Safe Is a Place, Until It Isn't

A country can run perfectly smoothly and be hollowing out underneath. Both things can be true at once. An essay on what “safe” actually means — from Venice and the Dutch Republic to the modern state — and on the stranger possibility the age of AI raises: not that the system crashes, but that it stays calm, comfortable, and steady forever, with most people looked after but no longer needed.

A country can run perfectly smoothly and be hollowing out underneath. Both things can be true at once — and the gap between them is what the age of AI is really about.

Picture a country running smoothly. Nobody is rioting. The shops are full. People get their money and watch their shows. Everything looks fine.

Now look again. Hardly anyone feels needed. Fewer people are having children. Nobody trusts the government, joins anything, or feels part of something bigger.

Same country. Two different stories. One person calls it stable. The other calls it hollow. The gap between those two words is what the age of AI is really about.

History’s lesson

Venice. The Dutch Republic. Singapore. People say they got rich because they were “safe” — clear rules, protected property, no nasty surprises. True enough. Money goes where the rules hold still.

The detail is worth holding for a moment, because it is more specific than the slogan. Venice dominated Mediterranean trade for four centuries not because Venetians were cleverer than anyone else, but because a merchant’s contract was reliably enforced and the gold ducat held its weight and purity for roughly five hundred years — a trader from Cairo or Bruges could plan around it. The Dutch Republic honoured its public debts and built the machinery of modern finance: a central exchange, tradable shares, marine insurance, a reliable bank. Singapore, a small port with no resources at independence in 1965, decided that what it could not get from geography it would manufacture as trust — clean courts, enforceable contracts, a currency that behaved as advertised.

Spain is the counter-example. It hauled mountains of silver out of the Americas and still went broke, over and over, because its rulers were unpredictable and might just take your wealth. Resources without reliability did not compound. Reliability without resources did.

And here is the part the slogan misses. Safety does not just sit there. It compounds. A stable country attracts talent and money. That talent and money pay for the courts, the roads, the institutions. Those institutions make the country more stable, which attracts more talent and more money. It is a flywheel. The uncomfortable part is that the same flywheel runs in reverse. A shaky country loses its talent and its money, which starves the institutions that made it stable, which makes it shakier, which speeds up the exit. Somewhere is always winning this. Somewhere is always losing it.

But “safe” had limits. The safety was for the people at the top — bought with lucky geography, naval power, and plenty of violence aimed at people far away. So here is the lesson: a place can be rock-solid at the top and hollow underneath, and still call itself a success. The people at the top write the story.

People follow safety. Money follows people.

Safe place? People move toward it. Unsafe place? They leave. Oldest pattern there is.

Capable people have always drifted toward wherever life was most predictable. Venice and Amsterdam did not just build their success — they imported it, taking in talented people fleeing somewhere worse. When Spain and Portugal expelled their Jewish populations, and when France later pushed out its Huguenot artisans and merchants, those refugees did not arrive empty-handed. They brought skills, trading networks, and capital. Amsterdam compounded. The cities that expelled them did not. Intolerance, in that light, was a capital export policy. Safety is a magnet, and it pulls in two things at once: talent and money.

It still happens. When a country goes bad, the first out the door are the most capable — doctors, engineers, business owners, investors. They have got the skills, the savings, the contacts to start over. Their money leaves with them. The safe place gets richer and sharper. The place they left gets poorer and emptier.

Now the uncomfortable part. Look at who stays. Not random. The people who can leave are the ones with options — the steady, capable middle. Plenty of thoughtful people stay too, out of loyalty or roots. But the calm center gets thinned out. What is left tilts poorer, angrier, quicker to boil.

And there is a sharper version of this that history keeps repeating. The dangerous group is not the poorest. France in 1789 was among the richest countries in Europe, not the poorest, and its revolution was not led by the starving — it was led by the frustrated bourgeoisie, the lawyers and merchants and mid-level professionals whose path upward was blocked by an aristocracy above them. Revolutions are rarely the work of the destitute. They are the work of the capable and the sidelined — people with enough education to see the fence clearly and no longer any reason to respect it. Hold on to that. It is going to matter when we get to what AI does.

And losing those people changes how the place gets run. A government staring at a frustrated, volatile population cannot govern gently. It clamps down — more police, more surveillance, more controlling what people see, more handouts to keep things quiet. Losing capable people does not leave an empty space. It leaves a heavier state.

So the danger is not one country quietly hollowing out. It is a split. The world sorts itself. Safe places pull in talent and money and stay open. The places left behind lose their problem-solvers and clamp down harder just to hold together. And because the flywheel runs both ways, once that sorting starts, it feeds itself.

Will AI take all the jobs?

The hopeful answer: every new technology was supposed to kill jobs, and new jobs always appeared instead. A farmer 500 years ago could not have dreamed up “software project manager.” So maybe AI creates work we cannot picture yet.

But it is worth being precise about what the earlier machines actually did, because it is not quite what the hopeful answer assumes. The Industrial Revolution replaced muscle. It took the heavy lifting — the field, the loom, the load — and in doing so it built a whole new middle: logistics, management, clerks, engineers, the entire apparatus of people coordinating complex work. The ladder lost rungs at the bottom and grew new ones in the middle. The thinking was left to us.

AI is different in a way that matters. It does not come for muscle. It comes for the thinking itself — the analysis, the judgement, the coordinating of complexity that the last revolution’s middle class was built to do. Is there a higher rung above that one, the way clerical work was a higher rung above the plough? Maybe. Maybe not. Nobody knows, and anyone who tells you they do is guessing.

So both things can be true at once. New jobs may appear over thirty years — and the next ten may still be brutal for millions of people mid-career, who cannot wait three decades for it to work out. “It worked out before” is an average. Averages do not tell you how much it hurt, or to whom.

And notice who is in the firing line. The cognitive middle — educated, articulate, mid-career, used to being useful. That is the same group, from two sections ago, that history says does not riot when it is poor but does turn when it is sidelined. AI does not just threaten jobs. It manufactures, at scale, exactly the demographic that has unmade settled arrangements before.

Why the government won’t fall apart

So is that the revolution, then? Probably not — and the reason is worth being clear about.

Ancient Rome found out about an angry mob when the mob was already in the square. A modern government is different. It sees the anger building online weeks ahead. It wires money straight to the towns that are hurting. It keeps people entertained almost for free, all day. And since everyone is buried in their own feed, the angry people never even find each other.

So forget the dramatic uprising — crowds in the streets, the system toppling. It probably is not coming. The modern state is very, very good at keeping the peace. And the clamping-down we just described? That is not the state failing. That is the state working.

But here is the catch, and it is the heart of this whole essay. The same tools that keep the peace are the tools that hollow things out. People who cannot unite to be angry cannot unite to fix anything either. People kept endlessly entertained stop showing up for real life — the groups, the communities, the families. People simply handed money slowly become managed instead of needed. None of it looks like a crisis. Everything looks like it is working. That is exactly the point. The country stays calm because it is emptying out.

There is one more thing the calm version leaves out. We are quietly assuming the winners can wall themselves off — that the talent and the machines can sit safe behind the fence while the rest drifts. A Venetian merchant could lock his gold in a vault. AI cannot be locked anywhere. It runs on physical things with fixed addresses: sovereign-scale energy grids, water for cooling, semiconductor supply chains stretched across oceans, data centres you can point to on a map. The owners may sit in the safe country. The machine they own is exposed to the weather, the grid, and the politics of the actual world. The fence, in the AI age, is harder to build than it has ever been.

Both sides end up in the same place

Here is the twist. The optimist and the pessimist now describe the exact same future: a small group doing the real work, a big group looked after but not needed, fewer kids, less trust, less taking part.

The optimist looks at it and says: stable. Nothing broke. Everyone is fed.

The pessimist looks at the very same thing and says: hollow. Nobody is needed anymore.

Rome could hand out free grain and still feel strong, because its people still had real roles — citizens, soldiers, part of the story. The grain dole, the annona, fed hundreds of thousands. But it was the floor of what a citizen was, not the whole of it. A modern country can hand out the money and leave out the rest. Same handout. Opposite feeling. Sustenance without standing is a thinner bargain than Rome ever offered.

It is worth saying plainly what a society actually runs on, because it is not just money. It runs on consent — on enough people broadly agreeing the arrangement is fair enough to go along with. And consent rests on three things: that the system is legitimate, that you can take part in it, and that you have some honest reason to expect tomorrow to be better than today. A society can absorb a startling amount of inequality while those three hold. What it cannot survive is losing all three at once. A population that is fed but has concluded the game is rigged, that it has been quietly turned into pensioners in its own country — that population has not been made safe. It has been switched off.

That is the contradiction sitting inside the winning country. It succeeds by being attractive to talent and machines: orderly, predictable, low-friction, concentrated. And succeeding in exactly that way hollows out its own middle and wears through the consent that “safe” was always built on. Carried all the way through, the strategy eats its own foundation.

So what’s the honest answer?

AI probably will not spark a revolution. The modern state is too good at smoothing things over. But revolution was never the real danger. The real danger is the hollowing — a society quietly losing the people, the trust, and the purpose that hold it up — and the splitting, as the safe places and the struggling ones drift further apart and the flywheel pulls them further still.

So the countries that are genuinely safe over the next few decades may not be the ones that build the highest fence. They are more likely to be the ones that never let the gap open far enough to need one — that keep enough of their people as participants rather than pensioners, as stakeholders rather than recipients, so the underlying agreement holds. Safe is not a wall you finish building. It is a consent you keep earning.

Is a calm, comfortable country where most people are not really needed a success or a slow failure? Facts cannot settle that. It comes down to one question: is being looked after a good enough replacement for being needed?

Two honest warnings, so this does not oversell itself. One: the whole worried case depends on AI hitting a ceiling humans cannot climb past — and we do not know that it will. There may be a higher rung. Two: “stable but hollow” is nearly impossible to prove wrong, and any idea you cannot prove wrong should be held loosely. The solid core is small but real: a big uprising is unlikely, talent really is draining toward safer places right now, and “nothing collapsed” is not the same as “everything’s fine.”

Which brings us back to the title. “Safe” is a place — a setup, a way of life that holds. We usually hear “until it isn’t” as a warning that it all comes crashing down. But here is the stranger, quieter possibility: it does not crash. It stays exactly as it is — calm, steady, forever.

And that’s the problem.