Vol. III, No. 66
Covering 16 June - 29 June 2025
Monday, June 30, 2025
Late edition · A Relentless publication
All the fortnight that mattered, in technology and in the world, read next to what we were building at the time.
THE MIDDLE EAST

America bombs Iran's nuclear sites, and then, abruptly, the guns go quiet

American stealth bombers struck Iran's most fortified nuclear facilities with the largest bunker-busting bombs ever used in combat. Iran fired back at a US base, telegraphing the blow to avoid casualties, and within days a ceasefire held. A war that could have engulfed the region ended, for now, in 12 days.

The United States entered the war directly on June 21 and 22, sending B-2 stealth bombers to strike Iran's three key nuclear sites, including the deeply-buried enrichment facility at Fordow, with massive bunker-busting bombs, the first combat use of the largest conventional weapons in the American arsenal, in an operation the president declared had "obliterated" Iran's nuclear programme. It was the gravest American military action in the Middle East in a generation, undertaken without a vote of Congress, and it raised immediately the spectre of a wider war. Iran's response, on June 23, was calibrated to avoid exactly that: it fired missiles at the largest American base in the region, in Qatar, but telegraphed the attack in advance, allowing the base to be evacuated and no one to be killed, a face-saving retaliation designed to satisfy honour without forcing escalation. And on June 24, astonishingly, a ceasefire took hold, and the 12-day war between Israel, America and Iran simply stopped.

The suddenness of the ending was as striking as the ferocity of the war, and it left the central question unanswered: whether the strikes had actually destroyed Iran's nuclear programme or merely delayed it. Intelligence assessments leaked in the days after suggested the damage might be less total than the president claimed, that Iran's stockpile of enriched uranium and its knowledge remained, that the programme had been set back months or years rather than obliterated. And the deeper strategic question loomed larger than the rubble: whether bombing a nation's nuclear sites destroys its will to build a weapon or guarantees it, whether Iran, having been struck precisely for approaching the bomb, would now conclude that only the bomb could deter such strikes in future. The war ended in 12 days. Whether it made a nuclear-armed Iran less likely or more, no one could yet say, and the history of such strikes offered little comfort.

WASHINGTON

The Supreme Court takes away the courts' power to stop the president everywhere at once

In a decision of enormous consequence for the rule-of-law confrontations of the past months, the Supreme Court ruled on June 27 to sharply curtail the power of federal judges to issue nationwide injunctions, the tool by which a single court had been able to block an unlawful government policy for the entire country while it was litigated. The ruling, arising from the fight over the president's attempt to end birthright citizenship, means that a policy blocked by one court can now often proceed everywhere except for the specific plaintiffs who sued, forcing challengers to fight the same battle court by court, state by state, and vastly expanding the president's practical ability to enact contested policies before the courts can stop them. It was a technical ruling with a profound effect: the weakening of one of the judiciary's most powerful tools for restraining the executive, handed down at the exact moment that restraint was being most tested.

IN BRIEF

An outsider wins in New York; a summit's spending pledge; a billionaire's wedding

Zohran Mamdani, a 33-year-old democratic socialist state legislator, won the Democratic primary for mayor of New York City on June 24 in a stunning upset over the establishment favourite, a former governor, powered by a youth-driven campaign focused on the cost of living, a sign of the generational and ideological churn in a Democratic party searching for direction after its 2024 defeat. NATO leaders, meeting in the Hague, agreed to a historic pledge to raise defence spending toward 5 percent of economic output, Europe's belated, expensive answer to the American retreat. And in Venice, the Amazon founder Jeff Bezos married in a days-long celebration of such lavishness that it drew protests, a gilded emblem of an age of extreme wealth.

The Column

The check that was disarmed, not defied

The Supreme Court curtailed the courts' power to block the president nationwide this fortnight. It arrived as a technical ruling about procedure, and that is exactly why it is more dangerous than open defiance, because it works on the machinery a check needs to function at all.

The Supreme Court ruled this fortnight to curtail the power of federal courts to issue nationwide injunctions blocking the president's actions, and the ruling arrived clothed in the dry language of procedure, a technical adjustment to the scope of judicial remedies, nothing so dramatic as a constitutional confrontation. That modesty is exactly why it deserves more alarm than a louder act would, because it is a weakening of a check that works not by defying the check but by quietly disarming it, disabling the machinery a check needs in order to function, and that method is both more durable and more dangerous than open defiance.

The principle at stake was stated most clearly by Montesquieu, whose analysis of the separation of powers shaped the very constitution now in question. Montesquieu's insight was not merely that power should be divided among branches, which is the version everyone remembers, but the subtler and more important point that each branch must possess the actual means to resist the others, that the separation is worthless unless each power holds concrete instruments capable of checking the encroachments of the rest. A judiciary that can declare an executive action unlawful but cannot actually stop it is not a check; it is a commentary. The power to halt an unlawful act, promptly and effectively, is the means of resistance that makes judicial review more than advisory, and the nationwide injunction, whatever its procedural imperfections, was one of the sharpest such instruments the courts possessed, the tool that let a single court stop an unlawful national policy before it did its damage. To curtail it is not to defy the judiciary's authority, which would provoke a crisis; it is to leave the authority formally intact while removing one of the means by which it is exercised, which provokes nothing, and that is precisely the danger.

The counterargument is genuine and I will state it at full strength, because the nationwide injunction really was a problematic instrument. It let a single district judge, of either party, block national policy for the entire country, invited forum-shopping for friendly courts, and produced a chaos of conflicting orders; serious legal thinkers across the spectrum had criticised it for years, and there is a real, non-partisan case that its curtailment is a sensible reform rather than an assault on the courts. I hold that this is not frivolous. But Montesquieu's framework cuts through the merits of the instrument to the question of the moment, which is not whether the nationwide injunction was ideal in the abstract but what its removal does to the balance of means at this particular time, when an executive is testing the limits of every check as aggressively as any in memory. A reform that trims judicial power is one thing in an era of executive restraint and quite another in an era of executive assault, because the same procedural adjustment that merely tidies the law in calm times removes a critical means of resistance in a confrontation, and the timing determines which it is. A tool can be imperfect and still be the one thing standing between an unlawful policy and its execution.

And that is the deeper lesson about how checks actually die, which is almost never by the dramatic defiance everyone is watching for. The cinematic version of constitutional crisis is the executive openly refusing a court order, and that would indeed be a crisis, visible and galvanising. But the more common and more effective method, the one Montesquieu's framework helps us see, is the quiet disarming, the procedural narrowing, the technical ruling that leaves the check formally standing while removing the means that made it work, so that the authority remains on paper and the power to exercise it drains away, and no single moment ever looks like the crisis it cumulatively is. A check that is defied provokes resistance; a check that is disarmed provokes a debate about legal procedure, and by the time the debate is settled the means of resistance is gone. The court did not defy the judiciary this fortnight; it disarmed it, slightly, technically, durably, and the modesty of the method is not evidence that little happened but the very reason so much did.

Field Notes
A Relentless build, told plainly

Replacing the system nobody dares touch, one careful piece at a time

This fortnight's front page is about tools of restraint being weakened. This is about the most dangerous project in enterprise software, replacing the old system the whole business runs on, and the disciplined, unglamorous way to do it without bringing everything down. No client is named.

Every large organisation has one: the old system that the whole business secretly runs on, that nobody fully understands any more, that was built years or decades ago by people long gone, that is fragile and creaking and increasingly unable to do what the business now needs, and that everyone is terrified to touch, because it is load-bearing and its failure would be catastrophic. Replacing it is the most dangerous project in enterprise software, and I have done it more than once, and the single most important thing I learned is how not to do it: you do not replace it all at once. The temptation, always, is the big-bang cutover, build the shiny new system in parallel, and then, on one terrifying weekend, switch everything over from the old to the new. And the graveyard of enterprise technology is full of organisations that tried exactly that and brought their whole operation down, because the old system had accumulated, over its long life, a thousand undocumented behaviours and edge cases and quiet dependencies that nobody remembered until the new system, which did not replicate them, went live and the business discovered, catastrophically, all the things the old system had been silently doing that no one had known to rebuild.

The disciplined way, the way that works, is the opposite of the heroic weekend: you replace the old system gradually, one piece at a time, while it keeps running. You identify one function the old system performs, build the new system to do just that one thing, route only that piece of the work to the new system while everything else stays on the old, watch it carefully, and only when it is proven do you move the next piece. Over months, function by function, you starve the old system of its responsibilities, moving them one at a time to the new, until eventually the old system is doing nothing and can be quietly switched off, having been replaced not in a big dangerous leap but in a long series of small, reversible, survivable steps. It is slower. It is less heroic. It lacks the drama of the big cutover weekend. And it is enormously more likely to work, because at every step you are risking only one small piece, and if that piece fails you can roll it back without bringing down the whole operation, and you discover the old system's hidden behaviours one manageable piece at a time rather than all at once in a catastrophe.

I raise this in a fortnight about the weakening of institutional checks because there is a connection, and it is about how you change load-bearing systems safely, whether technical or civic. The reason the gradual approach works and the big-bang fails is that a load-bearing system, one the whole operation depends on, accumulates over time a vast web of dependencies and behaviours that are not written down anywhere, that exist only in the fact of the system's continued functioning, and that you discover you needed only when they are gone. This is true of an old software system, and it is true, I would gently suggest, of old institutions and old checks and old restraints, the nationwide injunction, the federal structure, the norm, the treaty, that have accumulated, over long lives, a web of stabilising functions that nobody fully catalogues until they are removed. The disciplined engineer replacing a load-bearing system does it gradually, reversibly, one careful piece at a time, precisely because they respect how much the old system was silently doing that no one remembered. It is a discipline worth extending, I think, beyond software: before you remove a load-bearing thing, technical or institutional, that has worked for a long time, it is worth asking, humbly, what it might be silently holding up that you have forgotten to notice, and whether you can afford to find out the hard way.

The Ledger
AI
The intelligence dispute over the Iran strikes, whether they "obliterated" the programme or merely dented it, is a case study in the epistemics this paper tracks: confident claims, contested evidence, and the difficulty of establishing truth about a thing buried under a mountain and a fog of politics.
Data centers & power
Oil markets swung violently on the war and its sudden end; the energy volatility underlying the AI build-out's power question remains hostage to a Middle East that can go from war to ceasefire in 12 days.
Rates
Markets, having feared an oil shock, rallied on the ceasefire; the Fed holds, granted a reprieve from the inflationary threat of a wider war, though the tariff turmoil grinds on beneath it.
Real estate
US 30-year mortgage 6.77 percent (June 26). Range-bound; the American housing market's strange placidity amid global war and constitutional erosion continues to puzzle.
India tech
A quieter fortnight for India after the crash and the crisis; the technology sector watches the weakening of American institutional checks with the wariness of an industry whose fortunes depend on American stability and the rule of law.
What we called wrong
Nothing to retract. The SCOTUS injunction ruling extends this paper's long attention to "the layer that can say no": the courts' power to restrain the executive was not defied this fortnight but curtailed, a quieter and more durable weakening of the check than any single act of defiance.
The Back Page

The machine loses to the kid

The political story of the fortnight that will echo longest was not the war but a primary. Zohran Mamdani, 33 years old, a state legislator almost no one outside his district had heard of a year ago, a democratic socialist running on the cost of rent and groceries and the daily arithmetic of a city too expensive to live in, beat the entire Democratic machine to win the primary for mayor of New York, defeating a former governor with universal name recognition, vast funding, and the backing of every power broker in the state. He did it with an army of young volunteers, a relentless focus on the concrete cost of ordinary life, and a fluency in the media of the young that the establishment could not match. Whatever one thinks of his politics, and they will be fiercely contested, the result is a marker of something larger: the exhaustion of a generation with a political establishment that seems unable to address the thing crushing them, the sheer cost of building a life, and the arrival, in force, of a younger cohort that intends to do politics differently. This paper wrote, after the 2024 defeat, about a Democratic party that had lost touch with the concrete concerns of struggling people, that had asked voters to prioritise the abstract over the price of eggs and lost. Mamdani is, in a way, the answer from the other direction: a candidate who talked about nothing but the price of eggs, and rent, and buses, and won by doing so, against a machine that had forgotten how. The general election is another matter, and he may yet lose it. But the machine lost to the kid this fortnight, and the machines everywhere should take note of why: he talked about the thing that was actually hurting people, in the language they actually speak, and it turned out to be enough.