You can’t really get in trouble at any of those.

Is that a good thing or a bad thing, do you think?

I don’t know. Maybe both. I think you probably learn stuff from getting in small trouble. Like how far you can go.

I think you’re probably right.

My teacher says I’m very sensible. She means it as a nice thing. But I don’t think I want to be the sensible one. I think I want to be the one who does something a bit surprising sometimes.

There’s still time for that.

Is there? It feels like everything’s already quite planned out.

He pulls a piece of grass from the edge of the garden path and twists it between his fingers.

Granddad, do you think I’m going to be alright?

I do.

How do you know?

Because you ask good questions. And you pay attention to the right things. And you know the difference between good quiet and bad quiet, which a lot of grown-ups have never worked out.

That’s not really proof though is it.

No. But it’s enough.

Okay.

He seems to accept this. He leans slightly against his grandfather’s arm. Not dramatically. Just a small adjustment, like settling.

Granddad?

Yes.

Can we do this again next Sunday?

Every Sunday you like.

I might bring my drawings.

Please do.

They sit in the good quiet for a while. The afternoon does not require anything else of either of them.

The child in this account is not in crisis. He is not being abused, neglected in the legal sense, or failed by parents who do not love him. His parents love him very much and are doing what the world they live in requires of them.

What he is experiencing is the aggregate of every structural pressure described in the preceding articles in this series: the household under economic strain, the fear that keeps him indoors, the schedule that fills the time that boredom should occupy, the screen that substitutes for the unstructured world outside, the educational anxiety that has arrived two years early, the parental attention that is present but divided, the social complexity that he is navigating without a guide.

He will probably be alright. The grandfather is not wrong about that. The Sunday afternoons in a garden with someone who has nowhere else to be and nothing to assess are, for this child, doing what no tutoring session, coding club, or football practice can do: they are giving him an experience of being known, unhurriedly, by someone who finds him interesting exactly as he is.

The question the series asks — who is protecting childhood? — has a partial answer here. Not the state. Not the school. Not the algorithm. The grandfather, with a garden and a plate of biscuits and the patience to sit in the good quiet.

THE LONG VIEW — CHILDHOOD & SOCIETY

This article is a companion piece to Whatever Happened to Childhood? The characters are composite and fictional. The circumstances are not.

This essay is about that condition. It is shared, in different registers, by the child of immigrants who speaks the national tongue more fluently than the neighbours who distrust her; by the person whose faith marks them as forever suspect in a country their grandparents helped build; by the gay or transgender person raised on the same promises of belonging as everyone else, only to find those promises were not written with them in mind. These are not identical experiences, and it would be a disservice to flatten them into one. What unites them is not identity but the experience of conditional belonging the sense that acceptance depends on continual proof rather than being taken for granted. And what the conditionally accepted are finally denied, as we will see, is not grandeur or glory. It is something smaller and much harder to win back: the freedom to be ordinary.

The double vision

More than a century ago, W.E.B. Du Bois described this as “double consciousness” the sense of always seeing yourself partly through the eyes of others. He was writing about being Black in America, but the idea has proved remarkably portable, because the experience it names travels. To live this way is to carry two sets of eyes in one head: the way you see yourself, as a full participant with a stake in the common project, and the way you have learned that others see you, as a question still waiting to be answered.

The doubling reveals itself in ordinary moments the insider never has to notice. It is the form that asks for your nationality and then, a few lines down, asks again where you were born, as though the first answer needed a footnote. It is the pause before you introduce yourself at the school gate, deciding which version of your name to offer the one that is yours or the one that is easier for them. It is the job interview where you sense, correctly, that you must be not merely as good as the other candidate but visibly, reassuringly better, so that hiring you feels like no risk at all. It is the parents’ evening where you are the only mother in a headscarf, and you feel the room adjust by a fraction of a degree as you sit down. None of these is a slammed door. Each is the quiet, cumulative reminder that your presence is a thing to be explained.

The reminders are exhausting precisely because they are invisible to the people who deliver them. When a stranger asks “but where are you really from?” they imagine they are being curious, even warm. They do not feel the small door closing, because for them there is no door. You have answered the question five hundred times, and each answer chips away at the fiction that you might simply be from here, without qualification. In time you learn to hear the asterisk in almost anything even the compliment, “your English is so good,” which carries its own small insult inside it.

The cruelty of near-belonging

It might seem that the person who is wholly rejected the newcomer who speaks no word of the language, who never expected to belong suffers most. In material terms they often do. But there is a particular torment reserved for those who come close to belonging and are stopped just short: the ones who do everything right. Who master the accent, earn the degrees, join the institutions, marry into the culture, raise children who know no other home everything the society appeared to ask only to discover that the true price of entry was never effort but ancestry, and ancestry cannot be earned.

There is a bitter arithmetic in this. Each achievement was supposed to buy a little more security, and for a while it seems to. Then the successful outsider is called “a credit to their kind,” and the phrase gives the whole game away. No accomplishment dissolves the category; it only makes you an admired specimen of it. You become the example produced when someone wants to prove the system is fair evidence marshalled against the very grievance you feel most keenly. This is why the most accomplished outsider is often the loneliest: they have climbed high enough to see clearly that the summit they were promised does not exist.

For the member of a minority faith, the same dynamic takes the form of a loyalty perpetually on trial. Their patriotism is never simply assumed; it must be performed, and even then it is doubted the moment the news carries a story that implicates, however remotely, someone who shares their religion. They are asked to condemn, to explain, to distance themselves from strangers on the other side of the world with whom they share nothing but a label. The insider is never asked to answer for the crimes of people who merely resemble him. That, in the end, is what belonging is: the presumption of individuality, the right to be judged only for what you yourself have done.

For the gay or transgender person raised inside a tradition that has no place for them, the exclusion can be more intimate still, because it comes from within the very family that was meant to be the shelter from the world outside. This is a matter of lived experience rather than an argument anyone is being asked to settle here; readers may hold deep and sincere convictions about it, and the point is not to ask them to abandon those convictions. It is only to notice what the experience feels like from inside. Such a person grew up on the same expectations as everyone else: the wedding, the grandchildren, the ordinary continuity of generations and made those expectations their own before understanding that to fulfil them, as themselves, was not permitted. To be a stranger at your own dinner table is its own kind of homelessness. You can be surrounded by people who love you and still go unseen — loved, that is, as a version of yourself that does not exist.

The theft of the ordinary

What is taken, in every one of these cases, is chiefly the ordinary. Belonging, when you have it, is mostly invisible; it is the freedom not to think about it. The insider does not wake and weigh his relationship to the nation, his faith’s standing in the public square, whether his love is legible to strangers. He simply lives, moving through the day without translating, without scanning each room for the temperature of its welcome. That unconscious ease is the real privilege, and it is exactly what is withheld. The outsider is condemned instead to consciousness; the world will not let them stop noticing.

Brené Brown draws a useful line here between fitting in and belonging. Fitting in, she writes, is being accepted for being like everyone else; belonging is being accepted as you are. The outsider is often permitted the first and quietly denied the second welcomed on the condition that they file down whatever makes them distinct. And so they curate. They soften the accent in one room and thicken it in another, rehearse the safe answer to the intrusive question, and develop an exquisite sense of the moment a conversation is about to turn, when it is time to laugh along at their own expense to keep the peace. Onlookers call this being “well integrated,” not realising they are praising the successful concealment of a wound.

Against the false consolations

It would be easy here to reach for comfort to say that home is something you carry within, that belonging is really a matter of self-acceptance, that the outsider’s vantage is secretly a gift. There is truth in each of these, but offered too quickly they become evasions: they let the society off the hook by turning belonging into a private spiritual achievement rather than a shared obligation. Nobody should have to become a philosopher of exile merely to survive a place that will not accept them. A society that demands an endless supply of resilience from one group of people while asking none of another has simply organised its cruelty efficiently and called it strength of character.

And yet, because despair is also a form of surrender, it is worth saying what is equally true. The line between insider and outsider is not a law of nature. It is drawn and redrawn constantly, and it moves. Almost every group now comfortably inside the circle of belonging was, within living or recorded memory, the suspect outsider of its day the wrong religion, the wrong region, the wrong blood, the wrong love. The boundary that feels eternal to those it excludes is, on a longer view, embarrassingly recent and plainly provisional. That is no guarantee it will move again in time to comfort any particular life. But the door has been moved before, and it was moved by people who refused to accept that the guest room was theirs by right.

There is, too, a belonging the outsider builds without always naming it: the fierce, chosen kinship of others who share the condition. The friend who needs no explanation. The family assembled from those who see you as you are. The tradition remade from within by people who love it enough to insist it make room for them. These are not consolation prizes. They are among the deepest bonds people form, precisely because they are not inherited but chosen, tested, and defended.

What belonging gives back

The feeling described here is not, at bottom, a weakness or a pathology. It is a form of moral clarity. The person who feels the sting of conditional acceptance feels it because they have correctly perceived a promise being broken. The society told them in its schools, its anthems, its founding stories that they belonged, and they believed it, as children believe what they are taught. The pain lives in the gap between the promise and the practice, and is, in its way, a tribute to the promise. It is the insiders grown cynical about belonging who no longer feel it; the outsider still believes in the ideal, which is why its betrayal wounds them.

Social scientists have long argued that this is not a marginal need but a central one. Roy Baumeister and Mark Leary, in a landmark 1995 paper, described belonging as a fundamental human motivation, its absence linked to measurable harm to health and well-being. What the outsider is denied, then, is not a luxury. And the demand folded inside the wound is correspondingly modest: not pity, not tolerance, but the ordinary. Not to be admired as an exceptional specimen or thanked for their resilience, but released from the obligation of being remarkable permitted, at last, the insider’s unglamorous freedom to be mediocre on a bad day, to be from here without an asterisk, to be judged as an individual rather than a sample, to love and pray and speak without translating.

It would be a mistake to read all this as a plea made by minorities on their own behalf. Belonging is not a gift the majority bestows on the margins; it is something a whole society gains when fewer of its members are made to live as guests in their own home. The sociologist Robert Putnam spent a career showing how much a community’s health rests on the invisible ties of trust and reciprocity between its members. Every person kept permanently at the door is a thread of that fabric left unwoven a store of loyalty, talent, and attachment held in reserve, never fully given because never fully invited. A country that learns to move its boundary is not being charitable. It is repairing itself.

Whether that repair happens is not, in the end, up to the outsider. It rests with everyone who holds without having earned it the power both to draw the line and to erase it. The question the outsider poses, simply by continuing to live with dignity in a place that half-receives them, is not really a question about the outsider at all. It is a question about the society: whether its talk of belonging was ever meant, or whether it was only ever a house with a guest room, a smile at the door, and a lock that opens from one side.

The fear is usually narrated as a fear of malevolence: the machine will turn on us, the Terminator made flesh. That is the wrong fear, or at least the wrong emphasis. The thing on the screen has no malice in it and will never acquire any. The accurate fear is stranger and harder to dismiss that we are learning to build machines that can go almost anywhere and understand almost nothing, and that we are beginning to give those machines guns. The danger is not a mind that hates us. It is a body that has outrun its mind, with a weapon bolted to its back.

To see why that is the real shape of the thing, it helps to understand what these machines have genuinely mastered, and what they have not. The gap between the two is the whole of the matter. The same principle applies to aerial drones most readers will think of those first but legged robots make the problem unusually visible, because they move through human spaces in recognisably human ways.

What the machine has conquered

Begin with the astonishing part, the part that makes the footage possible at all. Until quite recently, getting a legged machine to cross broken ground was one of robotics’ standing humiliations; the robots of the old DARPA challenges lurched and toppled, and a stiff gust or a misplaced rock was enough to fell them. That problem is now substantially solved. Modern quadrupeds cross rubble, mud, stairs and slopes with a fluency that would have seemed like science fiction a decade ago and they do it, for the most part, without even looking. They feel their way through the force in each leg, reacting to the ground hundreds of times a second, yielding where no human engineer could have scripted the yield in advance.

They learned this, in the main, not from instruction but from practice — millions of hours of it, run not on hardware but inside simulations, where a virtual robot can fall ten million times, on ten thousand randomised surfaces, at no cost and no risk, until the reflexes that survive are burned into a controller and copied onto the real machine. This is why locomotion fell. Walking is, in the relevant sense, a closed problem: the goal is unambiguous, stay up and get there; the verdict is automatic, it fell or it did not; the physics of a foot on ground can be simulated; and the failures are cheap to rehearse. Given those four gifts, the machine can teach itself and it has.

Hold on to those four conditions, because everything that follows turns on their absence everywhere else.

What it cannot do

For the thing the machine has not mastered is the thing the word “understand” actually points at. It crosses the ground; it does not comprehend the ground. It will step around an obstacle with great elegance and have no reliable idea whether the obstacle is a rock, a sandbag, or a crouching child. The faculties that remain stubbornly unsolved recognising what a thing is, grasping what is happening in a scene, judging what a situation demands are precisely the faculties a weapon most needs and a walking machine least requires.

And they resist for the same reasons locomotion yielded, run in reverse. The open world offers none of the four gifts. It has no rulebook to rehearse against: predicting how a real, cluttered scene will unfold is not looked up but guessed — approximately, and wrongly in the ways that matter most. It offers no automatic verdict: a game scores itself, but nothing in the world announces whether a machine has correctly told friend from foe. Its mistakes are neither cheap nor safe to practise. And it holds an inexhaustible supply of situations no one anticipated the long tail of the unexpected that no amount of training quite empties.

This is why the triumph that conquered the game of Go has not conquered the battlefield. A program could master Go by playing itself because Go is closed perfect rules, instant verdict, no cost to a million games. The world is open, and an open world is one you can neither fully simulate, reliably score, nor safely practise in. It is the same reason aircraft can now be designed almost entirely inside a computer while robots cannot be raised inside one: air obeys equations we have written down and checked against reality; the contents of a street obey no equation at all. There is a physics of flight. There is no physics of what that shape in the doorway is.

Notice the cruel symmetry this produces. The very property that makes the machine untrustworthy that it acts on a picture of the world it cannot fully verify, in situations it was never shown — is the property that makes it dangerous once it is armed. We are not nervous about the robot that crosses the rubble. We are nervous, rightly, about what it might conclude is a target.

The worst possible order

Now set a second fact beside the first, and the unease sharpens into something more precise. The body has become cheap, and it has spread.

The surefootedness that took decades to achieve is now, in hardware terms, very nearly a commodity. The methods were published openly, in papers any competent laboratory on earth can read; the parts motors, batteries, sensors, a graphics chip rode down the cost curves of the smartphone and the electric car. A capable quadruped can now be bought for a few thousand pounds. Ghost Robotics’ Vision 60 has been adopted by the United States Air Force and Marine Corps for patrol and perimeter work; China has become perhaps the most prolific developer of armed robot “dogs”; and a clip of a rifle-bearing quadruped no longer signifies a defence-budget marvel so much as an off-the-shelf platform with an attachment.

What has not spread, tellingly, is the part that might one day supply the judgement. The frontier of machine intelligence the largest models, the world-models the whole industry is now racing to build still concentrates in a very small number of laboratories in two countries, because it alone still demands the rare ingredient: vast computation, the one resource that has not commoditised, and the one that export controls are now fought over. So the field has split into two tiers advancing at wildly different speeds. The bodies are everywhere and getting cheaper. The minds are scarce, concentrated, and for the open world not yet good enough.

This is, when you sit with it, very nearly the worst order in which these two capabilities could arrive. We are industrialising movement before we have mastered judgement fixing weapons to the competence while the comprehension that might make it safe lags behind, locked in a few hands. It is the movement we are arming, not the understanding.

The black box and the deadline

The attraction is obvious. A robot that can patrol a perimeter for twenty-four hours without fatigue, carry sensors into dangerous terrain, or enter a mined building ahead of soldiers offers genuine military advantages and saves the lives of the people who would otherwise have gone in first. The problem begins further along the chain, when observation becomes judgement, and judgement becomes permission to kill.

The institutions that might govern this have noticed, and are moving at the speed institutions usually manage against a problem moving faster than they are. Since 2014 the United Nations has discussed lethal autonomous weapons the systems that, once switched on, select and strike targets without a human pulling the trigger under the unlovely machinery of the Convention on Certain Conventional Weapons. The Secretary-General, António Guterres, has called such weapons “morally repugnant” and pressed states to conclude a legally binding instrument by the end of 2026 — this year on the argument that decisions over life and death should not be surrendered to a process no human can see into.

The diplomatic momentum is real but lopsided. A 2024 General Assembly resolution drew 166 states in favour and three against; a follow-up in 2025 won more than 150. But the holdouts are not marginal. The powers most reluctant to let international law constrain their autonomous arsenals are precisely the powers furthest ahead in building them the United States, Russia and China and the talks remain, as campaigners put it, consultations rather than negotiations. Meanwhile the Pentagon’s most recent budget sought a record sum for artificial intelligence and autonomy, and its “Replicator” initiative aims to field thousands of expendable autonomous systems at speed, even as elements of AI-assisted targeting have already been reported in conflicts from Gaza to Ukraine. The machines are arriving faster than the rules for them.

Underneath the diplomacy sits the conceptual knot, and it is the same knot this argument has been tying throughout. The accepted way to talk about these weapons sorts them by how much human remains in the decision: a human in the loop authorises each strike; a human on the loop merely supervises, and may intervene; a human out of the loop is gone the moment the machine is switched on. Everyone says they want “meaningful human control.” Almost no one can agree what it means and the deeper trouble is that meaningful control over a system whose reasoning you cannot inspect may be a contradiction in terms. You cannot truly oversee a judgement you cannot see. That is the black box the Secretary-General fears, and it is the engineering reality this whole subject keeps circling: the learned, judging layer is the powerful one and the unauditable one at once. Where it errs, the error is silent and a second problem follows it. If a machine no one can fully understand kills someone it should not have, who is to answer for it? The operator who trusted it? The programmer who could not have foreseen the case? The accountability does not so much fail as evaporate.

The case on the other side

It would be dishonest to present only the prosecution, because the case on the other side is not frivolous, and the people who make it are not monsters.

Human soldiers, the argument runs, are themselves a catalogue of failure. They tire, they panic, they misremember their training, they give way to fear and rage and the appetite for revenge; the historical record of what frightened, exhausted young men do to civilians is not a reassuring one. A machine does none of that. It does not grow angry, does not seek to punish, does not fire first out of terror for its own life — it has no life to fear for. A well-designed autonomous system, its defenders contend, might in principle keep the laws of war more faithfully than a human under fire: holding fire when the rules demand it, never shooting in revenge, applying proportionality without adrenaline. Some argue that to forbid such a system, if it genuinely spared civilians, would itself be a moral failure. And there is the hard realism: the technology cannot be uninvented, an adversary unburdened by scruple will build it regardless, and a power that unilaterally renounces it may simply lose. Existing humanitarian law, the holdouts add, already governs how force may be used; perhaps what is wanting is enforcement, not a new treaty.

These are real arguments, and the better ones turn on a single, genuine “if” , if the machine’s judgement worked. Which is exactly where the case meets the wall this essay has been building toward. Every defence of the autonomous weapon assumes the competence we have least achieved: the reliable discrimination of the combatant from the child, the surrender from the ambush, the funeral from the formation. Precision of movement is not precision of understanding. The machine that crosses the rubble flawlessly is the same machine that cannot tell you, reliably, what is standing in the rubble and a weapon that is superb at reaching its target and unreliable at identifying it is not a more humane soldier. It is a faster mistake. The soldier who panics can at least be tried; the black box that misjudges leaves no one to stand in the dock.

Coda

So return, at the last, to that small lurch of unease in front of the screen, and grant it its due. The fear is not superstition, and it is not really about the Terminator. It is the entirely rational recognition that we have built a thing which can go anywhere and understand almost nothing, and that we are beginning briskly, cheaply, and in many countries at once — to give it a gun.

The surefootedness is genuine, and it is ours; we are entitled to be proud of it. The blindness is genuine too, and we are choosing, for now, to look past it. There is a version of the future in which comprehension catches up to competence before we have committed ourselves in which the machine learns to read the ground it already crosses so well, the rules catch up to the machines, and the deadline that falls this year yields something more than a record of a chance not taken. There is another version in which we settle the matter the other way: by deploying first, and discovering afterwards what the machine could not see.

The deadline is real, and it is this year. The machines are already in the field. The question is no longer whether we can build something that walks into the dark without stumbling; we have. The question is whether we will arm it before it can tell us what is waiting there and whether, having armed it, we will still be able to name who is answerable for what it decides. The surefooted thing crosses the ground beautifully. It does not know what the ground is for. We do; and that, for now, is the only safeguard that counts.

John is thirty-six. Susan is thirty-five. They have a daughter, Ellie, who is three and currently conducting a serious investigation into what happens when you put raisins in a wellington boot. There is a second child on the way, due in four months. They live in a house they can afford. They are not wealthy. They are not stressed. These two facts are related, and the relationship between them was not accidental.

This is the story of a decision — not a dramatic one, not a crisis, not a breakdown that forced change — but a deliberate, quietly radical choice to organise their lives around what they actually value rather than around what the world around them assumed they should want.

They did not opt out. They are still employed, still connected, still very much part of the society they live in. What they opted out of was the specific version of that life in which the schedule never empties, the debt never shrinks, the evenings are too short and the weekday mornings are a controlled emergency, and somewhere in the middle of all of it there is a child who is technically supervised but not quite reached.

They know that world. They lived in it for three years before Ellie arrived. They decided, with some clarity and without excessive drama, that they did not want their children to grow up in it.

The Conversation That Started It

It was not a single conversation. It was several, spread over about six months, beginning when Susan was pregnant with Ellie and accelerating in the weeks after she was born, when the reality of what they had signed up for became concrete in ways the antenatal classes had not entirely prepared them for.

The surface question was practical: they could not afford full-time nursery on their current trajectory and maintain the mortgage payments on the house they had bought at the top of their budget, service the car finance on two cars that spent most of the week in the driveway, and keep up the standard of living — the holidays, the evenings out, the frictionless online purchasing — that they had built up incrementally and never quite examined.

But the surface question opened onto a deeper one. John, in one of those conversations, said something that Susan later described as the sentence that changed the direction of everything: ‘I’m working this hard so we can have a life, but the working is the life. Where’s the actual bit?’

It was not an original observation. But it was his, said at the right moment, and it made the right kind of sense.

The question was not how to earn more. The question was how much they actually needed — and what they were giving up to chase the rest.

What They Did

What followed was not a sudden lurch toward simplicity. It was a methodical, unglamorous, and at times uncomfortable audit of every financial commitment they carried and every assumption about lifestyle that underpinned it.

They started with the numbers, because the numbers were honest in a way that the feelings were not always able to be. They listed every outgoing — fixed costs, variable costs, the subscriptions that renewed automatically and were never quite cancelled, the habits that had calcified into expenses without anyone deciding they should. What they found, looking at it all together rather than in the separated-out form in which it normally appeared, was that a significant portion of what they were spending was not on things they particularly valued. It was on things they had acquired in a period when spending felt like reward, and which had become part of the background texture of their lives without being examined.

The two cars were the largest single item. One of them — John’s, used for a commute he was now doing twice a week since his employer had moved to hybrid working — was costing them, between finance, insurance, tax, and running costs, more per month than they spent on food. They sold it. John cycles on the days he goes in, takes the train when the weather makes cycling impractical. The exercise he was previously paying a gym membership not to do turned out to be available in the commute for free.

The house was the second conversation. They did not sell it. But they looked honestly at whether they had bought more house than they needed at the time they bought it, and concluded that they had — that the fourth bedroom they had vaguely told themselves was for ‘the future’ was in practice a room they heated and cleaned and paid interest on for no current purpose. They did not downsize, because with a second child coming the space would earn its place. But the exercise of looking at it clearly changed something in how they thought about the relationship between space and cost.

The subscription audit was both the most tedious and the most revealing part of the process. They found eleven recurring charges for services they either did not use or duplicated. Three streaming platforms where one would do. A meal kit delivery they had started in enthusiasm and continued in inertia. A cloud storage subscription for photos that could be stored elsewhere for nothing. A magazine they did not read. A fitness app used twice. The monthly total was not enormous. Cancelled across a year, it was a weekend away.

The rule they adopted: If it doesn’t actively make our life better, it goes. Not if it might be useful. Not if it was a good idea in principle. If it is not, in practice, making our life better.

What Susan Did

Susan had been working full-time in a marketing role she was good at but had never quite loved. She had also been privately calculating, for most of Ellie’s first year, the mathematics of full-time nursery fees against her post-tax salary and arriving at a number — the actual financial benefit of her full-time employment, after childcare — that was considerably smaller than the number her gross salary suggested.

This calculation is not universally discussed with the frankness it deserves. For many households with young children, the financial benefit of both parents working full-time, after childcare costs are properly accounted for, is materially smaller than the headline salaries imply. The emotional and practical costs — the compressed evenings, the logistical complexity, the particular exhaustion of the working parent collecting a tired toddler at six o’clock and having two hours before bedtime to be everything that needs to be done — are not reflected anywhere in that calculation.

Susan went to four days a week. The pay reduction was fifteen percent. The reduction in nursery fees was twenty-five percent, because Ellie went three days rather than five. The net financial effect was positive. The effect on the quality of the day that Ellie spent with her mother, and on the quality of Susan’s presence in that day, was not measurable in financial terms but was apparent to anyone who was around the two of them.

She is not, as she is at some pains to point out, a person who has given up a career. She has shaped a career around a life rather than a life around a career. There is a meaningful difference, and she is clear about it. She expects to return to full-time work. She expects, when she does, to be more focused and more effective than she was before, because she will have made a deliberate choice rather than an anxious drift.

This is not a template. The arithmetic of part-time work and childcare is different for every household, and there are households where reducing hours is not financially viable. What Susan’s choice illustrates is not the specific decision but the willingness to examine the calculation honestly rather than accepting the default.

John’s Half-Day

John works in project management for a mid-sized engineering firm. He is moderately well paid, reasonably well regarded, and genuinely interested in his work. He is also, by his own candid assessment, a person who used to allow the work to expand into whatever time he made available to it — checking emails at ten at night, taking calls on Saturdays, treating the boundary between working and not working as optional in a direction that only ever went one way.

He stopped. Not dramatically, not with a declaration, but by making a decision that after six-thirty in the evening he was not available for work, and that weekends were not a productivity reserve. The work did not suffer. The projects got delivered. What changed was that the evening — previously a depleted, half-attentive, screen-adjacent interval between the commute and bed — became available for something else.

What it became available for was Ellie’s bath, and the reading after it, and the particular kind of conversation that happens with a three-year-old when she is clean and warm and not yet asleep and the adult sitting with her is not also somewhere else in their head. John, who had been present in his household in the logistical sense for most of Ellie’s first three years, started being present in a different way. He describes it, with the slight awkwardness of someone not accustomed to this kind of language, as ‘actually being there instead of just being in the building.’

He has also started cooking properly, in the way his mother used to cook — from ingredients, on a budget, with the expectation that a good meal is not a special occasion but a Tuesday. The weekly food shop is smaller and better. Ellie, with the completeness of conviction available only to small children, has informed everyone in her orbit that her daddy makes the best pasta. This may or may not be objectively true. It is clearly true in the way that matters.

The children will not remember the car. They will remember the pasta. The reading. The bath. The particular quality of a parent who arrived fully, with nothing more pressing waiting behind them.

The Grandparents

Both sets of grandparents are close enough to be part of the daily fabric rather than occasional visitors, and John and Susan have been deliberate about making that relationship work in ways that are genuinely reciprocal rather than merely convenient.

The temptation, in a household trying to reduce pressure, is to treat nearby grandparents as a childcare solution — reliable, free, and emotionally invested. John and Susan have been careful to resist this framing. Ellie’s grandmother on Susan’s side takes Ellie on Tuesday mornings, which she does because she wants to and because the relationship between them has become one of the visible threads of Ellie’s week. But this arrangement was discussed, offered, and accepted as a gift rather than assumed as a resource. The distinction matters, both for the grandparents’ sense of their own role and for the sustainability of the arrangement over years rather than months.

What the grandparents offer is not primarily childcare. It is continuity — the particular experience of being known by adults who have known your parents since before you existed, who carry the family’s accumulated story, and who have, unlike parents, the luxury of engagement without the underlying responsibility. Ellie at her grandparents’ house exists in a different register than Ellie at home. She is less managed, more observed, more freely herself. The grandparents are not filling a gap in the schedule. They are providing something qualitatively distinct from what parents can provide, and John and Susan are clear about the value of that distinction.

In return, John and Susan are present for their own parents in ways that the previous version of their lives made difficult. A Sunday lunch that is actually on Sunday, not rescheduled twice and then abandoned. A parent who is unwell and needs someone to drive them to an appointment. The small but accumulating evidence that the relationship is a relationship and not a logistics arrangement.

The decision to prioritise the extended family as a living system rather than a support network changes the nature of what is available within it. Families that function as networks of mutual support do not require the state, the market, or the institution to fill the roles that families have filled across every culture in which human beings have organised themselves.

What They Gave Up

It would be dishonest to present this as costless, and John and Susan do not present it that way. There are things they have given up, and some of them were not trivial.

They no longer take the kind of holiday they used to take — the long-haul flight, the resort with the pool, the holiday that was partly about recovering from the pace of the preceding year. They go to places they can drive to. They stay in cottages and go to beaches and do the kind of holiday that requires more tolerance for unpredictability and less tolerance for disappointment when the weather fails to cooperate. They have found, with some surprise, that they prefer it — not abstractly, as a principle, but in practice, in the specific experience of it. Ellie is better on a beach than in a resort. She is three. She does not need a pool.

There are professional costs that are harder to assess. Susan’s step back from full-time work in the early years of her career is a decision whose full consequences will only be visible in retrospect. She is aware of this. She has made it with open eyes, and with the expectation that the career she returns to will be shaped by choices rather than default. Whether the arithmetic of that decision resolves well over a decade is not yet knowable. What is knowable is that the alternative — the full-time continuation, the compressed evenings, the nursery at five days — had its own costs, and that those costs were in many respects already visible.

John’s decision to hold the boundary on work hours has probably cost him something in the competitive assessment of people who have not made the same decision. He works in an environment where visible commitment — the late email, the weekend availability, the willingness to be always on — is still, beneath the official language of work-life balance, a form of currency. He has decided, consciously, to spend less of it. Whether that decision costs him a promotion he would otherwise have had is a question he genuinely cannot answer. He has decided it is worth the uncertainty.

The Practical Architecture

For families considering a version of what John and Susan have built, the architecture is less romantic and more methodical than the principle suggests. The decisions they made are reproducible, but they require the willingness to sit with the numbers honestly and to make choices that the surrounding culture will not always validate.

Start with the audit. Every outgoing, every subscription, every financial commitment written in one place and looked at in total. Not categorised or softened — just listed and seen clearly. Most households that do this find at least one surprise that is large enough to matter.

Separate wants from habits. Some spending is genuinely valued. Some is inertial — the accumulated residue of choices made in a different period, maintained because cancelling them requires a decision and the default is continuation. The question is not whether to spend but whether the spending is chosen.

Do the childcare arithmetic properly. For households with young children, the calculation of what both parents working full-time actually contributes, after childcare costs, tax, and the other expenses that employment generates, is often materially different from the headline income suggests. It is worth doing precisely.

Protect the evenings. The working day will expand to fill whatever time is offered. The decision to protect the evening — from work, from screens, from the management of logistics — requires an active choice and produces a compound return that shows up in the quality of the hours spent with children and partners, and eventually in the quality of the children themselves.

Involve the grandparents as partners. Not as resources. The relationship between grandparents and grandchildren is valuable in its own right and distinct from what parents provide. Families that cultivate it deliberately rather than instrumentally find that what it offers — continuity, unconditional interest, the transmission of a longer story — is something no market or state institution can replicate.

Expect scepticism. The choice John and Susan have made is not the one their social environment generally validates. People will ask, with varying degrees of directness, whether they are sure. Whether they have considered the long-term. Whether they are being sensible. The answer is that they have considered the long-term rather carefully, and this is what it looks like.

What Ellie Has

Ellie does not know that her parents made a decision. She does not know that the life she is growing up in is different from the one it might have been. She knows that when she wakes up there is usually a parent at home, that the evenings are not rushed, that her grandfather comes on Tuesdays and he always brings something from the garden, that her father makes pasta on Wednesdays, that the house is not particularly tidy but it is warm and it is hers.

She knows, with the completeness of knowledge available to very small children who have not yet learned to doubt it, that she is the most interesting thing in the room. Not because she is spoiled — she is not — but because the adults around her are genuinely paying attention. Not managing her toward the next transition. Not filling a gap in the schedule. Paying attention.

The research on what children need in the first years of their life is neither complicated nor particularly surprising when it is stated plainly. They need consistent, present, unhurried adult attention. They need the experience of being known by people who find them interesting. They need boredom, and the freedom to resolve it on their own terms. They need physical space, and the permission to use it. They need to see adults who are not continuously stressed as the natural condition of grown-up life.

Ellie has most of these things most of the time. Not because her parents are unusual, or exceptionally capable, or have made a discovery that was previously unavailable. But because they made a deliberate choice to organise their lives so that these things could be present.

The children do not need everything. They need enough. And they need the adults around them to be present enough, unhurried enough, and secure enough to give them what enough actually looks like.

The Second One

The second child is due in four months. Susan and John face this with a combination of genuine excitement and the clear-eyed realism of people who have already done it once and know what it costs in sleep, patience, and the particular cognitive load of two small people with different and sometimes competing needs.

They also face it with something that the version of themselves three years ago did not have: a household that has been deliberately simplified to absorb the disruption that a new child creates. The financial buffers are real. The schedule has slack built into it. The grandparents know their role and are ready for it. John and Susan have, between them, an ongoing honest conversation about what they need from each other that did not exist in the same form before they started the broader project of examining how they were living.

They are not in a perfect position. Nobody is. They will be exhausted. There will be months that feel unmanageable. The second child will be entirely different from the first in ways they cannot predict and will require adjustments they have not anticipated.

But they will face those months from inside a structure they chose, with clear sight of what matters and a reasonable understanding of what they are willing to give up for it. That is not nothing. In the landscape of contemporary family life, it is quite a lot.

Ellie, for her part, has been informed that a baby is coming. She has absorbed this information and returned, with the magnificent resilience of the genuinely occupied, to the wellington boot experiment. There is still quite a lot to find out about raisins.

John and Susan are fictional. Their circumstances are drawn from patterns visible across a growing number of households in which one or both partners have made deliberate decisions to trade income, status, or career velocity for time, presence, and reduced stress. The choices they have made are not available to every household in equal measure — economic constraint is real, and this article does not pretend otherwise.

What is available to every household, regardless of income, is the habit of examining what is actually valued against what is actually being done — and asking, with honesty, whether the gap between them is as unavoidable as it feels.

THE LONG VIEW — FAMILY, CHOICE & INTENTIONAL LIVING

This article is a companion to Whatever Happened to Childhood? and Granddad, Can I Tell You Something? It is the series’ attempt at an answer, not a conclusion.

The story goes like this: Earth is fragile. One serious asteroid strike, one engineered pandemic, one runaway climate tipping point, one nuclear exchange, and the experiment of human civilisation is over. Mars, therefore, is insurance. Establish a self-sustaining colony there and the species survives whatever befalls the home planet. The logic is clean, the ambition is stirring, and the people advancing it are serious, intelligent, and in several cases extraordinarily well resourced.

The story has one significant problem. Mars is not a lifeboat. It is one of the most hostile environments in the accessible solar system, it is separated from Earth by a journey that is reliably dangerous, and the proposition that a meaningful number of human beings could survive there independently — in the event of a catastrophe that had already destroyed the industrial civilisation whose supply chains and technology their survival would require — is not supported by any honest reading of what is actually known about the planet or the physics of keeping people alive on it.

This article does not argue that Mars should not be explored. Scientific exploration of the solar system is valuable in its own right and has produced returns in knowledge, technology, and perspective that justify its costs on those terms alone. What it argues is that the specific claim — that Mars constitutes a viable off-planet backup for human civilisation — is not currently true, is unlikely to become true within any timeframe that is relevant to the risks it is supposed to address, and that the framing of it as such deserves more scrutiny than it typically receives.

It also asks a question that the mainstream coverage of the Mars programme does not often ask: whose interests does the narrative of planetary colonisation actually serve, and are those interests fully aligned with the human survival rationale offered as its justification?

What Mars Actually Is

Mars is, by the standards of the planetary neighbourhood, a neighbour. At its closest approach to Earth it is approximately 54 million kilometres away. At its furthest it is over 400 million kilometres. The journey, with current propulsion technology, takes between six and nine months depending on the alignment of the planets — a window that occurs roughly every twenty-six months and cannot be moved.

The planet at the other end of that journey is not a fixer-upper. It is a world that, by every measure relevant to human habitation, is aggressively hostile to the biology of the organisms being proposed as its inhabitants.

The atmosphere is ninety-five percent carbon dioxide, with the remaining five percent split between nitrogen, argon, and trace gases. It is also extremely thin — surface pressure on Mars is less than one percent of Earth’s at sea level. A human being exposed to the Martian atmosphere without a pressure suit would not suffocate on carbon dioxide. They would experience explosive decompression within seconds. The atmosphere offers no protection and no breathable air. Every molecule of oxygen required for human respiration must be either transported from Earth or manufactured on-site using energy that itself must be generated from scratch on a planet whose distance from the sun delivers roughly half the solar energy available at Earth’s surface.

The temperature range is extreme. At the equator, a summer afternoon might reach 20 degrees Celsius. The same night, the same location will be minus 80. The average surface temperature is approximately minus 60 degrees. Most of the planet, most of the time, is far colder than Antarctica in winter, without Antarctica’s relatively thick atmosphere to retain any heat.

The dust is a specific and underappreciated hazard. Martian dust storms are the largest weather events in the solar system. Global storms can envelop the entire planet for months, reducing solar power generation to near zero and coating every surface with a fine, electrically charged particulate that is, based on current analysis, chemically reactive in ways that are damaging to both equipment and human lung tissue. The 2018 global dust storm lasted for over eight months and was the proximate cause of the end of the Opportunity rover mission. An Opportunity rover does not need to breathe. A human being does.

The dust is not merely an inconvenience. Martian regolith contains perchlorates — oxidising compounds toxic to the human thyroid and potentially to broader human metabolism — at concentrations that would be considered a significant contamination event on Earth. Any Mars habitat requires complete separation of its interior environment from the dust that will, inevitably, attempt to penetrate every seal, joint, and airlock. The long-term health consequences of perchlorate exposure in a closed habitat environment are not yet fully characterised because no human being has ever lived in one.

The Magnetosphere Problem

Of all the environmental challenges Mars presents to human habitation, the one most consistently underweighted in popular accounts of the colonisation project is the absence of a global magnetic field.

Earth’s magnetosphere — generated by the movement of liquid iron in the planet’s outer core — is not a passive feature of the planet’s geology. It is the active shield that deflects the continuous stream of charged particles emitted by the sun, and that prevents the solar wind from stripping away the atmospheric gases that make surface life possible. Without it, Earth’s atmosphere would be gradually eroded by solar wind in the same way that Mars’s atmosphere was eroded over the past four billion years.

Mars lost its global magnetic field approximately four billion years ago, when its core cooled and the dynamo that generated the field ceased to function. The consequence was the progressive loss of most of the planet’s original atmosphere, the loss of its liquid surface water, and the exposure of the surface to the full radiation environment of the inner solar system, unmediated by either magnetic protection or atmospheric shielding.

The radiation environment on the Martian surface is approximately forty times higher than on Earth’s surface for galactic cosmic rays alone, with periodic spikes during solar particle events that can temporarily raise that figure by orders of magnitude. The radiation doses accumulated by a Mars colonist over a two-year stay — the minimum meaningful mission duration — would, in current dosimetry models, significantly elevate their lifetime cancer risk. Long-term habitation, over years and decades, would represent radiation exposures for which there is no safe human precedent outside of specific medical contexts.

This problem is not unsolvable in principle. Underground habitation, or habitation beneath sufficient water shielding, could substantially reduce radiation exposure. But underground habitation on Mars means that the colony is not living on Mars in any meaningful sense. It is living inside Mars — in excavated tunnels or lava tubes, cut off from the surface environment, dependent on artificial light for any plant growth, and requiring the maintenance of sealed, pressurised, radiation-shielded volumes indefinitely and without failure.

Mars has no magnetic shield and will not have one on any human timescale. The planet’s surface is continuously bathed in radiation that damages DNA, causes cancer, and degrades the central nervous system. There is no engineering solution to this that does not involve essentially living underground — on a planet where underground habitation at any scale has never been attempted.

There is also no prospect of terraforming Mars on any relevant timescale. The calculations for generating a functioning magnetosphere, thickening the atmosphere to survivable pressures, raising the average temperature, and introducing a water cycle have been done, and the most optimistic peer-reviewed estimates place the minimum timescale for even partial terraforming — atmosphere only, not the full suite of conditions required for surface life — at several thousand years, requiring energy inputs and industrial capacity that do not exist and could not be created without first solving the problem it is supposed to solve.

The Journey and Its Costs

Before any colonist encounters the radiation environment, the dust storms, or the perchlorates, they must survive the journey. The transit to Mars, at current technological capacity, is a six-to-nine-month exposure to the deep space radiation environment — which is, if anything, worse than the Martian surface, since there is no planetary bulk absorbing radiation from one hemisphere.

The physiological consequences of prolonged microgravity are well documented from the International Space Station programme: bone density loss, muscle atrophy, fluid redistribution causing vision impairment in a significant proportion of astronauts, immune system suppression, and cardiovascular deconditioning. These consequences are manageable for the six-month missions conducted to the ISS, because the returning astronauts land on Earth and undergo rehabilitation in a full-gravity, full-atmosphere, medically equipped environment. The Mars colonist lands on Mars and cannot be evacuated. If their vision has been permanently affected — a known risk of long-duration spaceflight — there is no ophthalmology suite waiting for them.

The launch vehicle requirements for any meaningful Mars programme are extraordinary. A self-sustaining colony requires not dozens but thousands of people — estimates vary, but the geneticist minimum viable population for long-term survival without inbreeding depression is generally placed at several hundred individuals, and a colony capable of maintaining the industrial and technical complexity required for its own survival in the absence of Earth resupply would require considerably more. Each person requires life support, food, water recycling infrastructure, medical equipment, power generation, construction materials, agricultural systems, and the tools and knowledge to maintain all of the above.

The mass required to transport and supply a meaningful Mars colony is not currently calculable in terms of any existing launch system because it exceeds the capacity of any existing or planned launch system by several orders of magnitude. The most optimistic proposals involve hundreds of heavy-lift launches over decades, each carrying the risk of failure that is inherent in rocket flight, and each failure carrying consequences — loss of life, loss of critical equipment, loss of systems whose absence jeopardises the entire enterprise — that cannot be absorbed without setback.

The accident record of rocket launches is, in historical terms, remarkably good given the physics involved. But it is not zero, and the margin for failure in a mission architecture where every component of a habitat’s life support system is being delivered by a separate vehicle is substantially smaller than in any mission architecture that has previously been attempted. A launch failure that destroys a power generation system, a water recycling module, or a year’s food supply in transit does not merely delay the programme. It may kill the people waiting for it.

The Backup That Isn’t

The civilisational insurance argument for Mars colonisation rests on a specific chain of logic: that the risks to Earth are real, that they could be civilisation-ending, and that a self-sustaining off-planet population would survive whatever ended Earth’s civilisation and preserve the human species.

Each link in this chain deserves examination.

The risks to Earth are real. An asteroid impact of sufficient scale, an engineered pathogen of sufficient lethality, a nuclear exchange of sufficient scope, a climate transition of sufficient speed — any of these could, in the relevant scenarios, end organised human civilisation on Earth. This is not disputed. The question is whether a Mars colony, as currently conceived or plausibly achievable, addresses these risks.

Consider the most likely civilisational threat scenarios. A pandemic — even an engineered one of catastrophic lethality — does not end the possibility of human survival on Earth. Isolated populations, remote communities, submarine crews, and coincidentally immune individuals would survive, and the planet’s biosphere would remain intact. The argument for Mars as pandemic insurance requires the pathogen to be sufficiently lethal and transmissible to kill every human being on Earth without exception — a biological scenario that, while theoretically conceivable, is not supported by any historical or epidemiological precedent.

A nuclear exchange that produces nuclear winter — the most commonly cited existential risk scenario — would reduce Earth’s surface temperature, collapse agricultural systems, and kill the majority of the human population. It would not sterilise the planet. Humans survived the Toba supervolcano eruption approximately 74,000 years ago, an event that produced environmental conditions comparable to a severe nuclear winter and reduced the human population to an estimated few thousand individuals. The species survived. The planet was habitable. Mars, in the same scenario, remains exactly as hostile as it currently is.

The only scenario in which Mars provides meaningful civilisational insurance is one in which Earth itself is rendered permanently and completely uninhabitable — the planet’s surface destroyed or stripped of atmosphere. The candidate events for this outcome are an asteroid impact of a scale that has not occurred in the past four billion years, or a stellar event from a sun that is currently in the most stable phase of its lifecycle. These scenarios are real in the timescales of stellar physics. They are not relevant to the timescales being used to justify current Mars investment.

Mars is a backup for a catastrophe that is unlikely to occur within any timeframe that is relevant to current investment decisions. The catastrophes that are likely to occur within relevant timescales do not require a backup planet. They require better management of the one we have.

The Opportunity Cost

The total estimated cost of a human Mars programme — from current development through to the establishment of a colony of any meaningful size — varies by proposal and by how honestly the full supply and maintenance costs are accounted for. Conservative estimates for a crewed mission programme reaching the surface of Mars and returning safely are in the range of hundreds of billions of dollars. Estimates for a permanent colony of any self-sustaining size are in the range of trillions of dollars over decades.

These are resources being deployed in a world that simultaneously contains: one billion people without reliable access to clean water; a global food security system under structural strain from climate change and supply chain fragility; pandemic preparedness infrastructure that was demonstrated, within living memory, to be inadequate; biodiversity loss at rates that threaten the stability of the ecological systems on which the planet’s food production depends; and nuclear arsenals maintained by multiple states in a deteriorating security environment.

The argument that addressing these challenges and pursuing Mars colonisation are not in competition — that one does not come at the cost of the other — requires accepting that the capital, the engineering talent, and the political attention available for civilisational risk management are effectively unlimited. They are not. Every dollar and every engineer-hour directed toward a Mars programme is a resource not directed elsewhere. The opportunity cost is real.

The specific claim that Mars investment generates technological spinoffs that benefit Earth’s challenges has some validity and should not be entirely dismissed. Space programmes have historically generated genuinely useful secondary technologies. But the claim that the spinoff return justifies the primary cost requires comparing it to the return that the same investment, directed at the primary problem, would have produced — a comparison that is rarely made honestly in the advocacy for planetary colonisation.

The risk analysis is also relevant here. The resources required to meaningfully improve Earth’s resilience to the most probable civilisational risk scenarios — pandemic preparedness, early asteroid detection and deflection, food system resilience, nuclear risk reduction — are, by orders of magnitude, smaller than the resources required to establish a self-sustaining Mars colony. A fraction of the Mars budget, directed at Earth-based risk reduction, would produce a substantially larger reduction in civilisational risk than the full Mars budget directed at its stated goal.

The deflection of a civilisation-threatening asteroid — the one scenario in which off-planet alternatives are genuinely relevant — requires early detection and a relatively modest propulsive intervention, not a habitable colony on another planet. The global asteroid detection and deflection infrastructure that would provide genuine protection against this risk is funded at a level that would embarrass the Mars programme by comparison.

Why Are We Actually Doing This?

The question is not rhetorical. The Mars programme — both the public agency version and the private commercial version — is large, expensive, persistent, and supported by individuals and institutions with very significant power and resources. The justifications offered publicly are the exploration rationale, the science rationale, and the civilisational insurance rationale. The preceding analysis has argued that the insurance rationale, as commonly stated, does not survive serious scrutiny.

That leaves a question worth asking directly: what else might be driving the ambition?

The first alternative rationale is the simplest and the most benign: genuine exploratory ambition. Human beings have consistently sought to go further, see more, and extend the boundary of the known — not because it is strategically optimal but because it is, in some deep sense, characteristic of the species. The Apollo programme was not primarily justified by its practical returns. It was justified by the fact that human beings wanted to go to the Moon, and had the capacity to do it, and that this was felt to be sufficient reason. The same sentiment, scaled up and better resourced, is a plausible and not unrespectable driver of the Mars ambition.

The second rationale, less often stated and considerably more commercially significant, is resource extraction. Mars is not resource-rich in the ways that have historically driven exploration and colonisation — it has no immediately accessible deposits of materials not available on Earth in abundance, and the cost of returning any extracted resource to Earth would make it economically non-viable by any current calculation. But the trajectory of this calculation changes if in-space manufacturing and the space economy develop along the lines that some in the industry project. The ability to operate in deep space — to establish infrastructure, to refuel, to maintain presence — has long-term commercial and strategic value that is not captured in any current economic model but that serious people with serious money are actively trying to position for.

Speculative but worth examining: the first entity to establish sustainable off-planet operations — whether governmental or commercial — acquires a form of strategic advantage that is entirely novel in human history. Space law, as it currently exists, does not clearly resolve the question of who owns the resources extracted from celestial bodies. The Outer Space Treaty of 1967 prohibits national appropriation of celestial bodies but does not clearly prohibit commercial resource extraction. The race to establish presence — flags, infrastructure, and operational precedent — in advance of the legal framework catching up is a dynamic familiar from earlier periods of terrestrial expansion. The Mars programme, in this reading, is less about survival and more about position.

The Sovereignty Question

The overlap between billionaire private space ventures and the political systems of the countries in which they are based is not accidental. The largest private Mars programmes are operated by individuals whose wealth is denominated in the currency and regulatory environment of specific nation states, and whose ambitions in space are not separable from the geopolitical competition between those states and their rivals.

A private Mars colony, established and operated by a company incorporated in a single country, owned by a single individual, would represent something genuinely new in the political history of the species: a sovereign-equivalent entity operating outside the jurisdiction of any existing state, whose governance structure would be determined entirely by its founder, and whose relationship to the Earth-based legal and political order would be, at best, ambiguous.

This prospect has been discussed with varying degrees of candour by the individuals most likely to bring it about. The framing has generally been libertarian — freedom from the constraints of existing political systems, the opportunity to build new institutions from first principles, escape from the regulatory and democratic frameworks that constrain activity on Earth. These are not, on examination, arguments about survival. They are arguments about power.

The question of who governs a Mars colony — who makes the decisions, under what legal framework, with what accountability to the people living in it — is not a detail. It is the central political question of any serious colonisation programme, and it is one that the current advocacy for Mars settlement has not addressed with anything approaching the seriousness it deserves. A colony that is wholly dependent on resupply from Earth, whose residents have no avenue of political redress, and whose governance is determined by the entity that controls the resupply chain is not a free society. It is a company town at interplanetary scale.

The following is speculative but analytically coherent: for an individual with significant wealth and the political vulnerabilities that come with it — regulatory exposure, democratic accountability, the possibility of legal consequences that accumulate as political winds shift — the establishment of an off-planet presence represents something qualitatively different from the civilisational insurance being offered as public justification. It represents personal insurance: a jurisdiction that does not yet exist, a territory not subject to the courts that might otherwise have standing, a place to be that is outside the reach of the institutions of the society that generated the wealth now being used to leave it. Whether this is a motivation or merely a coincidental benefit of the stated programme is a question that the individuals involved are not well positioned to answer objectively.

What Should Be Done Instead

The argument of this article is not that space should be abandoned, that Mars should not be visited, or that scientific exploration of the solar system is a misallocation of resources. It is the narrower argument that the specific claim — Mars as civilisational insurance, as the backup plan for an endangered species — does not currently justify the scale of investment being directed at it, and that the resources available for civilisational risk management would produce a substantially larger reduction in actual risk if directed differently.

The practical alternative is not glamorous. It involves investment in early warning and deflection systems for near-Earth objects. It involves pandemic preparedness infrastructure capable of containing a pathogen of genuinely catastrophic potential, rather than the fractured and underfunded system that performed inadequately against a moderately serious one. It involves food system resilience — the agricultural and storage redundancy that would allow human civilisation to survive a multi-year disruption to growing conditions. It involves nuclear risk reduction, which remains, in the considered assessment of most analysts who study existential risk, the single most likely mechanism by which human civilisation could be ended in the next century.

These investments do not have the narrative power of a rocket to another planet. They do not offer the opportunity for individual heroism, or the frisson of genuine frontier. They are, in the main, the unglamorous business of making existing systems more resilient, filling known gaps in known defences, and applying the rigorous probability-weighted risk analysis that any serious infrastructure engineer would apply to a system whose failure modes include civilisational collapse.

They are also, by a significant margin, more likely to work. The probability that a well-funded global pandemic preparedness system would contain a civilisation-threatening pathogen is considerably higher than the probability that a Mars colony of any currently achievable scale would survive the loss of Earth resupply. The probability that an effective asteroid deflection capability would prevent a civilisation-threatening impact event is considerably higher than the probability that a Mars colony would be habitable by a meaningful human population within the timeframe of the risk.

The species does not need a backup planet. It needs better risk management of the planet it has. The difference between these two projects is not merely strategic. It is the difference between a solution that addresses the actual problem and a solution that addresses the problem in a way that is more interesting, more fundable, and more personally useful to the people funding it.

The Honest Assessment

Mars is a fascinating, scientifically significant world. The robotic exploration of Mars has been one of the genuine achievements of the space age, and the scientific return from missions to the planet has expanded human knowledge of planetary formation, the history of water in the solar system, and the conditions under which life might or might not persist in environments beyond Earth.

A crewed mission to Mars, conducted as a scientific and exploratory enterprise, would be a historic achievement and would generate knowledge not available from robotic exploration. This case can be made honestly, and it is sufficient to justify a programme of crewed Mars exploration on its own terms.

What it cannot honestly be made to justify is the civilisational insurance claim. Mars, as a backup for human civilisation against the threats most likely to threaten it within any relevant timescale, is not a viable option and will not become one in any timeframe that is actionable from the perspective of current risk management. The physics of the planet, the biology of the human organism, the engineering requirements of self-sufficient habitation, and the economics of the supply chain combine to make it so.

The people advancing the Mars programme are not fools. They know this. The question of why the civilisational insurance framing has become the dominant public justification for a programme whose actual drivers are more complex and more personal is one that deserves more sustained examination than it has received.

In the meantime, the asteroid is not tracked by ambition. The pandemic is not contained by narrative. The food system is not resilient by virtue of rockets. The actual problems require the actual resources. The resources are finite. The choices are real.

THE LONG VIEW — SCIENCE, POWER & PLANETARY FUTURES

This article presents analytical commentary and speculative analysis. The speculative sections are identified as such. Scientific data referenced is directionally accurate; readers seeking primary sources are encouraged to consult the peer-reviewed literature on Mars habitability, radiation biology, and existential risk.

Kampala, Uganda

MAMA AMARA

Kumi District village

Two women. One family. Three hundred kilometres apart.

The same decade. Entirely different exposures to its consequences.

Africa did not cause the climate crisis, the AI displacement wave, or the breakdown of international trade and pharmaceutical supply chains. It is absorbing consequences that were made elsewhere, by people who will not bear them.

THE WEIGHT OF IT

MOVEMENT ONE

What the World Does to Women Who Were Never Asked

Two voices from Uganda, 2025 to 2030 — the years when the accumulating pressures became impossible to absorb quietly.

Editorial Board

Filed from Kampala and Kumi District. The rains that were late are now simply not coming.

Summary

Grace is thirty-one. She came to Kampala from Kumi District at nineteen to work, to send money home, to build something. She works in the informal market at Nakasero, sells airtime and mobile money services, and remits a third of what she earns to her mother — Mama Amara — who farms a half-acre of matooke and beans outside Kumi town with Grace’s younger siblings. By 2030, the rains have become unreliable in ways that no accumulated knowledge can adapt to. The medicines for the children’s malaria are increasingly unavailable or unaffordable. The NGO that ran the village health post closed when its European funding was redirected. Grace’s mobile money business is the thread on which four lives hang.

She knows exactly how thin a thread it is.

race sends money every Friday. She has sent money every Friday for eleven years, through boom and through the grinding uncertainty of recent seasons. The amount changes — sometimes four hundred thousand shillings, sometimes less when the market has been slow, occasionally more when a customer from the office blocks nearby has paid a large bill. What does not change is the Friday. The Friday is the architecture of her mother’s week. Everything is planned around the Friday transfer.

GRACE — KAMPALA

Bwaise district, northern Kampala

The market is not what it was in 2023. The customers from the NGO offices disappeared first — the international organisations that used to send their staff past my stall for mobile money transfers back to their families have been closing or cutting staff since the Western funding dried up. Then the fuel prices went up after the trade disruptions, and the boda-boda drivers who used to top up their airtime three times a day now come once. I am still here. I am making less. I send what I can.

MAMA AMARA — THE VILLAGE

Kumi District, eastern Uganda

The first rains this season came three weeks late and then stopped for ten days when the matooke needed water most. The second planting of beans failed entirely. I have not told Grace the full situation because she will worry and send more than she can afford and go without herself. I told her the harvest was small. It was not small. There was almost nothing.

“Africa has always known how to survive scarcity. What it was not built to survive is scarcity arriving faster than knowledge can adapt — rains that come when the seed is already in the ground for the wrong season, heat that kills varieties that survived everything before.”

The medicines that stopped arriving

The collapse of global pharmaceutical supply chains, which began as a Western story about reshoring and trade fragmentation, arrived in Ugandan village health posts as a simpler and more brutal fact: the medicines were not there. Artemisinin-based combination therapies for malaria — the drugs that had transformed child mortality rates across East Africa since the mid-2000s — became intermittently unavailable as the logistics networks that supplied them fragmented. The WHO’s essential medicines list remained aspirational in the communities that needed it most. The gap between what the list said should be available and what was actually on the shelf at the sub-county health centre widened through the late 2020s into something that killed children who would not have died a decade earlier.

MAMA AMARA — THE VILLAGE

Kumi District, eastern Uganda

Amara’s youngest had fever for five days last month. I took her to the health centre. The nurse said the malaria test was positive but they had no ACT drugs. She gave me a paper to take to the private clinic in town. The private clinic wanted sixty thousand shillings. I called Grace. Grace sent the money. The child is better now. I do not know what I would have done if Grace had not answered the phone.

GRACE — KAMPALA

Bwaise district, northern Kampala

My cousin in Gulu told me last week that three children in his neighbourhood died of malaria in a single month. Three children. In 2015 this would have been a national emergency. There would have been an investigation, a ministerial statement, international attention. Now it is just what happens. The NGOs that used to fill the gaps between government provision and need have gone. The government provision was never enough on its own. The gap is now a hole.

[ HUMAN STORY — MEDICINE ACCESS ]

Space reserved: a nurse at a sub-county health centre; a mother who lost a child to a preventable illness; a pharmacist on supply chains.

The climate that nobody voted for

Uganda’s agricultural calendar has been shaped by two rainy seasons for as long as oral history reaches: the long rains from March to May, the short rains from October to December. The subsistence farming systems of the eastern and northern regions — the millet, the sorghum, the matooke plantations that form the nutritional foundation of rural life — were calibrated over generations to these rhythms. The rhythms are now unreliable in ways that agricultural extension services, themselves hollowed out by two decades of structural adjustment and NGO dependency, are not equipped to address. Farmers who have been farming the same land for forty years are finding that the knowledge they carry — knowledge that was once the most valuable thing they owned — no longer predicts what the soil and sky will do.

Lake Victoria, whose fisheries have supported millions of lakeshore families across Uganda, Kenya, and Tanzania for generations, has been showing the effects of warming water temperatures since the early 2020s. The Nile perch stocks, already depleted by decades of commercial overfishing, have been further disrupted by algal blooms fed by warmer temperatures and agricultural runoff. The fishing communities of the Ugandan lakeshore, who sent their young men to the lake because the lake had always provided, are finding it provides less. The young men are going to Kampala instead. Kampala is receiving them.

MAMA AMARA — THE VILLAGE

Kumi District, eastern Uganda

My brother John came back from the lake last year. Two years fishing and he came back with nothing — the catch is too small, the Chinese processing company that used to buy from the small fishermen closed its Kampala operation, and he could not make the boat payments. He is here now, on the land, helping me. Another mouth and another pair of hands is sometimes the same thing.

“The Chinese road came through in 2027. It brought trucks and it brought dust and it brought a mobile phone tower that works sometimes. It did not bring any of us jobs. The jobs went to the workers they brought with them.”

The girls who stopped going to school

In conditions of economic stress, girls’ education is the first sacrifice and the last investment to be restored. This is not a new pattern in Uganda or anywhere else. What is new in the late 2020s is the combination of pressures that makes the sacrifice more frequent and the restoration less likely. School fees that were manageable when harvests were predictable become impossible when two consecutive seasons fail. The informal social pressure to marry daughters early, which economic improvement had been gradually eroding, reasserts itself in communities where a bride price represents the only reliable cash income a family can anticipate.

MAMA AMARA — THE VILLAGE

Kumi District, eastern Uganda

My second daughter Stella left school in January. I did not want this. I kept her in school through the bad season, sold a goat, asked Grace for help. But the fees went up and the harvest went down and her father’s brother came with a proposal from a man in Soroti. Stella is sixteen. I told him no. I am telling you I told him no. But I do not know what I will tell him next time if the rains fail again.

GRACE — KAMPALA

Bwaise district, northern Kampala

I was lucky. My aunt in Kampala took me in when I was nineteen, let me sleep in her back room, helped me get my first market stall. Without her I would have been Stella — a village wife at twenty, children by twenty-two, no money of my own, no way out. I think about this every time I talk to my mother. I think about what I can do from here that will be enough. It is never quite enough.

[ HUMAN STORY — EDUCATION AND GIRLS ]

Space reserved: a secondary school teacher on dropout rates; a girl who came back to education; a mother on the choices she faced.

End of Movement One — The Weight of It

THE PROBABLE LIFE

MOVEMENT TWO

Holding the Line With What Remains

Two voices from Uganda, 2030 to 2034 — the years of damaged adaptation, of systems that partially held, of costs that were paid in ways that do not appear in any statistics.

Editorial Board

Filed from Kampala and Kumi District. The mobile money network went down for four days in March. It came back.

Summary

By 2032, Grace has survived the collapse of the mobile money platform during the coordinated East African cyber attack — three days without income, three days her mother went without the transfer that bought food. She has diversified: a small food stall alongside the mobile money, a savings group with six other market women, a careful relationship with a microfinance lender whose rates are punishing but whose money is available. Mama Amara has adapted her farming to shorter-cycle crops, joined a village water collective, lost one child to a preventable illness, and kept the rest in some form of education. This is not resilience in any celebratory sense. It is the cost of surviving a decade designed by forces that do not know they exist.

he cyber attack on the East African mobile money infrastructure in March 2031 lasted seventy-two hours. For Grace it was four days without income — the two days the system was down, and the two days after when the queues at her stall were so long with people trying to make the transfers they had missed that she could not process them fast enough to earn normally. She had told nobody about the attack before it happened, because she had not known. She told nobody about the four days of lost income after, because there was nobody to tell who could have helped.

GRACE — KAMPALA

Bwaise district, northern Kampala

My mother called on the second day. She did not say she needed money — she never says directly that she needs money, she talks around it, asks how I am, mentions in passing that the maize price has gone up. I knew what she was not saying. I borrowed from my savings group — three hundred thousand shillings from the pool we keep for emergencies — and went to the post office to send it by the old system, the slow one, the one that still worked. It arrived in two days. I paid the savings group back the following week. This is what a safety net looks like when there is no safety net.

MAMA AMARA — THE VILLAGE

Kumi District, eastern Uganda

When the phone money stopped coming I used what I had — a small amount I keep separate, rolled in cloth in the roof, that I do not tell anyone about. I have kept this since Grace started sending. It is for exactly this situation. I had enough for four days. On the fifth day I was beginning to think about asking the neighbour. Then it came through. I thanked God and I did not tell Grace how close it had been.

“The resilience of African women is real. What is also real is the cost of that resilience — the sleep lost, the food skipped, the calculations made at two in the morning about which child’s school fee comes first. This is not a strength to be celebrated. It is a tax on people who can least afford to pay it.”

The disease that came back

Malaria had been retreating in Uganda through the 2010s and early 2020s — not eliminated, never eliminated, but reduced to a manageable presence in the lives of families who had bed nets and access to treatment. By the late 2020s it was advancing again. The Anopheles mosquito range was expanding as temperatures rose, reaching altitudes that had previously been too cool for reliable breeding. The bed net distribution programmes, which had depended on a combination of government funding and international donor support, had been cut when donor countries turned their attention and their budgets inward. The treatment gap — the space between a positive malaria test and an available, affordable course of ACT drugs — was widening.

HIV/AIDS medication supply chains showed the same pattern in a slower and more politically complicated register. The antiretroviral programmes that had transformed HIV from a death sentence into a manageable chronic condition across sub-Saharan Africa depended on a global supply chain whose reliability was quietly assumed and whose fragility became visible only when it began to fail. Stockouts at district health centres, patients missing treatment cycles, the slow reemergence of drug-resistant strains in communities where adherence had been interrupted by supply failures — these were the downstream consequences of upstream decisions made in pharmaceutical boardrooms and trade ministries in Europe and America that never mentioned Uganda in their deliberations.

MAMA AMARA — THE VILLAGE

Kumi District, eastern Uganda

Akello, my neighbour, has been on ARVs since 2019. Last year she missed two months because the health centre ran out and the next delivery was delayed. She is sick again now — not the way she was sick before, not that bad, but sick in a way the nurse says is connected to the interruption. Twelve years of managing her health correctly and two months of a supply chain failure undoes some of it. Nobody will be held responsible. Nobody in the place where the decision was made even knows her name.

The money that flows one way

The remittance economy that connects Kampala to Uganda’s rural districts is one of the most significant financial flows in the country, larger in aggregate than most formal foreign aid programmes and considerably more efficient in its delivery. Grace is one of several million people in African cities performing the same function: absorbing urban economic shocks on behalf of rural families, smoothing consumption across the distance between the cash economy and the subsistence one. This function is invisible in economic statistics, unacknowledged in policy frameworks, and performed almost entirely by women. It is also the mechanism by which the costs of global economic disruption are distributed downward into the people least responsible for causing them and least resourced to absorb them.

GRACE — KAMPALA

Bwaise district, northern Kampala

Someone from a research organisation came to the market last year — a university student writing about mobile money and rural resilience. She was very polite. She asked good questions. I told her what I do, what my mother does with what I send, how the system works. She wrote it all down. She said it was remarkable, what women like me do. I wanted to ask her: remarkable to whom? To us it is just Thursday. It is just survival. There is nothing remarkable about it from the inside.

“The informal economy did not fail. It bent. It absorbed what the formal systems could not handle and charged the cost to the people doing the absorbing — their health, their sleep, their futures.”

What adaptation actually looks like

MAMA AMARA — THE VILLAGE

Kumi District, eastern Uganda

I have changed what I grow. Matooke needs more water than I can reliably get now, so I have reduced it and planted more orange-flesh sweet potato — it tolerates dry spells better and the children will eat it. I joined the women’s water collective — we share the pump maintenance cost and we have a rota for the dry season. I planted three moringa trees two years ago after the health worker told me about them — the leaves are nutritious and they grow even when other things do not. I am not thriving. I am adjusting. There is a difference.

GRACE — KAMPALA

Bwaise district, northern Kampala

I added a food stall to my business — I sell roasted maize and groundnuts in the afternoons when the mobile money is slow. The six women in my savings group — we meet every two weeks, we pool money, we rotate who takes the lump sum. It is not a bank. It is better than a bank because it does not require a credit score or collateral or a form in triplicate. We have kept each other solvent through two market disruptions and one illness and one death. The bank did not do that.

End of Movement Two — The Probable Life

WHAT COULD BE BUILT

MOVEMENT THREE

The World That Chose to See Them

Two voices from Uganda, 2034 to 2036 — not a solved world, but one where enough of the global decisions went differently to change what is possible on a half-acre in Kumi District and at a market stall in Bwaise.

Editorial Board

Filed from Kampala and Kumi District. The rains this season came on time. This was noted.

Summary

In this scenario, three decisions made between 2029 and 2032 changed the material conditions of Grace’s and Mama Amara’s lives: the African Medicines Supply Compact, which created a continental stockpile and distribution system for essential medicines outside the fragmented global chain; the East African Climate Adaptation Fund, which put actual resources into the hands of smallholder farmers through women’s cooperatives rather than agricultural ministry bureaucracies; and the Digital Infrastructure Resilience Framework, which made mobile money networks legally critical infrastructure with mandatory redundancy requirements. None of these decisions were made by people primarily thinking about Grace or Mama Amara. They were made by people who had finally looked at the data on where the costs of global disruption were actually landing and decided to do something about it.

race sent money on Friday as she always does. What is different in 2035 is that the amount she sends is slightly less than it used to be — not because she earns less, but because her mother needs less from her. Mama Amara has been part of the Kumi Women’s Agricultural Cooperative since 2032, when the East African Climate Adaptation Fund reached the district. The cooperative has access to drought-tolerant seed varieties, a shared irrigation pump, and an agronomist who visits monthly and whose advice is calibrated to what is actually happening with the rainfall rather than what the textbooks say should be happening. The matooke is still reduced. The total harvest is not.

GRACE — KAMPALA

Bwaise district, northern Kampala

My mother called last week not because she needed money but because she wanted to tell me that Stella — my sister who nearly stopped school three years ago — has been accepted into the nursing programme at Soroti Regional Referral Hospital. The bursary came from the women’s cooperative fund that the Climate Adaptation programme helped set up. I sat in my stall and I cried in a way that confused my customers. Stella will be the first person in our family to have a professional qualification. She will probably come to Kampala. Or she will stay in Soroti and treat the children who used to die of malaria because there were no nurses. Either way, this is what it looks like when something works.

MAMA AMARA — THE VILLAGE

Kumi District, eastern Uganda

The medicines are more reliable now. The sub-county health centre has not run out of ACT drugs in eighteen months — the new supply system comes through Kampala from the continental stockpile and arrives on schedule. Akello is stable. She has been stable for two years. She says she can think about the future now in a way she could not when she did not know if her medication would be there next month. Thinking about the future is not a small thing when you have been living week to week for a decade.

“Development is not what arrives from outside. It is what becomes possible when the obstacles built by outside forces are removed. Grace did not need to be given anything. She needed the system to stop taking things away.”

The supply chains that were redesigned

The African Medicines Supply Compact of 2030, negotiated through the African Union with technical support from the WHO and funding from a combination of member state contributions and a debt-for-health swap that converted a portion of Uganda’s external debt obligations into a health infrastructure investment, created something that had been technically possible for decades and politically impossible until the supply chain failures of the late 2020s made the cost of inaction visible enough to overcome the inertia. A continental warehouse system for essential medicines. A procurement mechanism that bought at scale, cutting the per-unit cost of ACT drugs by forty percent. A distribution network designed around the last mile — the sub-county health centre, the village health worker — rather than around the capital city logistics hubs that had always been the end of the road for international supply chains.

The compact was not a charity. It was an infrastructure investment, financed in part by the argument — which took longer to win than it should have — that the economic cost of child mortality and preventable adult illness in a continent of 1.5 billion people was larger than the cost of the medicines that prevented it. The argument had always been mathematically true. It became politically viable only when the countries that had previously supplied the medicines from outside stopped being reliable suppliers and the African Union was forced to build the alternative.

Necessity produced what generosity had not.

The land and the water

MAMA AMARA — THE VILLAGE

Kumi District, eastern Uganda

The irrigation pump the cooperative shares serves twelve families. We paid a portion of the cost — small, manageable, nothing like what a private pump would have cost — and the cooperative fund paid the rest. The agronomist told us which crops to plant for the changed rainfall. I planted orange sweet potato, cowpeas, and a drought-tolerant sorghum variety I had never grown before. The sorghum did well. I have enough stored now for six months. In thirty years of farming I have not had six months of stored food. I do not know what to do with the feeling this gives me. It is something close to safety. I had forgotten what safety felt like.

GRACE — KAMPALA

Bwaise district, northern Kampala

The mobile money network has not gone down since the Resilience Framework was implemented. I know this sounds like a small thing from the outside. From the inside it means I can plan. It means my mother can plan. It means the seventy-two hours of panic in March 2031 will not happen again — or if it does, there is a backup system, a redundancy, something that was built because someone finally counted the cost of it failing and decided that cost was too high. Someone counted the cost. This is all I ever wanted. For someone to count the cost of what happens to us when the systems fail.

What Grace knows that the economists do not

There is a version of the better scenario that is told as a story about institutional design, about supply chain architecture, about the political economy of African Union negotiations. That version is true and it matters. There is another version that is told in the twelve years of Friday transfers, the rolled cloth in the roof with the emergency money, the savings group meetings every two weeks, the moringa trees planted against an uncertain future, the phone call that Grace made when the fever lasted five days. That version is also true and it also matters — and it matters in a way the institutional version does not capture, because it is the version that was operating continuously, at full effort, regardless of whether the institutions were functional or not.

What changed in the better scenario is not that Grace and Mama Amara became stronger or more resilient or more remarkable. They were already all of these things to a degree that should have been embarrassing to the systems that required it of them. What changed is that the systems stopped requiring quite so much of them. The medicines were there. The seed was suited to the climate. The money moved reliably. The school bursary existed. Stella is going to be a nurse.

GRACE — KAMPALA

Bwaise district, northern Kampala

Someone asked me recently if I feel hopeful. I said: I feel like my mother has enough food stored for six months and my sister is going to nursing school and the mobile money has not failed in two years. Is that hope? I think it might be. It feels different from what I felt before, which was just the absence of immediate crisis. This feels like something I could build on. I have never built on anything before. I have only ever held things together.

MAMA AMARA — THE VILLAGE

Kumi District, eastern Uganda

In my grandmother’s time, this land fed this family and had surplus to sell. In my mother’s time, it fed the family and sometimes had surplus. In my time, it has fed the family — most years, most of the family, most of the time — and Grace has covered the rest. I want Stella’s children to grow up in a world where this land feeds the family and there is surplus again. I want them to know what surplus feels like. I think they might. I did not always think this.

THE LONG VIEW — COMPANION — END

Grace and Mama Amara are composite figures drawn from documented conditions in Uganda and East Africa. Their circumstances are specific; their situation is not unusual.

Throughout this article, discrimination simply means making decisions on the basis of group identity rather than individual merit.

The Core Problem: Discrimination Misallocates Talent

A market economy works because it tends to put people where they are most productive. Discrimination jams that mechanism. It assigns opportunities based on group identity rather than ability, which means the most capable person for a job, a promotion, or a loan is routinely passed over for someone less capable who happens to belong to the favoured group. Multiplied across millions of decisions, this is not a rounding error. It is a structural drag on productivity.

The clearest evidence comes from a landmark study by economists Chang-Tai Hsieh, Erik Hurst, Charles Jones, and Peter Klenow. In 1960, 94 per cent of American doctors and lawyers were white men; by 2010 that share had fallen to 62 per cent. Since there is no credible evidence that the aptitude for medicine or law is distributed differently across race or sex, the earlier figure meant a large pool of capable women and Black men had been steered away from the work they were best suited to. As artificial barriers fell and talent flowed toward its best use, the researchers estimate this improved allocation accounted for 20 to 40 per cent of the growth in U.S. output per worker between 1960 and 2010. A substantial share of half a century of American prosperity, in other words, came simply from discriminating less. Every remaining barrier pushes that engine into reverse: discrimination is not a private preference but a tax on the whole economy’s ability to use its people well.

Nepotism: When Loyalty Beats Competence

Nepotism is discrimination in favour of kin, and it is often defended as harmless or even prudent. History and evidence both point the other way. The most instructive case is what happens when societies abandon it. For centuries, appointments in the British state were handed out through patronage and family connection until the Northcote-Trevelyan reforms of the 1850s replaced them with competitive examination; the modern, capable civil service that resulted became a model copied around the world. Imperial China had grasped the same principle more than a millennium earlier, using written examinations to recruit officials on merit rather than birth. The lesson is consistent: institutions grow stronger the moment they stop reserving the best positions for relatives.

The contemporary evidence agrees. A study of more than 800 employees across 186 small and medium-sized Italian family firms found that favouritism toward relatives depressed the motivation of non-family staff, with measurable costs to productivity and stability. The damage runs along two channels. Directly, a less competent person occupies a role a better candidate could have filled, so decisions get worse. Indirectly and more corrosively, everyone else learns the rules are no longer fair and stops trying to win. Research links nepotism to lower innovation, higher turnover, and weaker long-run performance for exactly this reason: talented outsiders will not invest their best years where the top rungs are reserved by birth. Nepotism trades a durable meritocracy for a brittle loyalty, and the trade rarely pays.

Racism: A Sixteen-Trillion-Dollar Mistake

Racial discrimination is frequently imagined as a set of isolated bad acts. In aggregate, it behaves like a chronic, compounding cost. In a 2020 analysis, Citigroup estimated that discrimination against Black Americans had drained roughly 16 trillion dollars from the U.S. economy over the previous two decades, driven largely by lending that denied capital to Black entrepreneurs, along with gaps in wages, housing, and education. The mechanism behind it matters more than the headline figure.

That mechanism is visible in controlled experiments. In a now-classic field study, economists Marianne Bertrand and Sendhil Mullainathan sent nearly 5,000 fictitious resumes to real job openings, randomly assigning names that sounded either distinctively white or distinctively Black. Identical resumes with names like Emily and Greg received about 50 per cent more callbacks than those with names like Lakisha and Jamal, and a Black applicant needed roughly eight additional years of experience to draw the same interest. The employers making these calls were not only being unfair; they were systematically screening out qualified candidates and making worse hiring decisions than a name-blind process would have produced. Racism is a filter that removes ability while preserving the illusion of standards. Run that filter across an entire labour market, and the trillions follow.

Misogyny: Leaving Half the Talent on the Table

Discrimination against women is, mathematically, the largest single case of talent left unused, because women are half the population. The macroeconomic stakes are correspondingly enormous. The McKinsey Global Institute estimated that advancing women’s equality could add 12 trillion dollars to global GDP by 2025, and that a full-parity scenario could add as much as 28 trillion, roughly a quarter of global output. Within the United States, McKinsey found that closing gender gaps in workforce participation and hours could add some 2.1 trillion dollars to annual GDP. These are not soft benefits; they are the direct result of letting capable people contribute fully.

At the level of the individual firm, the pattern repeats. McKinsey’s multi-year ‘Diversity Wins’ research found that companies in the top quartile for gender diversity on executive teams were significantly more likely to post above-average profitability, while those in the bottom quartile were measurably more likely to underperform. These findings do not prove that diversity itself causes higher profits; successful organizations may simply be better at attracting and promoting talent in the first place. Either way, systematically excluding capable women is unlikely to improve performance, and the logic is intuitive: homogeneous leadership tends toward narrower thinking and shared blind spots, whereas teams drawing on a wider range of experience make more robust decisions. Excluding women does not protect performance. It caps it.

Misandry: Why the Case Cuts Both Ways

If the principle is that talent should never be wasted, it cannot depend on who is being overlooked. A consistent standard must apply consistently: if discrimination is counterproductive because it severs reward from merit, then it is counterproductive whoever the target is, including men. Dismissing men’s difficulties, or treating disadvantage as a problem only when it runs in one direction, produces the same waste from the opposite side.

The clearest contemporary example is education. In 1972, men earned 56 percent of American bachelor’s degrees; by 2019 women earned about 58 percent, a fifteen-point swing. When college enrollment fell in 2020, the decline was seven times larger for men than for women. As the scholar Richard Reeves has documented, many interventions designed to lift academic achievement help girls while doing little for boys, and the male shortfall in schooling feeds into weaker employment and health outcomes later. Reeves frames this as non-zero-sum, and rightly: recognizing that boys and men are struggling takes nothing away from the real and continuing gains women have fought for. A society that decides one group’s problems do not count forfeits that group’s talent just as surely as old-style prejudice did. Fairness is not a finite resource to be rationed between the sexes; treating it that way guarantees that someone’s potential goes to waste.

Religious Discrimination: Driving Talent and Capital Away

History offers a stark demonstration. When France revoked the Edict of Nantes in 1685 and stripped its Protestant Huguenots of religious freedom, some 200,000 of them fled, taking their skills in weaving, watchmaking, and finance to England, the Netherlands, and Prussia. France exported a generation of its most productive artisans and entrepreneurs to its rivals, who prospered accordingly. Persecution did not purify the nation; it impoverished it while enriching the competition.

Modern data confirm the pattern at scale. Analysing GDP growth across 173 countries while controlling for more than twenty economic, political, and demographic factors, researcher Brian Grim and colleagues found that government restrictions and social hostilities involving religion were associated with lower economic growth, while religious freedom correlated positively with ten of the twelve pillars of global competitiveness. The mechanism is intuitive: where people are constrained because of their faith, ambitious workers take their talents and their capital elsewhere, investment retreats from the instability, and communities fracture along sectarian lines instead of cooperating in commerce. Openness to belief is, among other things, a competitiveness strategy.

Beyond Economics: Trust, Belonging, and the Fabric of Society

The costs of discrimination are not only financial. They corrode the social trust and sense of belonging that hold a society together, and those are harder to rebuild than any balance sheet. When people believe outcomes are decided by identity rather than effort, they stop trusting institutions and one another, and a low-trust society is a high-friction one where everything from contracts to cooperation becomes more costly.

Belonging is not a luxury; it is a precondition for performance. Psychologists Gregory Walton and Geoffrey Cohen showed that when students carry ‘belonging uncertainty’, a nagging doubt about whether people like them fit in, their achievement suffers. A brief intervention that reframed early social struggles as normal and temporary rather than as evidence of not belonging raised African-American students’ grades over three years and cut the racial achievement gap roughly in half. The same dynamic operates in workplaces: employees who suspect they are judged by their group rather than their work underperform relative to their actual ability, a pattern researchers call stereotype threat. Discrimination thus creates a self-fulfilling prophecy, manufacturing the very underperformance it claims to predict.

This is where discrimination breeds uncertainty. In a fair system, people can plan: study hard, build skills, and expect roughly commensurate reward. In a discriminatory one, the rules are hidden and the outcomes arbitrary, so effort feels like a gamble, and that discourages exactly the long-term investment, in education, enterprise, and community, on which prosperity depends. Cohesion and unity are not sentimental add-ons. They are the substrate that lets a diverse society pool its talents instead of splintering into wary factions, each certain the game is rigged against it.

If It Is So Costly, Why Does It Persist?

A reasonable objection follows from everything above: if discrimination is this expensive, why has it survived for so long? Behavioural science offers an answer. Humans evolved in small groups where familiarity was a reliable proxy for trustworthiness, and favouring one’s own kin and tribe was, for most of our history, a sound survival strategy. Modern economies demand the opposite instinct, the ability to evaluate strangers by evidence rather than by identity. Our institutions have raced ahead of our psychology, which still whispers that the familiar face is the safer bet. Discrimination persists not because it is efficient but because it is intuitive, and the pull of that intuition is precisely why fair institutions, blind hiring, transparent criteria, and competitive examination have to be deliberately built. Left to instinct, we drift back toward the tribe. The cost of discrimination is real; it simply arrives later and more diffusely than the immediate, emotional comfort of preferring one’s own.

Conclusion: The Economics of Fairness

The forms of discrimination examined here look different on the surface. Nepotism wears the face of family loyalty, racism and misogyny of prejudice, misandry of neglect, religious discrimination of tribal suspicion. Underneath, they are the same mistake: each replaces the question ‘who is best?’ with the question ‘who is one of us?’ Every time that substitution is made, a society uses its people slightly worse than it could, and those small failures compound into misallocated talent, poorer decisions, eroded trust, and trillions in forgone prosperity.

The encouraging implication is that fairness is not charity a society extends at a cost to itself. It is enlightened self-interest, and the historical record shows meritocracy generating enormous gains whenever it expands. Every society must decide how opportunities are allocated. It can reward belonging, or it can reward ability. History suggests that the societies which prosper most are those that increasingly choose the latter, becoming wealthier not when they choose the right groups, but when they stop choosing groups altogether.

References

Hsieh, C.-T., Hurst, E., Jones, C. I., & Klenow, P. J. (2019). The Allocation of Talent and U.S. Economic Growth. Econometrica, 87(5). https://web.stanford.edu/~chadj/HHJK.pdf

Cirillo, A., et al. (via University of Bologna). Nepotism and Favouritism Reduce Productivity in Family Businesses. UniBO Magazine. https://magazine.unibo.it/en/articles/nepotism-and-favouritism-reduce-productivity-in-family-businesses

Citigroup / Citi GPS (2020). Closing the Racial Inequality Gaps. https://www.citigroup.com/global/insights/closing-the-racial-inequality-gaps-20200922 (summary: https://www.cnbc.com/2020/09/24/racial-inequality-cost-the-economy-16-trillion-over-the-last-two-decades-citi-finds.html)

Bertrand, M., & Mullainathan, S. (2004). Are Emily and Greg More Employable than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination. American Economic Review / NBER Working Paper 9873. https://www.nber.org/papers/w9873

McKinsey Global Institute (2015). The Power of Parity: How Advancing Women’s Equality Can Add $12 Trillion to Global Growth. https://www.mckinsey.com/featured-insights/employment-and-growth/how-advancing-womens-equality-can-add-12-trillion-to-global-growth

McKinsey & Company (2020). Diversity Wins: How Inclusion Matters. https://www.mckinsey.com/featured-insights/diversity-and-inclusion/diversity-wins-how-inclusion-matters

Reeves, R. V. (2022). Of Boys and Men. See also Harvard Magazine (2025), Why Men Are Falling Behind in Education, Employment, and Health. https://www.harvardmagazine.com/2025/05/harvard-men-gender-gap-education-employment

Grim, B. J., Clark, G., & Snyder, R. E. (2014). Is Religious Freedom Good for Business? A Conceptual and Empirical Analysis. Interdisciplinary Journal of Research on Religion. https://religiousfreedomandbusiness.org/2/post/2014/05/religious-freedom-linked-to-economic-growth-finds-global-study.html

Walton, G. M., & Cohen, G. L. (2011). A Brief Social-Belonging Intervention Improves Academic and Health Outcomes of Minority Students. Science, 331(6023). Summary via marycmurphy.com: Stereotype Threat in Organizations. https://marycmurphy.com/wp-content/uploads/2025/02/Stereotype-threat-in-organizations-Implications-for-equity-and-performance.pdf