Our older son turned 21 last week. We took him to a Mediterranean place owned by an Israeli chef, well-known for its food and unique adult beverages. I photographed him being carded over the hummus. Somewhere between that and the entrees, protestors with megaphones appeared outside the restaurant. They shouted slogans and waved signs, some I sympathize with, many I don't. But I admired the reach of their megaphones and the rhythm of their drumming. Smoky baba ghanoush, an unusual drink, and an angry helping of mixed political messaging. A very American cocktail to mark a transition to adulthood.
Gemini Notebook AI-generated podcast version of this week's newsletter. Latest episode at the top.
Listen: https://wair.ajwein.com/podcast
Science and Technology Trends
Several headlines this week cited a new cohort study on healthcare costs among patients taking Zepbound. GLP-1s remain expensive, and the CMS Bridge program now extends coverage to Medicare patients who lack a qualifying indication like diabetes or sleep apnea. Does the weight loss pay for itself?
The study reports that adults over 55 with obesity, or overweight plus a weight-related condition and no diabetes, who started tirzepatide between November 2023 and September 2025 and stayed on it, had $319 per month lower healthcare costs at 12 to 18 months. That figure excludes the cost of tirzepatide itself. The paper attributes it to fewer hospitalizations, though the inpatient signal disappears in the study's own deeper analysis.
The study has several problems. Every author is a Lilly employee or a Lilly-paid contractor. Both cohorts start at 15,843, and the GLP-1-treated arm is down to 1,181 participants by 12 to 18 months. Some of that is real-world discontinuation, and some is that the study ran out of time. Over 60% of the cohort started in 2025, and the data stop in September 2025. Either way, the people generating the $319 are early adopters who stayed on the drug, which is not the population Medicare is likely to enroll.
The Bridge program is funded for eighteen months, from July 2026 through December 2027. The savings in the published study do not appear until month twelve. Only a patient who enrolled in month one of Bridge can reach that window, and anyone starting in 2027 will have aged out of the program before any financial offset is achieved.
The clinical outcomes from these drugs are remarkable. Whether they return anything to the system at current pricing is a different question. It is also worth noting that the patient pays $50 a month and captures none of the $319 savings, which accrues to Medicare and other payers. The Bridge program's negotiated price is $245 a month with the $50 copay, so that is $195 net to Medicare. If the savings are real, the argument that follows is that CMS should fund the drug and run the program long enough to collect them.
Paper: https://dom-pubs.onlinelibrary.wiley.com/doi/10.1111/dom.71250
AI-supported analysis: https://wair.ajwein.com/tirzepatide-cost-offsets-2026-08-27-v2
The FDA approved the 2026-2027 COVID vaccines this week. Pfizer's Comirnaty and both of Moderna's products (Spikevax and the lower dose mNexspike) cleared on Thursday. Novavax's Nuvaxovid, the non-mRNA option now marketed in the US by Sanofi, followed on Friday. All four vaccines target XFG, the JN.1-lineage recombinant nicknamed Stratus. XFG is down to about 6% of cases, but its descendants, including SW.2, still account for roughly six in ten sequenced cases.
Because of more stringent FDA requirements, these vaccines will be harder to get and will almost certainly be used less. The criteria:
Somewhere between 30 and 60 percent of Americans have a qualifying condition, depending on how you count, and one qualifying condition is not exercising.
Attesting to a condition that is genuinely on the list is not off-label; it is the label. The indications list is now broad enough that up to two hundred million people qualify. It is as though on-label and off-label have quietly collapsed into each other.
These vaccines remain safe and effective. Even as infection rates fall, COVID vaccines reduce severe illness, hospitalization, and death. Nuvaxovid has consistently shown a lower side-effect profile than the mRNA products, and there have been signals it offers more durable protection (though this issue has never been settled and I have not seen this year's data).
Get whatever vaccine you can actually get.
https://www.nytimes.com/2026/08/27/well/live/covid-vaccines-fall-2026.html
Anti-Anti-Science
I like finding opinions I disagree with. This week I found myself rapt by the All-In Podcast interview with Eric Weinstein (no relation), a mathematician turned investor and podcast host. The podcast was hard to turn off. Stretching the intellectual glutes and lubricating the knowledge joints, it felt like yoga for the mind. But fifteen minutes in, the yoga class had become a circus sideshow, and I was the contortionist.
His premises are reasonable enough:
The trouble is the slide from observation (peer review as a form of intellectual tyranny, research careers constrained by normative expectations and social pressure) to the claim that the observations are evidence of a cabal. He says elite scientists are doing classified work on breakthrough propulsion, UAPs, and other civilization-altering discoveries. Asked where, he named Renaissance Technologies, the Long Island hedge fund, and then immediately backed off it: "I'm not saying that Renaissance Technologies is that. I am saying you have a secure campus with a story that makes no sense." He also repeatedly says the conventional academic establishment slighted him, which overtly colors his opinions (he sounds bitter).
He comes across as a smarter Joe Rogan, a carnival barker for my guilty-pleasure YouTube channels about aliens, breakaway civilizations, hidden physics, and cryptids, all wrapped in words like "heterodoxy" and physics jargon. His observations are worth having. His speculation is not. Weinstein is, after all, "just asking questions."
https://open.spotify.com/episode/0dZSW4TCSEbn6w3GzLj2mU
To keep an open and critical mind, I had Claude pull together resources that examine the specific claims in this podcast episode and, more broadly, address the current societal zeitgeist where data passes for proof and doing your own research passes for expertise.
https://wair.ajwein.com/weinstein-counterpoise-2026-08-29-v1
AI Impact
I found several articles and a podcast this week discussing AI as both a partner and a crutch in writing and thinking. Writing is thinking. Generative AI is shifting this notion; we are moving from authors to editors, and the move demands a new set of behaviors and disclosures.
I should state my own workflow. This newsletter is filled with my thoughts, my words, and mostly my sentences. Grammarly is my proofreader. Claude helps hone flow, clarity, consistency, and factual accuracy.
On Monday, financier Stanley Druckenmiller published an op-ed in the Wall Street Journal criticizing Treasury Secretary Scott Bessent. The piece contained obvious AI-isms that readers easily detected. Despite the mocking, Druckenmiller declined to be embarrassed, stating, "I write everything using AI now for the same reason I use a calculator when I do math problems. The only embarrassing thing is how many people reached out and told me how much better written this was than normal."
WSJ Editorial page editor Paul Gigot backed him, called AI "a fact of modern life," and printed no disclosure. The AI Daily Brief's episode on the affair is meta-commentary, and it is also the best discussion I have heard of the author-versus-editor question.
Separately, a loyal reader introduced me to ASD-STE100, Simplified Technical English. It is a controlled vocabulary of about 900 words governed by 53 rules, built for aerospace maintenance documentation. The mechanics reading the manuals often did not have English as a first language, and because the manuals had to be translated. A misread manual is a safety problem, so they engineered the ambiguity out on purpose. And now it is maintained by European engineers through the Aerospace, Security and Defense Industries Association of Europe (ASD).
These two stories overlap. Simplified Technical English narrows language deliberately, under a committee's governance and updated about every three years. Druckenmiller's op-ed narrowed his words and thoughts with generative AI to save time.
George Orwell wrote rules that apply to this topic in 1946, in "Politics and the English Language."
LLMs struggle with Rule 1. Models generate phrasing they have seen most often in training. That is why the Journal piece's AI author was detectable. But this story also has a larger philosophical dimension. Social critic Neil Postman also wrote about this type of future. He reflected that George Orwell feared the state taking away words to control thought, whereas Aldous Huxley feared the public would stop wanting them. The likelier outcome seems somewhere between; AI is helping us keep the words and while handing over the sentences to a tool that homogenizes them into a statistical mean of ideas. Nobody bans anything. But, we risk becoming non-authors and bad editors, while congratulating ourselves on using a calculator well.
The op-ed and the reaction
The episode
Simplified Technical English
Foreword to Neil Postman's Amusing Ourselves to Death (Viking, 1985): https://www.exodusbooks.com/Samples/Penguin/1235Foreword.pdf
AI Summary of this whole topic: https://wair.ajwein.com/llm-homogenization-controlled-language-2026-08-26-v2
The J&E Random Kidney Facts of the Week (JERKFoW!)
The summer before I left for college, high school chemistry lab and medical folklore inconveniently merged in my right ureter as a six-millimeter calcium oxalate kidney stone. While I have not birthed any children, the medical folklore that stone passage is equally painful tracks with the births I've seen (thank you medical school).
As described in a recent JERKFoW, urine contains a variety of minerals and molecules, including sodium, potassium, calcium, oxalate, phosphate, urea, uric acid, and bicarbonate. All of these need enough fluid to stay dissolved. (Now is the time to recall your high school chemistry lab on solutes and solvents.) If the urine you make has too high a concentration of the right types of minerals or molecules, or too little water, solutes start crystallizing into solids. Calcium, oxalate, phosphate, and uric acid are the usual culprits. Citrate blocks calcium crystal formation, so too little citrate is bad too. With the right imbalance, a deposit forms in the papilla tissue at the tip of the kidney's inner cone, then erodes through to the surface and anchors there while the stone grows. If that stone breaks off and moves, it enters the ureter, the muscular tube running between the kidney and bladder. The ureter is a few millimeters in diameter. As the ureter rhythmically squeezes on a stone (peristalsis), the pain can be intense, diffuse, and, as medical folklore says, similar to childbirth.
Typically, stones under 5 mm will pass, albeit with discomfort. Stones of 5 mm or greater often need to be removed. Sometimes they "go and get it" (I'll let you Google) using a variety of tools to break up the stone (urology lasers, pew pew).
While kidney doctors don't remove stones, we see chronic stone formers for biochemical prevention strategies. This means measuring the chemical composition of urine, counseling on dietary adjustments, sometimes adding medications that alter the urinary milieu, and, in almost all cases, encouraging extra water consumption to keep the "solution" dilute. This is one exception to the drink-only-when-thirsty advice I wrote about in previous weeks. The stone that never forms is the one that doesn't hurt.
Things I learned this week
We spent a few hours at the Barnes Foundation in Philadelphia with our now 21-year-old son. Albert Barnes earned his medical degree from Penn at twenty, never practiced, and made his fortune instead on Argyrol, a silver antiseptic dripped into newborns' eyes to prevent gonorrheal blindness. Neonatal eye drops bought 181 Renoirs, 69 Cézannes, 59 Matisses, and 46 Picassos, plus Modigliani, Van Gogh, Gauguin, Degas, Rousseau, and a little Monet and Manet for good measure. He was an early patron of Matisse, Modigliani, and Soutine; others he bought based on his taste. The installation is as interesting as the art. Barnes hung the art himself in what he called ensembles: wallscapes of paintings set among hinges, keyplates, kitchen utensils, and Pennsylvania Dutch chests, grouped by line and light rather than by artist or century, and he kept rearranging them for decades. John Dewey, a friend for thirty years and the Foundation's honorary director of education from 1923, supplied the philosophy. The curation is heterodox on purpose. Grouping a Cézanne with a door hinge argues something about looking that no art-history survey does. The wall arrangement is the lesson. Barnes's exact arrangements are still in the galleries today. The art moved to the current building in 2012, and the rooms were rebuilt to the same dimensions, so a window that faced south in his original home faces south in Philadelphia.
https://www.barnesfoundation.org/ and https://newcriterion.com/article/why-the-barnes-foundation-matters/
On September 5th, the 40th Marathon du Médoc runs through the French town of Pauillac, near Bordeaux. The course is the full 42.195 km, but runners stop at 23 châteaux for wine, plus oysters at kilometer 38 and steak at 39. This is not a race for a personal best. The goal is to endure 42.195 km while drinking and eating, while dressed in the year's theme (the 1980s this year). About 8,500 people enter, and roughly 90 percent of them are in costume.
https://www.marathondumedoc.com/en/faq-2/
A blog on the event with great photos: https://www.aladyofleisure.com/marathon-du-medoc/
I quickly found eight or nine more European races that run vineyard to vineyard with food along the way. Sadie, an American expat runner in Portugal and a House Hunters International alum, has compiled a thorough guide to 30 of them across 12 countries, down to medical certificate requirements and bathroom density along the routes. (What goes in must come out.)
https://jetsettingrunner.com/2026/08/05/european-wine-marathons/
My public elementary school's compulsory presidential fitness testing in the 1980s turned me off long running entirely. I was wrongly incentivized. Running so I wouldn't disappoint Ronald Reagan, who I doubt could run an 8-minute mile, was poor motivation. Running for wine, cheese, and French food may have changed my attitude. It is better to run toward something than away from something.
The Presidential Fitness Test exists because a 1955 study found that 58% of American children failed a basic mobility test, compared with 8.7% of European children, which Eisenhower read as a national security problem. https://www.mentalfloss.com/culture/education/presidential-fitness-test-origins-history
The television commercial I was exposed to in the 1980s: https://youtu.be/pcJVGiQ3PEg?si=xPpn1cjGNTeoQWFU
AI art of the week
A visual mashup of topics from the newsletter, and an exercise to see how various LLMs interpret the prompt. I use an LLM to summarize the newsletter, suggest prompts, and generate images with different LLMs.
Claude and I made a Barnes-like installation image of images this week. How meta.
See the full prompt and gallery of images: https://wair.ajwein.com/wair-art-barnes-ensemble-gallery-2026-08-31-v1
A straight-on photograph of a single dense gallery wall in an early-twentieth-century private collection, hung in the manner of Albert Barnes: coarse natural burlap covering the wall, symmetrical arrangement, works grouped by line and color rather than by artist or period, wrought-iron hinges, a keyplate, a spatula and a yarn spinner mounted flat between the frames, and a painted Pennsylvania German chest below the center. Eight framed paintings, each in a different idiom, all in warm gallery light.
Clean hands and sharp minds,
Adam