DUBAI — The United Arab Emirates halted all trade, commercial exchange and financial transactions with Iran in a foreign ministry statement issued in the small hours of Wednesday, closing without a stated expiry the largest commercial artery Tehran has left.

The stated cause was two ballistic missiles. The Emirati defence ministry said its air defences detected both launched from Iran on Tuesday, one falling outside Emirati territorial waters and one inside them, and that they were aimed at maritime traffic. Tehran called the accusation baseless. Neither version can be verified from outside, and in any case the important number in this story is not two.

It is roughly one third.

That is the share of everything Iran imports in a year that arrives, in one form or another, through the Emirates. Emirati exports to Iran grew from about $5.2bn in 2018, the year Washington left the nuclear deal, to something in the region of $23bn recently. Bilateral non-oil trade has run between $20bn and $30bn a year, which makes the UAE Iran's largest non-oil trading partner, ahead of China and Turkey. Around 10,000 Iranian-owned businesses operate inside the Emirates. Roughly half a million Iranians live there. Mark Kimmitt, the retired US general and former assistant secretary of state, put it to Al Jazeera plainly: the Emirati embargo is more significant than the American one.

He is almost certainly right, and the reason is architectural rather than political.

Western secondary sanctions have for two decades prevented most large suppliers from shipping directly to Bandar Abbas. So they ship to Jebel Ali. Goods land in Dubai, are re-documented, and are re-exported across the Gulf on smaller regional vessels and dhows. The Emirates did not smuggle for Iran. The Emirates provided the paperwork step that made the smuggling unnecessary. Layered on top is the financial half: Dubai's free zones host the entities through which oil and petrochemical proceeds have historically been converted into dollars, dirhams and euros outside a domestic banking system that no longer functions for the purpose. Iran imports something like $23bn of services a year, roughly a fifth of it Emirati. Separate estimates put smuggled imports at north of $20bn annually, much of it routed the same way.

What closed on Wednesday, in other words, is not a trade lane. It is the conversion layer between a sanctioned economy and a functioning one.

The timing is not accidental. The 60-day US-Iranian memorandum signed in June expired Monday without a successor, and President Trump declined to extend it. Brent settled just over $91 on Tuesday, its highest in about three weeks. The US naval blockade of Iranian ports continues and Washington has signalled a pivot from military pressure to economic pressure. The Emirati statement arrives inside that pivot, and it does more work than anything the Treasury Department has issued this year, for the simple reason that Treasury can only forbid Americans to trade with Iran, while Abu Dhabi can forbid the trade itself.

The consequences inside Iran are mechanical and fast. Electronics, machinery, spare parts and raw materials that arrive via Dubai now have to arrive via Oman or Turkey, at higher cost, lower volume and slower speed, if they arrive at all. Foreign exchange that cleared through Emirati institutions now has to clear somewhere else or not clear. Currency pressure follows within weeks, not quarters, because the rial's floor has for years been made of Dubai liquidity rather than central bank reserves.

The consequences on the Emirati side are smaller but not nothing. Dubai shipping agents, logistics houses and re-export traders lose a revenue line that many of them were built around. Iranian capital in Emirati property and retail freezes in place.

Which brings us to the only question that matters, and it is a question about precedent rather than principle. The UAE has done this before. It suspended direct cargo shipping with Iran in early March, days after the war began, and resumed it through Jebel Ali in late June — four months of closure, then quiet reopening, no announcement required. Nothing in Wednesday's statement suggests that pattern has changed. "Until further notice" is not a policy. It is a placeholder with a business model behind it.

So the thing to watch is not the statement. It is the manifests at Jebel Ali eight weeks from now, and whether the dhow traffic to Bandar Lengeh resumes before or after anybody in Abu Dhabi says so. Iran's problem this morning is real, immediate and severe. Whether it is durable depends on how long a trading city can tolerate not trading, and the historical answer to that question is: not very.

Ukrainian forces launched almost 800 drones into Russia overnight into Tuesday, the second-largest such attack since January 2025 by an Associated Press count, and the second salvo of that size in three days. Russia's defence ministry said 791 were shot down or intercepted across sixteen regions, Crimea, the Sea of Azov and the Black Sea. Saturday night's attempt was larger still: 822.

More than 600 of them went towards Moscow. Mayor Sergei Sobyanin reported 637 launched at the capital region between Monday evening and Tuesday morning, with 180 destroyed over the region itself. Officials reported no deaths and no major damage in the capital, which is the part of the story most coverage stopped at.

The part worth reading is what did get hit. A Wildberries warehouse caught fire. Wildberries is not a defence contractor; it is Russia's dominant e-commerce company, and a satellite image circulated by Ukrainian monitors showed one of its distribution facilities near Podolsk destroyed outright in Sunday's attack. Regional power infrastructure took damage in the Orekhovo-Zuyevo and Likino-Dulyovo districts, knocking out substations and cutting supply to tens of thousands.

This is what a mature drone campaign looks like when neither side can move the line. You cannot take Moscow with a Shahed-class airframe. You can make the delivery of a parcel to Moscow cost more, arrive later, or not arrive. Ukraine has stopped aiming at the war and started aiming at the economy that pays for it.

Russia is doing the same in reverse, with more lethal results. Two Banderol cruise missiles struck a busy intersection in Pechenihy, in the Kharkiv region, on Tuesday morning, killing at least ten and wounding seventeen. Governor Oleh Syniehubov said the site was a post office and shops surrounded by housing. Five more were injured in Izyum, including a nine-year-old.

Against that, the strategic assessment out of Kyiv this week was unusually blunt. Major General Vadym Skibitskyi, deputy chief of Ukraine's military intelligence directorate, told RBC-Ukraine on Tuesday that his analysts see no battlefield indicator that Russia is preparing to stop, and expect no peace agreement before the end of 2026. Russia's general staff, he said, still intends to finish taking Donetsk Oblast this year.

He also produced the number that undercuts the plan. Russian personnel losses exceeded contract recruitment in June and July. Moscow has a 2026 target of 409,000 recruits and cannot form new units or upgrade brigades to divisions without mobilisation. Recruitment into its unmanned systems forces stands at about 30 percent of plan, and is being backfilled from other branches. If Russia mobilises after September's parliamentary elections, Ukrainian intelligence believes it could assemble a serious strike group by roughly February 2027.

That is the shape of the thing. One side cannot generate infantry and is spending missiles on intersections. The other cannot generate a breakthrough and is spending drones on distribution centres. Meanwhile the American diplomatic track has been parked since Washington's attention went to the Gulf, and France, Germany and Britain are quietly building a separate format specifically so that Europe holds a seat when it restarts.

Everybody involved is now planning around a war that outlives the year. The drones are not an attempt to end it. They are an attempt to make it expensive enough that somebody else decides to.

For years, the standard way to say when heat becomes unsurvivable has been the wet-bulb temperature: the reading that combines heat and humidity to estimate the point at which a body can no longer shed heat and core temperature simply climbs. It is the number behind heat warnings, labour rules, migration projections and a great deal of climate journalism.

It was established in laboratories, on young healthy adults.

A modelling study published Tuesday in The Lancet Planetary Health, led by researchers at Stanford, ran the numbers again using newer chamber experiments that tested how people over sixty tolerate heat and humidity. Older bodies sweat less, move less blood to the skin, and pump less. They cross into what the authors call incompensable thermal stress at lower wet-bulb readings than the standard assumes.

Correcting for that, and for the fact that the world's population is ageing, changes the picture enormously.

At 1.5C of warming above pre-industrial levels, the study projects that about 290 million people aged 60 and over, roughly 22 percent of that age group worldwide, would face at least 180 hours a year of heat their bodies cannot compensate for. That is a month of unsurvivable conditions, annually, at a warming level frequently described in public debate as the safe target.

The sentence in the paper that ought to travel is this one: older adults face greater heat risk at 1.5C than young adults do at 4C.

At 3C, the model puts global exposure at 2.03 billion people, with India, China, Pakistan and Bangladesh accounting for 1.48 billion of them, or 72.6 percent. In the worst-affected regions, South Asia and the Persian Gulf, older people would face dangerous heat day and night for three consecutive months a year. The countries carrying the most risk are the ones with the least cooling: India, Pakistan, Bangladesh, Myanmar, Niger.

The authors are careful about their own uncertainty. This is a model, its outputs depend on model choices, and they say so. They ask for two things: expand access to cooling and public health protection now, and cut emissions to hold down the long-run rise in humid heat. They also note, in the register scientists use when they mean something alarming, that incompensable heat on this scale could increasingly drive internal displacement and cross-border climate migration.

What makes this a front-half story rather than a science brief is the methodological point underneath it. Every threshold in this field was set by measuring the most resilient available subject and calling the result human tolerance. The correction is not that the climate got worse on Tuesday. It is that the ruler was wrong, and it was wrong in the direction that flatters us.

One further note on geography. The Persian Gulf appears twice in today's paper: once as the region where a trade war is being fought over a waterway, and once as one of the two places on earth projected to become physiologically intolerable for its older residents for a quarter of the year. The same coastline, two different clocks.

Tuesday's tape was a straightforward transmission mechanism, and it is worth watching it work.

The 60-day US-Iran memorandum expired Monday without extension. Brent added 0.24 percent to settle at $91.09; WTI rose 0.60 percent to $84.24. The S&P 500 energy sector gained around 1.8 percent, the only meaningful advance on the board. The 30-year Treasury yield reached its highest level since 2007, a nineteen-year high, and the 10-year touched levels last seen in January 2025. And then equities did what equities do when the discount rate rises: the S&P 500 fell 0.69 percent to 7,691.76, the Nasdaq Composite dropped 1.33 percent to 26,289.71, the biggest one-day fall for both since 29 July, and the Dow lost 0.22 percent to 53,343.40. Semiconductors led the decline.

Four steps, each mechanically following the last, and the first step happened in Abu Dhabi and Tehran rather than in Washington.

This is a genuinely awkward position for the Federal Reserve, and the awkwardness is not about the level of rates. It is about authorship. Fed funds futures continue to price roughly 67 percent odds of a hold at the 16 September meeting against roughly 32 percent for a 25 basis point hike, essentially unchanged for days. Nothing the committee has said is moving those numbers. What is moving the long end is a risk premium priced off a shipping lane, and no amount of forward guidance addresses it, because the Fed cannot open the Strait of Hormuz and everyone in the market knows it.

Minutes from the July 28–29 meeting arrive Wednesday afternoon, with three dissents in favour of a hike to explain. They will be read, inevitably, as an oil document. Anything in them that sounds tolerant of energy-driven inflation will be treated as dovish; anything that sounds worried will push the long end higher again. This is a strange fate for a set of minutes written before the memorandum lapsed.

The forecast to interrogate is the EIA's, which has Brent averaging about $78 a barrel in the fourth quarter on the assumption that Hormuz traffic improves and disrupted production returns. Read that carefully. It is not a supply model. It is a behavioural prediction about the Iranian government, wearing a supply model's clothing. Every institutional forecast in the energy complex right now contains one of these, and almost none of them label it.

The test for the next few weeks is narrow and specific. A crude spike alone is survivable and the Fed can look through it. Crude, long yields and inflation expectations rising together is a different animal, because that combination is not a shock, it is a regime. On Tuesday all three moved the same way at the same time. That is one observation, not a trend. Two more and the September pricing stops being 67-32.

Small batches of Nvidia H200 accelerators have entered mainland China, the Financial Times reported Tuesday, citing two people with knowledge of the shipments. ByteDance and Tencent have each received roughly 10,000 processors in recent weeks. Several other Chinese firms may be cleared for similar volumes shortly.

The interesting figure is the one they did not receive. Under the American licensing terms, each approved customer may buy up to 75,000 to 100,000 H200s. Nvidia is sitting on something like 500,000 of them, built largely for Chinese demand. The constraint is not supply and it is no longer Washington. Beijing has told its own champions to keep the hardware outside the mainland to protect the growth of domestic chipmakers, and has pointed them at Hong Kong, which sits outside the mainland customs border and can therefore host the machines without technically importing them.

So the current state of the world's most consequential technology control regime is this: the United States permits the sale, China restricts the purchase, and the chips end up parked in a jurisdiction chosen precisely because it is neither.

How we arrived here is worth restating, because the sequence is faintly absurd. Export controls on advanced AI chips began in 2022 to slow Chinese model development. In December 2025 the president authorised H200 sales to China on the condition that Nvidia hand 25 percent of the revenue to the US government. The H200 is roughly six times the capability of the H20 that was previously the ceiling. Blackwell remains banned, and in May the Commerce Department's Bureau of Industry and Security moved to extend licence requirements to Chinese-headquartered entities operating outside China, closing the offshore-subsidiary route. Beijing is now using a version of the same route in the opposite direction.

Both governments are doing industrial policy through the same instrument. Washington wants the revenue and the leverage without the capability transfer. Beijing wants the capability without letting Nvidia entrench itself against Huawei and Cambricon domestically. The overlap is a set of chips that exist, are paid for, are legal, and are physically stranded in Hong Kong.

A report from the US-China Economic and Security Review Commission, published Monday, argues that this entire apparatus may be aimed at the wrong bottleneck. Chips constrain training. But once a moderately capable model exists, the advantage comes from deploying it across millions of industrial endpoints and harvesting the feedback, and China is deliberately building for that: standardising data assets across manufacturing, transport, finance and healthcare, and listing proprietary datasets on data exchanges in Shanghai, Shenzhen and Beijing. The commission notes that this is the kind of non-scrapable operational and physical-world data that embodied AI, robotics and autonomous systems actually need, and which no export control touches.

It also documents the mirror-image cost. China's tightened cross-border data rules have forced foreign multinationals to localise their Chinese subsidiaries and wall off Chinese customer data from global operations, with limited exemptions granted later.

Which leaves a supply chain in which the hardware cannot fully enter and the data cannot leave. Both sides have built the same wall, from opposite sides, and are now discovering that they are standing behind it together.

Three federal judges reached three different conclusions this week about whether the administration may reshape the physical fabric of the capital, and the pattern in the answers is not ideological. It is temporal.

On Monday, US District Judge Theodore Chuang blocked the relocation of FBI headquarters to the Ronald Reagan Building, ruling that the FBI and the General Services Administration exceeded their authority in discarding the Greenbelt, Maryland site selected in 2023, and that reprogramming $555m toward the Reagan move, including $323m Congress appropriated for Greenbelt, was arbitrary and capricious. Nothing has been built. The ruling therefore bites cleanly.

On Tuesday, US District Judge Dabney Friedrich allowed paint tests to proceed on the Eisenhower Executive Office Building, the nineteenth-century landmark next door to the White House, finding that preservation groups had not shown minimal testing would cause irreparable harm. The larger question, whether GSA may repaint the building white at all, remains live. The tests are small. They are also the first physical act, and physical acts have a way of establishing what the argument is subsequently about.

And on Tuesday the National Trust for Historic Preservation asked the Supreme Court to suspend construction of the $400m, 90,000-square-foot White House ballroom while the court decides whether it may continue. A divided appeals panel had upheld Judge Richard Leon's order pausing aboveground work, and Leon's line has been quoted everywhere it deserves to be: no statute comes close to giving the president the authority he claims to have. That order was itself stayed to permit appeal. The administration argues that renovation of the White House is the president's alone, that the Trust lacks standing, and, in a filing that repays a second reading, that the site's current condition as an open construction pit makes the building harder to protect.

That last argument is the one to sit with. The East Wing was demolished to make room for the ballroom. It is gone. The Justice Department is now citing the consequences of the demolition as grounds to permit the construction. Whatever the legal merit, the structure of the claim is that the hole in the ground has become a fact the court must accommodate.

Lawsuits over a proposed triumphal arch and a statuary garden are pending. Every project, the administration notes, is proceeding.

This is the asymmetry that matters more than any individual ruling. Litigation moves at the speed of briefing schedules, and each of these judges is doing careful, bounded work within that speed. Excavators move at the speed of excavators. A judgment can void an appropriation, as Chuang's did, because money can be moved back. It cannot restore a demolished wing, and once a building is painted the case becomes a case about repainting it.

So the courts will keep issuing correct rulings, and the capital will keep changing, and only some of that change will be reversible by the institution formally charged with reviewing it. Read the calendar, not the docket.

The trailer for *Oasis: Don't Look Back in Anger* arrived Tuesday, and everyone filed the same piece about it, so here is a different one.

The facts are tidy. Two hours, directed by Dylan Southern and Will Lovelace, written and produced by Steven Knight, produced by Magna Studios, presented by Sony Music Vision, distributed by Walt Disney Studios. World premiere out of competition at Venice on 5 September. Cinemas and IMAX on 11 September, with the UK two days early and IMAX early access screenings in North America on 9 September. Hulu and Disney+ later this year. It covers the Live '25 reunion tour, the first time Noel and Liam Gallagher had shared a stage in sixteen years, and it contains their first joint interview in more than twenty.

The line they chose for the trailer is Noel's: "The songs are one thing, but I'd forgotten what he means to people. I certainly wasn't prepared for the intense outpouring."

It is a good line and it is also a slightly startling one, because it is a man describing his brother as a market. Not with contempt — the affection is audible — but the sentence is structured around demand. What he means to people. The outpouring. The reunion tour was the most efficiently monetised nostalgia event in modern British music, and the film's own press language, an unapologetically uplifting account of the biggest musical event of 2025, is not the language of a documentary. It is the language of an aftermarket.

Which is what makes the directors an interesting choice rather than an obvious one. Southern and Lovelace made *Shut Up and Play the Hits*, still the best film anyone has made about a band ending, precisely because it declined to be uplifting: LCD Soundsystem's James Murphy, the morning after Madison Square Garden, alone with his dog and a room full of equipment he no longer needs. They also made *Meet Me in the Bathroom*, which is largely about how scenes curdle. These are filmmakers with a demonstrated feel for the hangover.

Now they have been hired to shoot the opposite ceremony, by the company that will stream it.

The reunion has already been consumed. Whatever it meant happened in Cardiff in July 2025 and in the stadiums after, and 2026 is the year it gets converted into a durable asset: a theatrical window, an IMAX premium tier, a Venice slot for prestige, a streaming tail. The joint interview is the crown jewel, and it is worth noticing that the most private thing these two men have done in two decades is now a scheduled release event with advance tickets on sale.

None of which means the film will be bad. Southern and Lovelace could make something genuinely strange out of this, and the material is there: two men in their fifties discovering that the thing they spent sixteen years refusing to do was the thing everyone wanted, and having to feel that in public, on camera, on a schedule.

But watch the edit. If the reconciliation lands as catharsis, it is a promotional film with excellent directors attached. If it lands as slightly uncomfortable, if there is one shot where you can see the transaction, then they got away with something. The whole question of the movie is whether Disney bought a celebration or accidentally bought the truth.

Consider what the United States has spent to pressure Iran since February: a carrier presence, a naval blockade, a sanctions apparatus refined over two decades, an expired memorandum, and a president announcing that the Strait of Hormuz is essentially American. Brent at $91. Six thousand seafarers stuck in a waterway. A war.

Now consider what the United Arab Emirates spent on Wednesday morning: one sentence.

All trade, commercial exchange and financial transactions with Iran halted until further notice. That is more damaging to the Iranian economy than anything Washington has issued this year, and it took a foreign ministry statement of under a hundred words. A retired American general said as much on the record within hours, which tells you the assessment is not controversial inside the system that spent the two decades building the alternative.

The lesson is not subtle and it is not new, though we keep declining to learn it. Sanctions imposed by an adversary are a tax. Sanctions imposed by a neighbour are an amputation. Washington can forbid Americans from trading with Iran, which Iranians long ago stopped noticing. Abu Dhabi can forbid the trade itself, because the trade physically happens there: at Jebel Ali, in the free zones, in the currency houses, in the ten thousand Iranian-owned companies operating inside the Emirates and the half-million Iranians who run them.

American sanctions built that. Every secondary sanction that made it illegal to ship to Bandar Abbas created a business in Dubai. The architecture of denial produced its own workaround, and the workaround became the single point of failure, and this week the single point of failure was pulled by a government that Washington does not control and did not need to persuade.

So: an extraordinary result, and I expect it to last roughly a quarter.

Dubai is not a policy. Dubai is a market position, and the position is *we handle what others cannot handle*. That is the product. Not the port, not the tax regime, not the towers: the willingness to be the place where a difficult transaction becomes an easy one. A city that has spent forty years selling that cannot spend two years not selling it without the demand relocating, and the candidates are already obvious. Sohar and Duqm in Oman. Turkish overland routes. More dhow traffic, less documentation, worse margins for everyone and better margins for whoever accepts the risk.

The Emirates already ran this experiment. They suspended direct cargo with Iran in early March, days into the war, and resumed it through Jebel Ali in late June. Nobody issued a press release about the reopening. That is the tell. Announcements are for closures; reopenings are for manifests.

What should worry Iranian planners is not the embargo. It is the second-order fact that Iran's foreign exchange floor has for years been made of Dubai liquidity rather than central bank reserves, and that this arrangement was always a courtesy extended by a state with entirely different interests. The rial's real reserve asset turned out to be Emirati patience. That is not a currency policy, it is a hostage situation that everyone had agreed to call trade.

And what should worry American planners is the reverse. The most effective instrument in the entire campaign is one they do not hold, cannot renew, and did not design. If Abu Dhabi decides in November that the war has become somebody else's and the money is better in the ledger, no one in Washington gets a vote. You cannot build a strategy on a favour.

The hard version of this argument, the one I would rather people forwarded, is that the whole Western sanctions edifice has been a twenty-year exercise in outsourcing enforcement to states with commercial reasons to enforce it and commercial reasons to stop. It works beautifully until the day it doesn't, and then it works beautifully for someone else. This week it worked. Write the date down. Then check the manifests in October.

The study in *The Lancet Planetary Health* is being covered as a climate story, and it is one. But strip the warming scenarios out and what remains is a finding about method, and the method is used everywhere.

For years the threshold at which heat becomes physiologically unsurvivable was derived from chamber experiments on young healthy adults. It then travelled outward, as thresholds do, into heat warnings, into projections of habitability, into the sentence "parts of the world will become uninhabitable by" that you have read a hundred times. Nobody hid the sample. It simply stopped being mentioned once the number acquired a name.

Redo the experiments on people over sixty and the picture changes so much that the paper contains this: older adults face greater heat risk at 1.5C of warming than young adults face at 4C. Two and a half degrees of difference, produced not by the atmosphere but by whose body you measured.

This habit has a long service record.

Crash test dummies were modelled on the fiftieth-percentile male for decades, and women were measurably likelier to be seriously injured in comparable collisions. Clinical trials underrepresented women until regulators forced the issue, and dosing guidance inherited the gap. Occupational heat limits are written around workers assumed to be fit enough to hold the job, which is a definition that quietly excludes the people most likely to be hurt. Evacuation models assume walking speeds. Thermal comfort standards in buildings derive from studies of office workers, which is to say of the employed, which is to say of the healthy.

In each case the error runs the same direction. We measure the person best equipped to survive the thing, call the result the human limit, and then design the world to that limit. It is not malice and rarely even carelessness. It is that the fittest subject is the easiest to recruit, the cheapest to insure, and the least likely to produce an inconvenient result.

The consequence is that our safety margins are systematically thinnest exactly where the exposure is highest, because ageing, poverty and heat are correlated and the standard is calibrated to the intersection of none of them. The countries the study flags — India, Pakistan, Bangladesh, Myanmar, Niger — are not merely hot. They are hot, ageing, and short of cooling, and the threshold that governs how urgently anyone treats that was set on a treadmill in a temperate laboratory.

I want to be careful about the size of the claim. This is a modelling study, its authors say plainly that the outputs depend on model choices, and one paper does not resettle a field. The correction may itself be corrected.

But the direction of the error is the point, and the direction is knowable in advance. Any standard derived from the most resilient available body will understate risk for everyone else, always, in every domain, and the understatement will be largest for the people with the least capacity to compensate. That is not an empirical question. It is arithmetic about sampling.

So the useful response is not to argue about 1.5 versus 3. It is to ask, of every threshold currently governing anything you care about, a single question: measured on whom? Ask it of your building's heat plan, your city's warning trigger, your employer's outdoor work rules. The answer is usually available and usually unflattering, and unlike the climate it can be changed by Friday.