2026-08-19 ATS briefing¶

The global diesel crunch: Russia's refinery crisis meets Middle East turmoil¶

The Ukrainian drone campaign against Russian refineries has done what sanctions never quite managed: it has turned a major fuel exporter into a country scouting for imports. Moscow's ban on diesel exports through the end of 2026, alongside the reported talks with India's Vadinar refinery, marks a reversal that would have seemed absurd two years ago. Russian refined product loadings fell 67% month-on-month in July, with diesel bearing the brunt.

The Strait of Hormuz closure compounds the problem. Two of the world's major refined product suppliers are simultaneously offline, and the shipping industry is discovering that tight product markets do not automatically translate into a tanker boom. The contraction in trade volumes from both regions means fewer cargoes to move, even as refining margins fatten. The conventional product tanker market gets the worst of both worlds: less business, and what little import demand Russia generates is absorbed by the shadow fleet.

The United States is the designated safety net, but the arithmetic is unforgiving. Utilisation rates near 96% and a string of refinery closures leave little room to expand output. American diesel buyers will now compete with former Russian customers in Turkey, Brazil and West Africa for the same limited barrels. The US can redirect supply, not create it.

What stands out is the strategic logic of the Ukrainian campaign. Targeting export infrastructure at Tuapse, Ust-Luga and the pipeline-linked refining hubs has proven more disruptive to Russian state revenues than any price cap. The war economy's fuel supply is now a vulnerability, not a strength. For the broader crisis, the episode shows how easily the physical infrastructure of energy trade can be severed, and how thin the global margin of spare refining capacity has become. The system is one refinery attack away from a genuine shortage, and no amount of rerouting fixes that.

Putin Has Run Out of Moves¶

Source: Project Syndicate

Åslund's diagnosis rests on a familiar liberal assumption: that Putin could simply choose peace and prosperity, and that his refusal to do so is a matter of stubbornness rather than necessity. The budget deficit of 2.8% of GDP, alongside inflation stuck at 6% and growth below 1%, is presented as evidence of a leader who has painted himself into a corner. But the framing misses what the war actually is for the Kremlin. It is not an expensive policy error that can be reversed when the costs become inconvenient; it is the regime's core project, the mechanism through which the Russian elite has consolidated its power and legitimacy since 2022. Ending the war would not restore the pre-war status quo, it would trigger a domestic reckoning that threatens the entire ruling arrangement.

The stagflation figures describe a real constraint, but they describe it from the outside. For the Russian state, the choice is not between a healthy economy and a costly war, but between managing the contradictions of a war economy and managing the contradictions of a lost one. The drone strikes on energy infrastructure and the inflation rate are pressures, certainly, but they are pressures the system has been built to absorb. The real question Åslund's analysis cannot answer is what happens when the war economy itself becomes the only economy the state knows how to run. That is not a problem of fiscal discipline, it is a problem of political survival, and it has no peacetime solution.

The Myth of the China Shock¶

Source: Project Syndicate

The opening of trade with China did not destroy American manufacturing, Michael Strain argues, and the political consensus that it did has become a serious obstacle to clear thinking about AI. Manufacturing output in the US has roughly doubled since 2000 while employment fell by a third, a divergence that points to productivity gains rather than collapse. The jobs that vanished were concentrated in routine production work, and their disappearance tracks automation and capital substitution at least as much as import competition. Strain's point is not that trade was costless, but that the costs were misdiagnosed: the China Shock became a convenient villain for a structural transformation that would have happened anyway.

The political afterlife of that misdiagnosis matters now because AI is generating the same panic. Policymakers who believe trade caused deindustrialization will reach for protectionist responses to automation, treating a productivity revolution as an external threat to be fended off. Strain wants them to see that the manufacturing jobs of the 1950s were never coming back, and that the real task is managing the transition to whatever comes next.

What the article leaves unexamined is who absorbed the costs of that transition. Productivity gains accrued to capital; displaced workers were offered retraining programmes that never matched the scale of the shock. The working class paid for the efficiency of the system, and the bipartisan consensus Strain criticises was the political form of that settlement. If AI repeats the pattern, the question is not whether to resist automation but whether the gains will be distributed any more fairly this time.

A China Strategy That Lasts¶

Source: Foreign Affairs

Cassidy's pitch to Congress is disarmingly simple: the executive branch flip-flops, so the legislature should lock in a durable anti-China framework. The de minimis closure in July 2025 is his model, a rare instance of bipartisan action that survived the changeover. But the asymmetry he cites cuts deeper than he lets on. Four million packages a day entered the US duty free while American exporters faced tariffs on goods worth more than $6.90. That is not a loophole Beijing exploited; it is a subsidy Washington wrote for Chinese e-commerce giants, and closing it merely restores a level field that never existed.

The substance of his programme is administrative modernisation: electronic filing, verified seller data, AI-driven customs screening, codifying the June 2026 executive order on bonds and penalties. None of this touches the structural source of Chinese overproduction, the state-directed buildout of industrial capacity that floods global markets with cheap solar panels and chemicals. Cassidy treats enforcement as if the problem were smuggling, when the real issue is that Chinese firms can profitably export at prices that undercut American manufacturers because their domestic market is saturated and their capital costs are politically determined.

His Western Hemisphere gambit, the Americas Act, offers near-shoring incentives to wean regional economies off Beijing. Yet the same logic that makes Chinese goods cheap makes Chinese credit attractive to governments in the region, and Cassidy's own provisions against corrupt partners concede that the US cannot match Beijing's willingness to lend without political conditions. The contradiction is concrete: Congress can build a durable legal framework, but durability is precisely what Chinese state capitalism possesses in abundance, and no statute can outlast the price advantage that overcapacity generates.

The Abandoned Order¶

Source: Foreign Affairs

The Chinese term lanwei lou, "rotten-tail buildings," is doing heavy lifting in Mark Leonard's argument, and it earns its keep. Thousands of unfinished apartment blocks, homes for people who improvise walls from bed sheets, become the governing metaphor for an international order whose financiers have stopped paying. The American people, Leonard argues, have lost interest in underwriting the rules-based system, and no successor has stepped in to take over the mortgage.

The strength of the piece lies in its documentation of allied denial. NATO's July 2026 summit is the sharpest example: European leaders pledging arbitrary defence spending increases, unconnected to their own analysts' assessments, as tribute to Donald Trump in the hope he will not withdraw. Denmark buys American arms while Trump threatens Greenland. Australia subsidises the U.S. industrial base through AUKUS even as commitments falter. These are not miscalculations; they are acts of faith performed by people who cannot afford to examine their own premises.

Leonard's central distinction between a disordered world and an "un-order" is worth taking seriously. In disorder, actors violate rules they still recognise. In un-order, they act without reference to any common framework. The first implies a shared grammar, however broken; the second suggests the grammar itself has dissolved. His conversations with leaders worldwide, who agree only on their uncertainty, give the argument an empirical weight that abstract theorising would lack.

What Leonard does not ask is whether the old order was ever as rules-based as its self-description claimed, or whether its abandonment by Washington is a betrayal or a culmination. The United States built the system to serve its interests; when those interests shifted, the system became a cost rather than an asset. The allies now discovering their dependence were always dependent. The question is not whether the order can be salvaged but what replaces it when the hegemon decides the rules no longer bind it either.

AI and the New Age of Bioweapons¶

Source: Foreign Affairs

Sherwood-Randall's scenario is precise enough to be useful: a jailbroken model designs a more transmissible H5N1, a cloud lab synthesises it anonymously, and a livestock farm's air-handling system does the dispersal. The weakness in her own account is that the 2024 H5N1 outbreak she cites as the "base case" was naturally occurring, and the federal response she coordinated still struggled with incomplete data and slow industry cooperation. If the system cannot handle a virus that emerged without malicious intent, the gap is not primarily about AI safety controls.

Her proposed answer, "deterrence by resilience," deserves scrutiny. The logic is that if attacks cannot be prevented, the state should build monitoring, attribution, and surge manufacturing so that the payoff for an attacker shrinks. But this converts the entire civilian economy into a target for the purpose of making retaliation credible. The same biomanufacturing surge capacity that produces vaccines in a crisis is the capacity that produces the reagents and equipment an attacker needs. The tools are identical; only the governance differs.

The article's silence on who governs that difference is telling. Sherwood-Randall writes as a former homeland security adviser, and her institutional reflex is to call for more federal coordination, more pre-negotiated contracts, more integration of scientific and intelligence data. She never asks whether the pharmaceutical companies holding those contracts have an interest in maintaining scarcity, or whether the "cloud labs" she wants to regulate are the same venture-capital-backed startups the US has been actively promoting as strategic infrastructure. The state that wants to contain synthetic pathogens is the same state that funds their development.

The deterrence framework itself is inherited from the nuclear age, where a small number of states could be identified and threatened. Her own argument is that AI dissolves that condition, yet she reaches for the same vocabulary. What she describes is not deterrence but a permanent state of emergency management, with the public absorbing the costs of a threat that the innovation economy continues to generate.

Demand for degree apprenticeships in England quadruples in three years¶

Source: BBC News

The numbers are stark enough on their own: 11.3 applications for every degree apprenticeship place this year, up from 2.8 two years ago, while the actual number of positions has nearly halved. But the more interesting figure is the one buried in the middle of the BBC's report - the 8,100 applicants who used the government's platform three years ago. That cohort was small enough to suggest degree apprenticeships were still a niche curiosity. The tripling since then marks the moment they became a mass phenomenon, and the mass phenomenon is precisely what employers cannot accommodate.

The applicants themselves describe the experience in terms that should worry the government. Daniel applied to university through Ucas as a "backup" because the apprenticeship application process was like "applying for a real job". Kieran ended up at Birmingham sharing lectures with the apprentices who beat him for a place, watching them get the same education for free while drawing a salary. The jealousy he describes is not personal envy but a structural recognition: two students in the same room, same lectures, same labs, one paying for the privilege, the other being paid.

What the state offers in response is telling. The DWP's answer is hiring incentives worth up to £8,000 and free training for 16-to-24-year-olds. That is a demand-side subsidy for employers whose stated reason for cutting back is that employment costs are too high. The apprenticeship levy was supposed to fund training, but businesses like Fives say the syllabus cannot keep up with AI-driven changes in work. The training system is too rigid to serve capital's needs, so capital simply withdraws from it, and the state responds by paying them to come back.

The class dimension runs underneath the whole story. Stephen Evans notes you are less likely to get a degree apprenticeship if you come from a working-class family. The students at Hereford Sixth Form College are competing against more experienced candidates for positions concentrated in the South East. The government's bursary for families on Universal Credit addresses the cost of taking up a place, not the barriers to securing one. The route that was supposed to rival university is becoming another filter, sorting by the same advantages that already determine who gets into the best universities.