Vol. III, No. 76
Covering 3 November - 16 November 2025
Monday, November 17, 2025
Late edition · A Relentless publication
The Relentless Dispatch
All the fortnight that mattered, in technology and in the world, read next to what we were building at the time.
AMERICA

The voters deliver a rebuke, and a 33-year-old socialist wins New York

In the first big elections since Trump's return, the backlash arrived: a Democratic sweep, capped by the victory of Zohran Mamdani as mayor of New York, on a message of affordability that the party's establishment had forgotten how to deliver.

American voters delivered the sharpest rebuke yet to the Trump administration on November 4, in the first major elections since his return to power. Democrats swept the governorships of Virginia and New Jersey by comfortable margins, and, most strikingly, Zohran Mamdani, the 34-year-old democratic socialist who had stunned the establishment in the summer primary, won the mayoralty of New York City, becoming the largest city's youngest mayor in generations and its first Muslim one, on a relentless, disciplined message about the sheer cost of living, rent, groceries, childcare, transit. The results, in an off-year that historically punishes the party in power, were a clear sign that the coalition Trump assembled in 2024 was fraying, that the affordability crisis he had promised to fix had not eased, and that a Democratic party lost in the wilderness since its defeat might have found, in the politics of the concrete and the affordable, a way back.

Mamdani's victory in particular carried a lesson beyond one city. This paper wrote, after the 2024 defeat, of a Democratic party that had lost touch with the concrete concerns of struggling people, that had asked voters to prioritise abstract threats over the price of eggs and lost. Mamdani won by doing the opposite with almost monomaniacal discipline: talking about nothing but the concrete cost of ordinary life, in the language ordinary people speak, against a machine that had forgotten how. Whether his ambitious and contested programme can survive contact with the reality of governing a vast city, and a hostile federal government that has already threatened to move against him, is the open question. But the political lesson of his rise was unmistakable, and both parties should read it: the voters, exhausted and squeezed, are looking for whoever will address the thing that is actually crushing them, and they will reward, across the ideological spectrum, the candidate who speaks to the concrete over the one who lectures them on the abstract.

WASHINGTON

The longest shutdown in history ends, having proved nothing

The longest government shutdown in American history ended after 43 days, when enough Senate Democrats agreed to a deal to reopen the government in exchange for a promised future vote on the healthcare subsidies that had been the sticking point, a resolution that left neither side satisfied and the underlying dispute unresolved. The shutdown had furloughed hundreds of thousands of workers, disrupted air travel and food assistance, and cost the economy billions, all to arrive at a compromise not obviously different from what had been available six weeks earlier, a now-ritual demonstration of a governing system that inflicts enormous self-harm to accomplish nothing, the manufactured crisis this paper has written about before, run to its longest and most pointless conclusion.

IN BRIEF

The files come out; a bubble questioned; a broadcaster in crisis

The long-running Epstein affair convulsed Washington anew as emails released this fortnight showed the disgraced financier's references to the president, and the House moved, with even Republican support, toward forcing the release of the full investigative files, a rare defiance of the president by his own party. Fresh doubts gripped the AI market as prominent investors, including the famed short-seller Michael Burry and the Japanese giant SoftBank, moved against the AI darlings, betting the boom had become a bubble. And the BBC was plunged into crisis, its director-general resigning over accusations that a documentary had deceptively edited a Trump speech, a scandal the president gleefully amplified.

The Column

The circular river of money

The AI boom is being questioned again, and this time the worry is not just the valuations but the strange, circular way the money flows. I want to explain what that means, because circular financing is one of the oldest warning signs there is.

By Rebecca Cho

The doubts about the AI boom that flared and faded a year ago returned with force this fortnight, as some of the market's most respected sceptics, the investor who famously foresaw the 2008 crisis among them, placed public bets against the AI darlings, and the concern this time was sharper and more specific than mere high valuations. It was about the way the money moves, and specifically about its circularity, and I want to explain the worry, because circular financing, money flowing round in a loop that makes everyone's numbers look bigger, is one of the oldest and most reliable warning signs in the history of financial manias, and it is now visibly present in the AI economy.

Here is the pattern that has people worried, simplified. A handful of giant companies are the AI economy: the chipmaker that sells the hardware, the labs that build the models, the cloud providers that host them, the companies that supply the power and the data centres. And increasingly, the money flows among them in circles. The chipmaker invests in the AI lab; the AI lab uses the investment to buy the chipmaker's chips; the chipmaker books the sale as revenue, which lifts its stock, which it uses to invest more. The cloud provider commits to buy the lab's services; the lab commits to buy the cloud provider's computing; each counts the other's commitment as demand. Enormous deals are announced among the same small group of players, each deal making each company's numbers look bigger, each company's rising value justifying the next deal, the money flowing round and round the same small circle, and from outside the circle it looks like a booming economy of vast and growing demand, when it may be, in part, the same money circulating among a handful of mutually-dependent giants, each propping up the others' numbers.

Why is circularity a warning sign? Because it can manufacture the appearance of real, external demand where there may be much less than it seems, and it is self-reinforcing on the way up and self-destroying on the way down. On the way up, the circle makes everyone's numbers look wonderful: revenues rising, investments flowing, valuations soaring, all apparently validating each other, and the validation draws in outside money that assumes the boom reflects real external demand. But if much of the demand is internal, the circle propping itself up, then the whole structure is more fragile than it looks, because the moment one player falters, the money stops flowing round the circle, and every company that was counting on the others' commitments finds the demand was partly illusory, and the numbers that looked so wonderful on the way up collapse together on the way down, each falling company pulling down the others it was entangled with. Circular financing turns a network of mutual support into a network of mutual contagion the instant confidence breaks.

I do not know whether the AI boom is a bubble; I wrote a year ago, and I maintain, that the technology is real and the question is only whether the price and the pace of return justify the spending, and that remains genuinely uncertain. But the circularity is a real and specific reason for heightened worry, because it is exactly the structure that can make a boom look more solid than it is and then fail faster than anyone expects. The warning sign is not that AI is worthless, it plainly is not, but that the money increasingly flows in a loop among a handful of entangled giants, each validating the others, and that this circular structure both inflates the appearance of demand on the way up and guarantees mutual collapse on the way down. When the money starts flowing in circles among the same small group of players, each propping up the others, it is time to ask how much of the boom is real external demand and how much is the circle admiring itself, because history's answer, when the circle finally breaks, has rarely been kind.

Field Notes
A Relentless build, told plainly

Automating the work that people were doing by hand, and the trap of automating the mess

This fortnight's front page is about money moving in circles and a party rediscovering the concrete. This is about the concrete, unglamorous work of automation: taking the tasks people do by hand and letting the machine do them, and the trap that swallows most attempts. No client is named.

By Alosh Paul

A great deal of the value I have delivered over my career came from a single unglamorous activity: taking work that people were doing by hand, laboriously, repetitively, error-prone, and building the systems to do it automatically. In one large operation, integrating a set of disconnected systems and automating the manual work that had connected them, people copying data from one place to another by hand, roughly halved the operational effort and eliminated a mountain of errors. Automation of this kind is one of the highest-value things software does: it takes the repetitive, mechanical, soul-deadening work that humans do badly because humans are bad at repetition, and gives it to a machine that does it perfectly and tirelessly, freeing the humans for the work that actually requires judgment. It sounds simple. And there is a trap in it that swallows most attempts, and I learned it the hard way.

The trap is this: automating a bad process just gives you a faster bad process. When you take a manual workflow, all the steps people currently perform by hand, and simply automate it as it is, you preserve, and now entrench in software, every inefficiency, every unnecessary step, every workaround and habit and piece of accumulated nonsense that the manual process had accumulated over years. The manual process was full of steps that existed only because of some old limitation, or some person's preference, or some problem that no longer exists, and if you automate it faithfully, you build all of that cruft into the machine, permanently, and now it is harder to fix because it is code instead of habit. The naive approach to automation, "let us make the computer do exactly what the people currently do, but faster," produces a system that does the wrong things efficiently, and is worse in a way than the manual process, because at least the manual process could be changed by asking people to work differently, while the automated bad process is now frozen into software.

The discipline, the thing that separates automation that delivers value from automation that entrenches waste, is to understand and redesign the process before you automate it, not after. Before you build the machine to do what the people do, you ask, ruthlessly, why they do each step, what it is actually for, whether it is still necessary, whether the outcome could be achieved a better way, and you throw out everything that exists only through habit or history, and only then, once you have the leanest correct process, do you automate that. Automation is not the act of making the computer do what people did; it is the act of understanding what actually needs to happen, designing the cleanest way to make it happen, and then building the machine to do that. The companies that get enormous value from automation are the ones that used the automation project as a forcing function to finally fix the process; the ones that get little are the ones that automated the mess and now have a faster mess. This is, I have come to see, a specific case of the deeper principle that runs through my whole trade, and that I will soon give its full argument: that the point of building your own system, rather than accepting a generic one, is precisely to shape it around the correct process for your actual work, and that the discipline of doing so, of understanding and redesigning before you build, is what separates technology that transforms an operation from technology that merely digitises its existing dysfunction.

The Ledger
AI
The return of bubble fears, sharpened by worry over circular financing among the AI giants, marks a shift in market mood from euphoria to scepticism. The technology remains real; the question, as ever, is whether the money, increasingly flowing in loops among a few entangled players, reflects real demand or its own reflection.
Data centers & power
The circular-financing worry connects directly to the build-out: the vast data-centre commitments that justify the spending are, in part, deals among the same handful of companies, and if the circle breaks, the half-trillion-dollar infrastructure vision, and the power plans behind it, could deflate with it.
Rates
The shutdown's end removes one uncertainty; the Fed continues easing. Markets, torn between AI-bubble fears and rate-cut relief, grow volatile as the year nears its close.
Real estate
US 30-year mortgage 6.24 percent (November 13). Steady near the year's low; the easing that has been the year's one reliable trend continues into winter.
India tech
The AI-bubble worry matters for India's services sector, which has bet heavily on AI-implementation work; a deflation of the boom would hit the firms riding the hype hardest, while rewarding those, in the frugal Indian tradition, delivering real and profitable results rather than circular promises.
What we called wrong
Nothing to retract, and a thread continued: the affordability lesson of the 2024 defeat, which this paper drew at the time, was vindicated in reverse this fortnight, as a party that finally spoke to the concrete cost of living was rewarded for it.
The Back Page

The kid becomes the mayor

In the summer, this paper marked a primary upset: a 33-year-old democratic socialist named Zohran Mamdani beating the entire machine of New York politics on a message about the cost of rent and groceries. We wrote then that the machine had lost to the kid, and that machines everywhere should note why. This fortnight the kid became the mayor. Whatever one makes of his politics, and they will be tested brutally by the reality of governing the largest, most complex, most ungovernable city in America, with a hostile federal government openly threatening to move against him, the arc of his rise is a genuine political phenomenon and a marker of something larger. A year ago the Democratic party was a smoking ruin, having lost the presidency to a convicted felon and lost its sense of what it was for. And the first real sign of its revival came not from the party's establishment, its consultants and its funders and its careful centrists, but from a young outsider who ignored all of them and talked, relentlessly and only, about the concrete, unglamorous, universal problem of how expensive it has become to simply live. He may fail as mayor; the job has broken more experienced people than he. But his rise is the clearest evidence yet of the political hunger this paper has tracked all year: for someone, anyone, of any ideology, who will address the thing that is actually crushing ordinary people, in language they actually speak, rather than lecturing them about abstractions from a comfortable distance. The kid talked about the rent. The kid became the mayor. It is not complicated, and the fact that so many powerful people find it baffling is exactly why the kid keeps beating them.