Two sentences from the Oval Office on Wednesday, spoken minutes apart, describe this war better than any briefing has.

"It was a very heavy attack last night, and we're prepared to do another one any time we want," President Trump told reporters. Then he said the renewed campaign would not go on "too long."

Read as forecasts, those statements contradict each other. Read as a description of an option, they agree exactly. An option is the right to act without the obligation to act, held open for a fee. Six months after the 28 February strikes that opened the war and killed Ayatollah Ali Khamenei, the United States has finished converting a military campaign into one.

The paperwork surfaced the same day. The Wall Street Journal reported that Defence Secretary Pete Hegseth is extending American troop deployments in the Middle East into next year, described by people familiar with the planning as a strategy to leave the President "many open-ended options." That is an unusually candid statement of purpose and it repays slow reading. The troops are not there to achieve a stated objective by a stated date. They are there so that striking stays available.

The premium

It came due late on Monday in Kuhestak, a coastal town of no strategic consequence overlooking the strait, where a US munition hit a house holding a wedding. The Iranian Red Crescent said four people were killed — two women and two children, aged four and sixteen — and more than fifty wounded; state media later put the wounded at 68. Ahmad Nafisi, the province's deputy governor, had told Iranian state television that five were killed. The figure subsequently became four. Nobody explained the revision, and we have not been able to reconcile it. Across the two nights, Iranian officials counted 18 dead and at least 108 wounded.

It came due at 23:40 local time on 31 August aboard the crude tanker Sidr, where two Filipino seafarers died in a fire their fourteen surviving crewmates contained themselves.

It came due in Tehran's currency market, where traders exchanged 2.20 million rials for a dollar on Wednesday, ten per cent weaker than last week's record.

And a smaller, more genteel instalment falls due in Washington on 16 September, when a Federal Reserve now roughly two-thirds priced for a quarter-point rise decides whether to raise the cost of an American mortgage because of a shipping lane on the other side of the world.

The count, not the price

The honest measure of what six months of optionality has purchased is not the oil price. It is the transit count.

Crude tanker transits through the Strait of Hormuz averaged roughly 24 a day in January and February. Preliminary Kpler data show four commodity vessels crossed on Tuesday and ten the day before, against a ten-day average near 13. Iran has widened the list of vessels it deems non-compliant and liable to fine, detention or seizure.

Against that, the US military escorted 40 vessels carrying 18 million barrels through the strait, and the President said American forces are "helping bring lots of boats out every day with millions of barrels of oil." He also said the United States controls the waterway.

Both accounts can be accurate simultaneously, and that is the point. Convoys move oil. A waterway you can cross only in a guarded column, after clearance by a navy, is not an open waterway. It is a toll bridge with a queue and a casualty rate. Brent traded near $95.50 on Thursday, more than thirty per cent above where the war started, which is approximately what a toll bridge of that description costs.

Everything here has a date except the war

Consider the calendar everyone else is keeping. August payrolls land on Friday. August CPI lands on 11 September. The Federal Open Market Committee meets 15 and 16 September. Saxony-Anhalt votes on 6 September. The Jewish high holidays begin in mid-September, and Israel's defence minister has said his country is prepared to respond "with great force" if attacked during them. Americans vote on 3 November. The Texas campus that will host $35bn of Anthropic's contracted compute energises in the first quarter of 2027.

Every institution in this issue is planning against a date. The exception is the war, which now has a troop rotation schedule running into 2027 and no terminal condition anyone in the government has been willing to say out loud. A deployment extension is not a plan. It is a renewal notice.

There is one more entry on the calendar and it belongs to the reader. Tell us what you have already rescheduled, cancelled, repriced or refused because of this war — a shipment, a wedding, a mortgage, a voyage, a hire. We print letters, including the ones that show us the frame is wrong.

MANILA — The two men who died aboard the Sidr were doing the job that moves the world's oil, which is to say they were Filipino.

At an online briefing on Wednesday evening, Migrant Workers Secretary Hans Leo Cacdac laid out what his department knows. Sixteen Filipino seafarers were aboard the Saudi-flagged crude tanker when it was struck by what he called "still unknown projectiles" late on 31 August as it approached the Strait of Hormuz. A fire broke out. The crew contained it. Two of them did not survive. Their remains have been brought ashore in Oman; the fourteen survivors are safe and awaiting repatriation.

Bahri, the Saudi national shipping company that operates the vessel, confirmed the deaths in a statement timed to 23:40 local. Saudi Arabia said on Wednesday the attack was Iranian. The projectiles have not been independently identified, and Greek maritime risk firm Marisks reported two tankers struck overnight Monday to Tuesday without attribution.

Then Cacdac gave the aggregate, which is the story.

Since 28 February, 31 vessels carrying 421 Filipino seafarers have been attacked in incidents involving the Strait of Hormuz. Of those 421: 406 unharmed, ten injured and since recovered, three killed, one still missing. Two hundred and sixty-eight have come home. An estimated 3,000 Filipino seafarers remain in the Persian Gulf, though the department stresses that does not mean 3,000 are in the strait.

Set that ledger against the traffic. Four commodity vessels crossed the strait on Tuesday, against a pre-war average near 24. Transits have collapsed by roughly four-fifths. The casualty list has not.

The right, and its price

The Department of Migrant Workers keeps the strait under red alert, classifies it as a war-like high-risk zone, and requires shipowners to conduct risk assessments, implement security measures and consider escorts before sailing. It also, repeatedly and in writing, reminds the industry that Filipino seafarers have the right to refuse to sail into designated warlike areas.

"If a seafarer exercises the right to refuse to sail, that must be respected," Cacdac said.

That sentence deserves far more attention than it will receive. In a war where the Pentagon has extended deployments specifically to preserve the President's options, where the President declines to say what would end the campaign, and where Tehran promises a "new strategy" it will not describe, a Filipino able seaman holds the only documented, individual, legally recognised right to decline.

He is also the person least positioned to use it. A refusal is unpaid. Filipinos make up roughly a quarter of the world's merchant crews by most industry counts, and their remittances are load-bearing for the national economy. A right exercised alone, by one man, against a manning agency he will need again next contract, is not a veto. It is a resignation with additional paperwork. It becomes a real right only when it is exercised collectively, or when the war-risk bonus is large enough that saying no is a choice rather than a sacrifice.

Cacdac said he intends to meet Bahri's senior management personally, and called the firm "a very good employer of Filipino seafarers." Both the meeting and the compliment are the diplomacy of a state that cannot order a ship to stop.

What the risk is worth to everyone else

The market has already priced what these crews are absorbing. Earnings on the Baltic Exchange's benchmark very large crude carrier route reached a record $647,000 a day. Mizuho estimates insurance alone adds $5 to $15 to the delivered cost of a barrel. Brent sits near $95.50.

Every one of those figures is a mechanism for paying somebody to accept a risk. The person actually accepting it has not been asked what it is worth.

If you crew, man, insure or charter through the Gulf: write to us. We want to know the war-risk bonus on your contract this month, and whether anyone aboard your ship has said no.

MERSEBURG — Two million people in a mostly rural eastern state will vote on Sunday, and the German political establishment has spent the summer preparing for the possibility that the vote will be decisive rather than merely worrying.

Alternative für Deutschland has polled around 40 per cent in Saxony-Anhalt, with recent surveys placing it between 40 and 43. Its lead candidate, Ulrich Siegmund, is targeting an outright majority in the state parliament. If he gets one, he becomes premier, and a party classified as unequivocally right-wing extremist by the state's own domestic intelligence service will hold significant governing power in Germany for the first time in the postwar period.

The incumbent coalition of CDU, SPD and FDP, led by Minister-President Sven Schulze, is running on economic stability and education reform. Siegmund is running on a 100-day programme that includes "remigration," an end to "political correctness," the abolition of inclusive schooling, and expanded Russian-language instruction. The BBC has reported the party's plans to change how German history is taught in the state, and its pledge to cut funding for what it calls anti-German art.

The contingency file

What is unusual is not the polling. It is how far the mainstream response has already moved past persuasion.

Politico reports that officials in Berlin and several state capitals are weighing operational containment. The most powerful instrument is Bundeszwang — federal coercion under Article 37 of the Basic Law, which allows the federal government, with Bundesrat consent, to take "necessary measures" to compel a state to meet its legal obligations. Less dramatically, security services are considering what one official called "Chinese walls" inside shared intelligence systems, barriers that would withhold certain information from a state government deemed Moscow-friendly. And national leaders have discussed whether transfers Saxony-Anhalt receives under Germany's interstate fiscal equalisation system could be cut on rule-of-law grounds. Officials concede the legal hurdles there are high.

Meanwhile the AfD has been preparing to receive power rather than merely win it. Party parliamentarians from Saxony-Anhalt, Siegmund among them, travelled to Salzburg early this year for instruction from Austria's FPÖ on running a state civil service. In April, national AfD leaders and top candidates from three states voting this September met a former FPÖ state secretary in Berlin for governing advice, including on how to persuade other parties to work with them. The party has created a Working Group on Government Participation to draft legislation and to plan legal resistance to obstruction.

The thing the party's own strategists fear

Buried in the reporting is the detail that complicates every containment plan: AfD's intellectual leadership is reportedly more worried about the ordinary difficulty of governing than about anything Berlin might do. Potholes, school administration, permits and payroll are unforgiving. A state government is judged on garbage collection.

That leaves the mainstream with a genuinely hard choice, and only three days to make it. Every tool in the contingency file is coercive, and every coercive tool confirms the party's founding grievance — that the establishment will not accept it even when it wins.

The firewall was built as a coalition norm: a promise by other parties not to govern with the AfD. It was never designed for the case where the AfD does not need them.

Kevin Warsh went to Jackson Hole and said the quiet part in the ordinary voice of a central banker: the Federal Reserve's focus belongs on price stability, inflation data are "concerning," and the labour market is "broadly consistent with full employment." He did not promise a hike. He removed the excuse for not delivering one.

The repricing was immediate. CME FedWatch showed a 64.2 per cent probability of a quarter-point increase from the current 3.50 to 3.75 per cent target range on Wednesday. A week earlier the same tool showed a 63.4 per cent chance of no change at all. Futures had put the odds near 35 per cent before the speech. J.P. Morgan Wealth Management abandoned its on-hold base case for 2026 and now pencils in 25 basis points this month. Barclays and Société Générale see a September move followed by a second in December. Evercore ISI reversed.

The committee is not unified. July's hold passed 9-3, with three members dissenting in favour of an increase, and the minutes recorded another group prepared to tighten if inflation did not come down. Core PCE, the Fed's preferred measure, was last 3.7 per cent above a year earlier. August payrolls arrive Friday; August CPI on 11 September, four days before the meeting opens.

What the instrument reaches

Here is the awkward part, and it is not a criticism of Warsh so much as a description of his tools.

The inflation currently doing the damage is not coming from American demand. It is coming from a naval chokepoint. Brent is near $95.50, up more than thirty per cent since 28 February. Insurance costs add an estimated $5 to $15 to the delivered price of a barrel. Four commodity vessels transited the Strait of Hormuz on Tuesday against a pre-war average of 24. State Street's Michael Metcalfe attributes the bond selloff to exactly this: energy prices driving hike bets, wrapped up in longer-term fiscal anxiety. Treasury yields sit near multi-year highs.

A 25-basis-point increase in the federal funds rate will not clear a mine, escort a tanker, or restart QatarEnergy's cargoes. It reaches American mortgages, business credit lines, and the government's own interest bill. It cools a demand side that is not the problem in order to offset a supply side the Fed cannot reach.

Central bankers know this. They do it anyway, and the reason is not stupidity. It is the second-order variable: expectations. A supply shock becomes a durable inflation only if people start behaving as though it will. Warsh cannot produce a tanker. He can produce evidence that the Fed will not accommodate.

The cost of proving it

That evidence has a price and a date. Raising rates on 16 September — seven weeks before congressional elections, into a war-driven energy shock, against a President who wants cheaper money and a rising debt-service bill — is a public statement that the administration has not contained inflation and will not get help. Warsh's alternative is to hold and be branded inconsistent within two weeks of his own speech.

He has essentially bought an option on his own credibility. The strike date is 16 September. The premium is paid by anyone refinancing this autumn.

A downside surprise in next Friday's CPI would let him out of the trade. That is a lot of institutional weight resting on one print.

There are two numbers on the Strait of Hormuz this morning and they are not describing the same thing.

The first is official. American forces escorted 40 vessels carrying 18 million barrels of oil through the waterway, and President Trump said on Wednesday that the United States controls the strait and is "helping bring lots of boats out every day with millions of barrels of oil."

The second is commercial. Preliminary Kpler data show four commodity vessels transited on Tuesday, down from ten the day before, against a ten-day average of about 13 and a January-February average near 24 crude tankers a day. Iran has expanded its list of vessels it considers non-compliant and subject to fines, confiscation or detention.

Both figures can be true. An escorted convoy is a discrete event that clears a batch of ships; a daily transit count is a rate. The first is a policy output. The second is a market input. Traders are pricing the second.

Why the tape wins

A convoy is an achievement. It is not a shipping lane. Charterers do not book cargoes against the possibility that a navy will assemble a column on a date they cannot predict, and underwriters do not withdraw a war-risk classification because a government describes a waterway as controlled. What the market needs is the boring thing: a sustained daily count, independently observed, trending back toward 24.

Until it gets that, the premium stays embedded. Mizuho puts the insurance component alone at $5 to $15 on the delivered cost of a barrel. Earnings on the Baltic Exchange's benchmark VLCC route have hit a record $647,000 a day — a figure that is not a freight rate so much as a hazard payment with a shipping label. Lloyd's List Intelligence has counted roughly 200 internationally trading crude and product tankers effectively stranded in the Gulf.

Brent traded around $95.20 to $95.50 on Thursday, easing marginally after Trump indicated the renewed strikes would be short-lived. WTI was near $90.80. The session before, crude moved in a $3 range on nothing but the question of whether Tuesday night's exchange would produce another one.

The asymmetry

This produces a market with an unpleasant shape. Good news is verbal and arrives from podiums. Bad news is physical and arrives as a fire on a tanker. Every peace headline compresses the premium; every mine restores it. Traders have learned to fade the podium.

The useful discipline for readers is the same one we are applying on the front page: distinguish what an institution announced from what an institution's own operating data implies. Forty escorted vessels is a claim about capability. Four transits on a Tuesday is a claim about behaviour.

When those two numbers converge, the war premium goes away. Not before.

A Texas television spot this cycle opens with a person typing into a chatbot: "Why is my electric bill going up?" An Ohio ad blames a Republican senator for the proliferation of server warehouses and calls them "a total scam." In the top House battlegrounds of Michigan, Wisconsin and Pennsylvania, every Democratic challenger is running against data centres.

Sixty-one days from the midterms, the most capital-intensive build-out in the American economy has become a campaign liability, and the polling behind it is not close.

A Gallup survey in March found 71 per cent opposed to local data centre construction. The Annenberg Public Policy Center measured opposition in respondents' own areas rising twelve percentage points between February and the summer. A Reuters/Ipsos poll found 64 per cent saying it is not a good thing to build data centres rapidly, 77 per cent worried about their electricity rates, and 14 per cent willing to live near one. Goldman Sachs projects roughly a six per cent national increase in electricity bills over the next year, concentrated near the facilities.

The cross-party character is the notable part. A Washington Post analysis of candidate posts from 2024 to 2026 found both parties moving sharply negative, with Republicans more negative than Democrats. In Florida, Byron Donalds spent more than $2m on a single primary ad promising to protect the state from data centres that "jack up utility rates." In Ohio, Jon Husted introduced a Ratepayer Protection Act while maintaining that siting should remain a local decision. In Maryland, three Calvert County commissioners lost their seats over the issue, and a Frederick County council president finished third of four in his primary after a permitting vote; a citizen referendum with more than 20,000 signatures was kept off the November ballot by a state Supreme Court ruling on 30 June.

The federal answer is a hearing requirement

On 22 July the House Energy and Commerce Committee advanced H.R. 9340, the Ratepayer Protection Act, by 52 votes to nothing. Unanimity on anything touching this industry is worth noting. What the bill actually does is narrower than the vote implies: it directs state regulators to hold a hearing on whether data centres should pay the full cost of the generation and transmission upgrades built to serve them. It does not require them to adopt the standard.

That sits alongside the administration's voluntary ratepayer protection pledge, which covers 263 million Americans and roughly 80 per cent of US electricity. Critics note there is no enforcement mechanism, and that about a fifth of US power delivery falls outside the pledge entirely, where the old cost-spreading logic applies in full. Oregon's POWER Act took the harder route and wrote a large-load tariff into statute.

What is actually being allocated

Strip out the politics and the fight is about who holds an option on future electricity.

Anthropic has contracted roughly 350 megawatts in Nueces County, Texas, under a reported $35bn deal, plus about 460 megawatts in West Virginia. Neither energises before 2027. The company has purchased certainty about power it will need in eighteen months. The residential customer on the same grid has purchased nothing and, absent a binding tariff, will discover the price at a rate case.

That is the asymmetry every one of these ads is gesturing at, in fewer words. A hyperscaler can hedge fifteen years of power. A household cannot hedge a month of it.

The first real test arrives when a pledge signatory files for interconnection in a state whose commission disagrees about cost allocation. Every utility commissioner in the country will be watching that docket. So, now, will a great many voters.

Two Anthropic announcements landed within days of each other, and the interesting thing is not either one. It is the difference in their tenor and their tense.

The first: a reported $35bn, six-year cloud agreement with Lambda, the Nvidia-backed provider, for roughly 350 megawatts at a Nueces County, Texas campus being developed by the bitcoin miner Hut 8. Bloomberg reported the deal on 31 August; the Wall Street Journal put the megawatt figure to a person familiar. Lambda installs Nvidia chips; Nvidia holds the underlying lease from Hut 8 and supplies the silicon. Initial energisation is expected in the first quarter of 2027. It follows an approximately $45bn, six-year agreement with UK-based Nscale, reported 26 August, for about 460 megawatts in West Virginia running Nvidia's Vera Rubin chips, online by end-2027. None of the companies has confirmed terms on the record.

The second, on 31 August: Anthropic said it had resumed external cybersecurity testing of its models after introducing new safeguards, roughly a month after pausing evaluations following security incidents disclosed on 30 July.

Two clocks

One of these is a commitment measured in years and gigawatt-scale infrastructure that cannot be reversed once concrete is poured. The other is a process that was switched off for about four weeks and switched back on.

That asymmetry is structural, not scandalous. Capital expenditure has to be contracted years ahead because transformers, turbines and grid interconnects have lead times. Evaluation regimes are software and process; they can stop and start. But it produces an odd governance shape: the company's compute posture through 2027 is now effectively fixed, while the assurance layer sitting on top of it is a monthly-cadence question.

The corporate structure around the Texas deal has the same double-booked quality. Nvidia sells the chips, leases the building, backs the provider and is a shareholder in the customer's competitors' suppliers. The one input in that arrangement that cannot appear on two balance sheets is the 350 megawatts.

The financing question underneath

Anthropic confidentially submitted a draft Form S-1 on 1 June, four days after closing a $65bn Series H at a $965bn post-money valuation, with run-rate revenue reported to have crossed $47bn in May. As of late August no public S-1 had appeared on EDGAR, so there is no confirmed date, ticker, share count or price. Bloomberg has reported the company expects to match or exceed SpaceX's $75bn offering. Treat all of that as people-familiar reporting, not company disclosure.

What the public filing will eventually have to contain, and what no press report can supply, is the risk section: how much of the multi-year compute commitment is take-or-pay, what happens if a campus energises late, and what an interruption to safety evaluations means for a company whose product differentiation is partly that it runs them.

Those are the numbers worth waiting for. The megawatts are already spoken for.

VENICE — The 83rd Venice Film Festival opened on Wednesday evening with a film about a man who worked out, in 1969, that a newspaper could be a machine for producing feeling. It is distributed in the United States by Netflix.

Danny Boyle's Ink, adapted by James Graham from his own Tony-nominated play, follows Rupert Murdoch's poaching of Larry Lamb from the Daily Mirror and the launch of The Sun. Jack O'Connell plays Lamb; Guy Pearce plays Murdoch; Claire Foy plays Jules Davies. It drew a five-minute standing ovation and warm reviews, with Deadline calling it thrilling and reading it as a warning about the present.

Boyle was careful at the morning press conference not to "monsterize" his subject, describing the young Murdoch as a breath of fresh air against the existing order. He was less careful about the economics of getting the film made: "We made it with not a lot of money, and it wasn't easy for all sorts of reasons. And then you think nobody's going to see this; it's about Britain, and it's about Fleet Street. Nobody knows what Fleet Street is anymore. You say to everybody you're making a film called Ink. They say, 'Is it a film about tattoos?'" It opens in cinemas on 11 December and lands on Netflix on 8 January.

Earlier in the evening, George Clooney accepted an honorary Golden Lion from Laura Dern and spent it on a subject other than himself. "Stories are always dangerous to people who peddle in fear," he said. "The people who are afraid of books and music and stories, they're never the heroes in history. So yes, movies don't stop wars. They don't lower inflation. They don't fix elections, but they do do something remarkable. They remind us that a stranger is never a stranger and that he might just be the hero in somebody else's story."

It is a good speech, and the honest clause is the middle one. Movies do not stop wars. It is worth noting that this is the concession an actor makes in a room full of people whose entire professional apparatus is built on the opposite assumption.

The rest of the lineup

Twenty-one films compete for the Golden Lion, decided by a seven-member jury under Maggie Gyllenhaal and announced on 12 September. Two of the twenty-one are directed by women: May El-Toukhy's Woman Unknown and Naza, the Gaza documentary co-directed by Rachel Szor and Yuval Abraham of No Other Land. "There's a systemic problem," Gyllenhaal said. "It's harder for women to get films financed."

Gyllenhaal was also asked whether the wars in Gaza and Ukraine and American politics should be allowed into the frame, and gave the maximalist answer: films are "fundamentally, unavoidably political," no subject off limits, including the most corrupt and most perverse people, "because it opens up the opportunity to find empathy for those people."

Still to come: Robert Pattinson in Lance Oppenheim's Primetime, Martin McDonagh's Wild Horse Nine, Florian Zeller's Bunker, Lee Chang-dong's Possible Love, Casey Affleck's Company, Werner Herzog's Bucking Fastard, plus new work from Guadagnino, Schrader and Wenders, a Laura Poitras short on ICE protests in Minnesota, and a documentary on the Oasis comeback. Studio titles are largely absent.

A festival that opens on the birth of the tabloid and closes on a Gaza documentary is not being subtle about its own thesis. It does not have to be.

There is exactly one participant in the Hormuz war who possesses a written, individual, legally acknowledged right to decline, and he is an able seaman from a country that is not a belligerent.

The Philippine Department of Migrant Workers repeats it in every briefing: Filipino seafarers have the right to refuse to sail into designated warlike areas, and that refusal must be respected. Secretary Hans Leo Cacdac said it again on Wednesday, hours after confirming that two Filipinos died aboard the tanker Sidr.

I want to be precise about why this is admirable and why, in its current form, it is not enough.

It is admirable because nobody else has anything like it. The Pentagon extended deployments to preserve the President's options; no soldier was consulted. Iran's security chief promises a "new strategy" and does not describe it to Iranians. The commercial fleet obeys charterparties. Into that, Manila has inserted an individual with a veto and stated in writing that the veto is real.

It is not enough because of who holds it and what it costs to use.

A refusal is unpaid. It is made by one man, in a manning agency's office, against an employer he needs again in six months, in an industry where roughly a quarter of the world's crews come from his country and where his remittances are structural to the national economy. He is being asked to make a foreign-policy judgement, alone, priced entirely to his own household. Call that what it is: not a veto but a resignation, filed on grounds his family will have to live on.

Look at what everyone else is paid to take the same risk. Earnings on the benchmark very large crude carrier route hit a record $647,000 a day. Insurance adds $5 to $15 to the delivered cost of a barrel. Brent sits above $95. A whole industry has monetised the danger of the strait with considerable sophistication, and the one person physically inside it holds an option with no premium attached.

We have discovered the cheapest naval escort ever devised. It is a manning contract.

So make the right expensive to override. Three things, none of them exotic.

First, a statutory war-risk floor. If a flag state or a labour-supplying state designates a zone as warlike, crew sailing into it should receive a mandated multiple of base wage and a mandated death benefit, published, non-negotiable, indexed to the underwriter's own war-risk rate. If the hull is insured at three to ten per cent of value, the crew should not be insured at goodwill.

Second, collective refusal. An individual right of refusal exercised individually is defeated by the labour market. Exercised through a union or a labour-supplying state acting for a whole cohort, it becomes a genuine instrument. The International Transport Workers' Federation exists. Manila has more leverage over the world's oil than any of its ministers has been willing to spend.

Third, publish the refusals. Not names — counts. How many Filipino seafarers declined a Hormuz voyage last month, how many were re-berthed, how many were quietly not rehired. That number would tell us more about this war than the transit tape does, because it measures what people do when the choice is actually theirs.

The serious objection: price the risk properly and some cargoes stop moving, oil goes up, and the cost lands on people who are not rich either. True. It is also the argument made against every safety cost ever imposed on shipping, and the answer has not changed. The cargo was never cheap. The cheapness was a transfer, and the people it was transferred from were on the deck.

Three Filipino seafarers have been killed since 28 February and one is still missing. Roughly 3,000 remain in the Gulf. They are, at present, the only volunteers in this war who were told they could leave and cannot afford to.

Let me concede the worst case immediately, because arguments of this kind are usually made by people who skip it.

The Alternative für Deutschland is classified as unequivocally right-wing extremist by Saxony-Anhalt's own domestic intelligence service. Its 100-day programme includes "remigration," the abolition of inclusive schooling, and a rewriting of how German history is taught. It has pledged to cut public funding for art it deems anti-German. Its lead candidate is polling around 40 per cent, three days out, aiming at an outright majority. If he gets one, a party of that description will govern a German state for the first time since 1945.

I am not going to argue that this is fine. I am going to argue that Berlin's contingency file would make it worse.

The file, as reported, holds three instruments. Bundeszwang under Article 37 of the Basic Law, which lets Berlin compel a state to meet its legal obligations. "Chinese walls" inside shared intelligence systems, withholding material from a state government considered Moscow-friendly. And cutting the interstate fiscal equalisation transfers a poor eastern state depends on, on rule-of-law grounds officials admit face high legal hurdles.

Consider how each reads in Merseburg on the Monday after the vote.

Article 37 has essentially never been used. Deploying it against the first government a state has elected in defiance of national preference would convert a constitutional emergency brake into an ordinary tool of partisan containment, available in perpetuity to whoever holds Berlin. The AfD's Working Group on Government Participation is already drafting litigation strategies. It would not merely survive that fight. The fight is the product.

The intelligence walls are more defensible and, narrowly, probably necessary; a premier who wants Russian-language expansion in schools is a real counter-intelligence question. But partial sharing has a way of becoming a documentable list of things the federal government will not tell a legitimately elected state government. That list will be read aloud at rallies.

The fiscal threat is the worst of the three. Withholding equalisation money punishes two million people, most of whom did not vote AfD, for the votes of those who did. It hands the party the most effective grievance in eastern German politics — that the west sends money as a favour and withdraws it as discipline — with a signed receipt.

Every one of these instruments confirms the same sentence: they will not let you have what you voted for. That sentence is the party's entire growth strategy. Berlin is preparing to write it for them.

The alternative is the one AfD's own intellectual leadership is reportedly most worried about. Let them govern.

Not as abdication — as a test with published results. A state government is judged on road resurfacing, hospital rosters, teacher vacancies and whether the trains run. Siegmund has promised to abolish inclusive schooling; that policy has parents, those parents have children with disabilities, and the outcome will be visible by spring. He has promised expanded Russian instruction; the staffing requirement is a number. He has promised remigration; the constitutional limits on a state government there are severe and will become publicly severe the moment he tests them.

So the federal response I want is the opposite of coercion. Pre-commit, publicly and in advance, that Article 37 will not be invoked for anything short of a demonstrated breach of federal law — and then invoke it without hesitation the instant there is one. Keep the equalisation money flowing, and say why. And publish, monthly, a plain-language ledger of what the state government has done: budgets moved, posts unfilled, contracts awarded, court rulings lost. Make governing legible rather than making it a martyrdom.

The firewall was always a coalition norm — a promise by the other parties not to govern with the AfD. It worked for thirteen years, but it was designed for a party that needed partners. It has no answer to a party that gets 40 per cent, and it never did.

The replacement for a promise is not a wall. It is a record. Berlin should start keeping one on Monday.