THE YEAR

2025: the strongman's year, the checks that bent, and the machine that runs the economy now

A returned president governed by decree and bent the institutions that once restrained him. Wars ended and began. And the AI build-out grew so vast it became the pillar the economy leans on. The year the order built after 1945 came apart in earnest.

The year that ends this fortnight was the year the postwar order came apart in earnest. A returned American president governed from the first hour by decree, pardoned an insurrection, deployed troops to his own cities, defied the courts, fired the officials who produced inconvenient facts, and bent, one by one, the institutions built to restrain executive power, testing whether any check could still say no and make it stick, and finding, too often, that it could not. Abroad, he dismantled the alliances of 80 years, courted the aggressor in Ukraine and punished the invaded, treated allies as marks and adversaries as partners, launched a trade war against the world, and pushed a proud India toward the very powers America had spent decades drawing it away from. The order that the United States built after 1945, of alliances, institutions, free trade and the rule of law, an order this paper has watched fray for three years, came apart this year not gradually but in great deliberate tearing pieces.

It was a year of wars ending and beginning: the Gaza war, after two years and a famine, reached a ceasefire and the last hostages came home; Israel and Iran fought a 12-day war and America bombed Iran's nuclear sites; India and Pakistan clashed and stepped back; Ukraine ground on toward a settlement that rewarded its invader. And it was, above all, the year artificial intelligence stopped being merely a technology and became infrastructure and economy both: the half-trillion-dollar data-centre bets, the cheap Chinese model that humbled Silicon Valley for a day, the synthetic video that ended "seeing is believing," the deregulated race to build faster than anyone could govern, and, by year's end, the startling realisation that the AI build-out had become the pillar a striking share of the entire economy now leans upon. The machine the year began arguing about had become, by the year's end, the thing the economy runs on, a concentration of hope and capital and risk of a scale the world has never before staked on a single unproven bet.

THE WORLD

A hard year closes with the usual cascade

The year's final fortnight brought the now-customary run of hard news: the Venezuela confrontation building toward an uncertain climax, the Ukraine settlement grinding toward a bitter conclusion, the immigration crackdown deepening in American cities, and the markets closing a third straight strong year, propelled once more by a handful of AI-driven giants, on a foundation that a growing chorus of sceptics warned was more precarious than its heights suggested. The rituals of the year's end, the honours, the retrospectives, the resolutions, proceeded against a backdrop of a world that had spent the year learning how much of what it took for permanent was, in fact, a choice that could be unmade.

IN BRIEF

The machines advance; the year's dead honoured; a fragile hope held

Artificial intelligence closed the year at full sprint, the models more capable than ever, the spending more vast, the questions of profit and power and governance more urgent and more unanswered. The year's obituaries honoured, among many, a pope who preached mercy in an age of strength and a partner whose empty chair had marked the passing of an era of investing. And in Gaza, the fragile ceasefire held, the hardest and least certain work, of turning a stopping into a peace, only just beginning as the year that saw the war finally end drew to its close.

The Column

What the year of the strongman taught us

The Dispatch closes its third year. The news was the darkest we have chronicled, and the patterns beneath it were the ones we have traced from the beginning, now arriving at a scale we had hoped not to see.

By The Dispatch

Three years ago this paper began with a wager: that setting each fortnight's headline news beside the unglamorous work of building the systems a modern world depends on would reveal patterns the news alone misses. A third year confirms the wager and darkens it, because the patterns we have traced from the beginning, the fragility of concentration, the failure to heed warnings, the erosion of the checks that restrain power, the crisis of shared truth, all arrived this year at a scale we had documented in miniature and hoped never to see in full. So let us name, as we close the volume, what the year of the strongman taught, because the lessons are the ones this paper was founded to teach, now written large.

The first is that the checks that restrain power are more fragile than we believed, and that their erosion, once begun, accelerates. We have written for three years about "the layer that can say no," the court, the board, the auditor, the norm, the treaty, the independent statistician, the institution that can overrule the powerful and make it stick. This year we watched those layers tested as never before, and bent, and in too many cases broken: the court order defied, the injunction curtailed, the statistician fired, the alliance abandoned, the norm shattered. The checks held, mostly, in the places where enough people inside them chose to make them hold, South Korea removing its coup-plotting president, the voters rebuking the powerful, the occasional judge who would not yield. But the direction was unmistakable, and the lesson is one democracies forget at their peril: the checks on power are not self-sustaining machines; they are conventions held up by the willingness of people to honour them, and when a sufficiently determined power declines to honour them, they hold only if enough people, at real cost, insist that they must.

The second lesson is the one about concentration and fragility that this paper was practically founded on, arriving this year at the scale of the whole economy. We watched a cloud region take down the internet, a software update crash the world, cheap drones destroy irreplaceable bombers, a single strait threaten global trade, and, by year's end, an entire national economy grow dependent on a single unproven bet on AI. The pattern is always the same: concentration is efficient, so everyone concentrates, and the aggregate result is a world of single points of failure, invisible because they work almost all the time, catastrophic when they fail. And the AI build-out is concentrating everything further, compute, capital, economic growth, hope, into fewer hands and fewer bets, building the largest single point of failure the world has yet assembled, and calling it progress.

And the third lesson is the oldest, the one about warnings written and unread, and it broke our hearts this year more than any other, because this year it took children. The dam, the bridge, the bank, the airspace, and this year the flood that drowned the girls at the Texas camp in a place everyone knew would flood, the fire that climbed the bamboo chimney the rules forbade, the famine forecast for months and named only when it was too late to be a warning. Over and over, the catastrophe was foreseen, documented, warned of, and the boring, unglamorous, inexpensive thing that would have prevented it was deferred because its cost was present and its benefit was future, and human institutions discount the future toward zero when the present is loud. It is the most human failure and the most lethal, and we have chronicled it for three years, and the species does not learn it, and this year it took the children, and we do not have a solution, only the same weary plea: find the warning already written about your own vulnerable place, and act on it, before it becomes the thing you grieve.

We do not know what next year holds, though the wars unresolved, the checks weakened, the economy staked on a single bet, and the technology accelerating past our capacity to govern it suggest it will not be gentle. We know only that we will keep doing this, fortnight by fortnight, the headline and the work beneath it, on the wager that seeing the pattern is worth something even when the world will not act on it. Thank you for reading the third year of it. The warnings for the year to come are already written, as they always are. As ever, we will keep reading them out loud.

The Ledger

The year AI became the economy's load-bearing pillar. From DeepSeek's cheap challenge to Sora 2's synthetic flood to the Action Plan's deregulated race, the technology advanced relentlessly while its governance eroded and its economics grew more circular and more concentrated. 2026's question: whether the great bet pays off, or is called.

Data centers & power

Promoted this year from the economy's footnote to its foundation. The build-out now props up national growth, which means the power question and the profit question beneath it carry the weight of everything. The wire under the boom is now the wire under the economy.

Rates

The year of easing: three Fed cuts brought relief to a market and a housing sector battered by the prior tightening, even as tariffs and deficits kept an inflationary floor beneath it all. The soft landing held, propped, ominously, by the AI investment itself.

Real estate

US 30-year mortgage closes 2025 near 6.15 percent, the year's low, the one consistent comfort in a turbulent year, driven, as all year, by an economy cooling beneath a market climbing.

India tech

India's hard year: courted, then punished, by America; pushed toward China and Russia; its technology sector battered by tariffs and the H-1B wall, and forced toward the diversification and self-reliance its instincts had always favoured. The strategic autonomy Delhi never abandoned looks, at year's end, vindicated.

What we called wrong

The volume's honest ledger: we were slow, early in the year, to grasp how fully the checks would bend; we underestimated, perhaps, the economy's resilience, though we now see how much of it rests on a single bet. We log our errors as we log them all, and carry the humility into a fourth year that will, we suspect, test every lesson this paper has learned.

The Back Page

The warnings for 2026

It is the end of another year, and this paper closes it as it closed the last, with the one resolution it believes in. Somewhere right now there is a report on a levee, a filing on a bank, a study of a fault line, a warning about a system, a risk register with a red line that everyone has learned to scroll past. Somewhere an engineer or a scientist or an auditor has done the patient, thankless work of seeing the next catastrophe clearly and writing it down, and their document sits, unfunded, in an inbox, because the thing it asks for is expensive and invisible and the cost of ignoring it will not arrive until later, on someone else's watch. We have spent three years learning that these documents almost always exist before the disaster, and almost never get read in time, and this year we learned, at Camp Mystic and in a Hong Kong tower and in a Gaza starved past the point of warning, that the price of not reading them is sometimes children. So here is our resolution for the fourth year, offered to anyone who runs anything: find the warning that is already written about your own operation, your own vulnerable place, the thing you have been scrolling past, and read it, and do the dull, expensive, life-saving thing it asks. It is cheaper than the emergency. It is always cheaper than the emergency. It was cheaper than the emergency this year, too, and we did not do it, and the emergencies came, as they always do, as they were warned they would. Happy new year, from all of us at the Dispatch. May you, this year, sound your sirens before the fire. Someone has already written down where the fire will start. Go and read it, before it burns.