Thursday, August 6, 2026

The big shift since 8/05: the denial is the confirmation

Yesterday this brief argued the buried lede was the magazine — CNN's Pentagon inventory numbers (THAAD ~80% depleted, Patriots ~50%, ATACMS/PrSM "virtually all" fired) mattered more than the Muscat diplomacy on the front page, and that an empty arsenal, not a change of heart, was why Hormuz was de-escalating.

Today the President denied it. The BBC's lead: "Trump denies US weapons shortage and says information 'leakers' being 'hunted down.'" Two things arrived with it. The American Conservative, on Iran War Day 160, reports him privately frustrated with the Pentagon over munitions shortages — you are not frustrated with your own department over a shortage that does not exist. And National Review, which is not a hostile outlet, ran "The President's Military Options Are Dangerously Limited." Vox asked the question in plain words: "Is the US running out of missiles?"

A denial and a friendly-press admission printing the same morning is not a contested story. It is a confirmed one with an official position attached. The leak-hunt is the tell: you hunt leakers over numbers that are true and damaging, not numbers that are false — a false number is refuted, not prosecuted.

The post-1975 analog is close enough to be uncomfortable. In December 2004 a soldier asked Donald Rumsfeld why troops were scrounging armor out of landfills, and got "you go to war with the Army you have, not the Army you might want." The Pentagon downplayed the shortage for roughly two years. The fix — the MRAP crash program, eventually tens of billions — came only once the political cost landed, and Rumsfeld was gone by November 2006. What we know now that the actors did not: the denial did not buy time, it spent it. The procurement clock started two years late and men died in the gap.

Where it diverges, and this is the part that should worry you more than the shortage: in 2004 the shortfall was armor for one war, and American industrial capacity was not the binding constraint — attention was. Today it is interceptors for two simultaneous wars. The Atlantic Council today: "Ukraine cannot defeat Putin's ballistic blitz by defensive measures alone." That is a think tank saying the quiet part — defensive interception is losing on arithmetic, because interceptors cost more and take longer to build than the missiles they stop. You cannot surge your way out of that inside an election cycle. The 2004 problem was solvable with money. This one is solvable with money and four years.

Hormuz is not reopening — it is being converted into a tollbooth

This is the story of the day and it is being filed under de-escalation.

The Washington Post: "Iran says Hormuz deal with Oman won't fully reopen strait." NPR, more specifically: Iran aims to ban US and Israeli ships from the Strait of Hormuz and charge others a toll. The South China Morning Post confirms the same shape from the Asian side. Trump says he thinks the war will end "pretty soon." Foreign Policy asks the right question outright: "Is Iran Winning the Battle Over the Strait of Hormuz?"

Read the mechanics rather than the mood. A closure is a wartime act — expensive, illegal, and temporary by nature, because the closer earns nothing from it and the world organizes against it. A toll is a peacetime institution. It generates revenue, it implies a schedule, it requires counterparties to acknowledge the tollkeeper's authority, and every dollar collected converts Iran from a wrecker into a stakeholder with an interest in traffic flowing. Iran is not giving up the closure. It is monetizing it and making it permanent, and it is doing so under a headline that reads like retreat.

The analog is not approximate; it is nearly exact. Egypt closed the Suez Canal in June 1967 and reopened it on 5 June 1975 — eight years. The reopening did not restore Israeli access. Israeli ships and Israeli-bound cargo stayed barred, and that ban ended only in 1979, as one clause of a comprehensive peace treaty at Camp David, four years after the waterway itself reopened. The sequence is the lesson: the shipping arrangement came first and cheap, the exclusion of the enemy came last and expensive, and it took a full political settlement to buy it.

If that holds, the US/Israel exclusion announced today is not an opening bid to be bargained down next month. It is the final concession, and it will not move for a maritime agreement, a toll schedule, or a Muscat communiqué. It moves for a treaty or it does not move.

The second half of the Suez precedent is more encouraging and less discussed. Once canal revenue became a structural pillar of the Egyptian budget, Egypt never closed it again — not in any subsequent regional crisis. The toll did what sanctions and naval escorts could not: it gave the chokepoint's owner more to lose from closing it than from keeping it open. If Iran genuinely stands up a toll regime and starts banking the receipts, the strait becomes more stable, not less, and it becomes so for the least idealistic reason available. That is the strongest argument I can make for letting the toll stand, and I think it is a better argument than the people who will reflexively oppose it are going to admit.

The cost is precedent. A toll accepted at Hormuz is a template for every chokepoint with a revisionist state on one bank — Bab el-Mandeb, the Bosphorus, Malacca. Japan is already exposed and already moving: NHK reports declining US crude exports and the knock-on for Japanese alternative sourcing, which is the 7/12 thread (roughly 90% of Japan's crude transits that strait) grinding forward with no good options.

The war widens south while the strait de-escalates

While Hormuz gets the headlines, the shooting is moving. Reuters: Saudi Arabia expects imminent attacks from the north and the south — Iraq and Yemen, which is to say the two theatres where Iran fights through others rather than directly. Bloomberg: the Houthis carried out a deadly attack in Yemen and hit Saudi territory. IRNA puts 58 dead in a strike on Saudi-backed forces in Yemen.

This is the standard shape of a conflict entering its negotiated phase, and it is why "the war will end pretty soon" and "attacks are imminent" are not contradictory statements. As the direct state-to-state exchange winds down, the proxy periphery heats up, because that is where positions get taken that will price the settlement. The 1988 Iran-Iraq endgame ran the same way: the Tanker War and the Gulf attacks intensified as the ceasefire approached, not after it receded.

Also today, and belonging to the same ledger: Al Jazeera reports Israel resuming strikes in south Lebanon with residents saying there is "no red line," while ECFR argues Hamas has agreed to disarm and Israel should now agree to peace, and Al Jazeera Arabic frames the same story as Hamas accepting weapons-confinement in exchange for withdrawal while Israel holds out for retribution. One front proposes to end; two others widen.

Birthright citizenship: the executive routes around the Court

Trump signed executive orders targeting birthright citizenship and "birth tourism" — carried by Reuters, NPR, Al Jazeera, DW and France 24, which is itself a signal about the international read. The Washington Post supplies the frame that matters: "Trump tries new tack against birthright citizenship after Supreme Court snub."

The interesting question is not whether the orders are constitutional. On the Fourteenth Amendment's text they are very likely not, and everyone involved knows it. The interesting question is what the executive gets from an order it expects to lose.

The template is February 2019. Congress appropriated a fraction of the requested border-wall money; the President declared a national emergency and reprogrammed military construction funds instead. The Supreme Court let the spending proceed in 2019 on standing grounds without reaching the merits; the merits went badly in the lower courts; and the policy ended in 2021 because an election ended it, not because a court did. The consequence we now know and the 2019 actors did not: routing around a court works in practice even when it fails in law, because the clock beats the docket. Money moved. Concrete got poured. The constitutional question was never definitively answered, because the political question was answered first.

Apply that here and the orders do not need to survive. They need to create facts — administrative practice at hospitals, passport agencies and consulates — that persist through the years of litigation, and to move the Overton window on a clause that has not been seriously contested since 1898. Judged as law this is weak. Judged as an instrument it is well-designed, and the people treating the likely court loss as the end of the story are making the 2019 mistake again.

Google recentralizes on a founder as the diaspora completes

Yesterday: Hassabis out as DeepMind CEO, and Jeff Dean and Sanjay Ghemawat leaving to found Discovery Loop. Today the FT supplies the other half: "Google shifts AI power back to Brin as DeepMind's Hassabis steps aside."

That is a founder reassuming direct operational control of the core technical function at a company of Alphabet's size, which is rare and always means the same thing — the institutional layer is no longer trusted to move fast enough. It is the same instinct that returned Jobs to Apple in 1997 and Schultz to Starbucks in 2008, with the difference that both of those were rescues of a failing business and this is not. Alphabet is not failing. It is losing people, which is a different disease with a different prognosis: rescues respond to founder attention, and talent hemorrhages mostly do not, because the thing the leavers wanted was authority, and a founder taking authority back is the opposite of supplying it.

The competitive frame arrived the same day and is worse than the org chart. The FT also ran "Alibaba's latest AI model puts it back in the great game." Foreign Policy ran "China's AI Acceleration." Hacker News has Qwen3.8 Max ranked best overall on the agentic index — a Chinese model at the top of a public leaderboard. And ByteDance now forbids distillation of rival models, which is what you do when you believe you are the one being copied rather than the one copying.

Under the consequential-action lens: AMD acquiring Taalas to etch models directly into silicon is the most interesting item in the technology scan and got the least attention. Etching a fixed model into hardware trades all flexibility for enormous efficiency, and it is a bet that model architectures have stopped changing fast enough to matter. If that bet is right, it moves inference economics by an order of magnitude and the industry's cost structure with it. If it is wrong, it is a very expensive paperweight. Either way it is a real position taken with real capital, which is more than most of the AI commentary this week represents.

What others miss

AI has designed new, replicable viruses — and it ran in Persian. BBC Persian today: artificial intelligence has, for the first time, designed completely new and replicable viruses. I found no equivalent in the English-language tier-1 scan. Set aside how it is framed and hold the claim still: the step from generating a plausible genome to a self-replicating organism is the step that turns biosecurity from a screening problem into a synthesis problem, because you can no longer defend by comparing candidate sequences against a list of known threats. That this is a Persian-service story today and an English-language story in three weeks is itself the finding — the regional services routinely carry science that the flagship desks reach late.

Ebola may be mutating mid-outbreak. The Guardian reports the virus behind the large DRC outbreak could be mutating; DW adds that Africa CDC plans to deploy a vaccine developed for a different strain. That combination — a possibly drifting pathogen and a mismatched vaccine — is the specific setup that turns a containable outbreak into a long one, and it is being covered as a routine health item.

Europe's grid is failing on temperature, not demand. Le Monde: heatwave forces power plant shutdowns across Europe. DW: drought threatens electricity supplies. The Daily Mail has tourists unable to cool off because the Mediterranean itself is too warm, and a power cut took out rail across north-west England. These are filed as weather. They are the same story: thermal plants need cool water to reject heat, and when rivers run hot and low the plants derate or stop — so supply fails hardest exactly when air-conditioning demand peaks. That is a structural coupling, and it will repeat every summer regardless of how much generation gets built.

Quick hits

Meta fined $567m in the largest child-safety ruling against a social media company. The Senate committee voted to hold Fauci in contempt. The Trump administration will deregulate Head Start — an antipoverty program created under RFK Jr.'s uncle, which the NYT could not resist noting. Jay Clayton sworn in as DNI. Abdul El-Sayed's Michigan win consolidates, with The Nation running two pieces on whether his centrist opponents will back him — and Politico California reports AIPAC's next target is state senator Aisha Wahab, which is the 8/05 El-Sayed/AIPAC thread moving from federal to state races inside one day. Germany still investigating the explosive drone at the Leipzig cargo airport. US-backed Venezuela transition talks began without María Corina Machado. The US Navy reclassified 19 Virginia-class submarines, which the Atlantic Council calls a breakthrough. Trump announced tariffs on polysilicon — 15%, with price floors — aimed at Chinese solar and semiconductor supply. California's cyclosporiasis outbreak is running "higher than expected" with no explanation. And 567-million-year-old fossils are being read as rewriting the dawn of animal life, which is the rare science story where the number in the headline is the argument.

My read

Two calls, one graded and one new.

The 8/05 call on the magazine was right, and today upgraded it from inference to fact with an official denial attached. I will take the win and name the part I got wrong in emphasis: I read the empty arsenal as the reason Hormuz was de-escalating. Today suggests it is the reason the United States is de-escalating, while Iran is not de-escalating at all — it is converting a temporary closure into a permanent institution. Those look identical for about a week and then diverge completely.

The new call: the toll sticks. Not because anyone concedes the principle, but because the alternative to a tolled strait is a closed one, the American magazine cannot fund a sustained escort regime, and every party that actually ships oil — including the Gulf states now taking fire — would rather pay a fee than wait out a blockade. Expect loud rejection of the toll's legality, quiet payment of it, and a US/Israel exclusion that outlasts the war by years. Suez took four.

What I am less sure of than yesterday: whether "war ends pretty soon" means anything. It is compatible with a settlement and equally compatible with a frozen conflict where the shooting moves to Yemen and Iraq and the strait becomes a permanent revenue line. The second is cheaper for everyone with a vote and worse for everyone without one, which is usually how these resolve.

The open questions I would actually pull on: does a tolled Hormuz stabilize the Gulf the way a tolled Suez stabilized Egypt, or does the US/Israel exclusion make it a permanent casus belli that Suez never was? And how long does the interceptor arithmetic hold before somebody concludes that the only affordable defense against cheap ballistic mass is offense?

The kicker

In a converted monastery in Halberstadt, Germany, an organ is playing a piece of music that began in 2001 and will not finish until 2640. It is John Cage's ORGAN²/ASLSP — "as slow as possible" — and they took the instruction literally: 639 years, chosen because the monastery's first organ was built in 1361, exactly 639 years before the performance started. The piece opened with a rest that lasted seventeen months. Notes are held for years at a time, sandbags weighting the keys, and when a chord finally changes people travel from all over the world to stand in the room and hear it. The last change drew a crowd. The next one is in 2028.

Somewhere in Germany there is a piece of music with a maintenance schedule, an audience that plans its travel around a single chord, and a completion date that no one who started it, or who will attend the next four hundred chord changes, will ever see.