During my recent time off, my wife encouraged me to learn mahjong, so I built an app to teach myself on the plane.
I don’t find joy in games. I am annoyed by arbitrary rules (healthcare policy and tech are my day job). I suspect the others know a hidden winning strategy, which gets in the way of social bonding (being picked last in middle school gym class left a mark). And I can’t multitask while playing. (I learned that drinking counts. Sake does not improve my mahj.)
These are moments for introspection, and the line between reconciliation and stubbornness is fine. I have a friend who is a trained opera singer and many friends who are amazing visual artists. My wife can look at 13 randomly selected mahjong tiles and, turn by turn, discern how to reach one of the 1,077 winning combinations. She’s lowkey Bayesian. I am good with words. The app is called Mahjong for the Unwilling.
Gemini Notebook AI-generated podcast version of this week’s newsletter.
Latest episode at the top
https://wair.ajwein.com/podcast
Science and Technology Trends
Peer-reviewed research on GLP-1s is now shifting to sustainability and tactics. I found this Nature Medicine study of British women prescribed tirzepatide and managed entirely through a telehealth service. The two companies that ran the program each have an employee among the authors, and the lead author consults for one of them. Of about 16,000 women screened, roughly 8,100 started treatment, and more than nine in ten were still taking it at six months. Average weight loss was 15.5%, and blood sugar and cholesterol improved, with no serious treatment-related adverse events. Results held across levels of social deprivation. Nonwhite women lost somewhat less (13.5% versus 15.8%) and were more likely to stop. The takeaway is that we (the medical community) are moving from asking whether we should prescribe these medications to the right patients to asking how we can get them to the right patients.
Paper
https://www.nature.com/articles/s41591-026-04687-x
AI-Supported Analysis
https://wair.ajwein.com/glp1-digital-women-shemed-2026-10-17-v1
I have sat through several talks (from former government officials) about satellite warfare. I don't spend time specifically looking for articles in this space; however, I recently found an article from the Center for Strategic and International Studies noting that, in recent weeks, Russia has maneuvered one of its satellites (Cosmos 2612) close to a U.S.-registered commercial radar imaging satellite that supports Ukraine. I was both amused and alarmed by this sentence from the piece, which is stuffed with ten-dollar words: “Close-proximity [satellite] operations could have a benign purpose, such as satellite servicing or debris removal, but might also presage the deployment of a kinetic or non-kinetic counterspace weapon or a hostile grappling operation. “ Somehow I doubt Russia is offering to service our satellite. I don't want to make light of the threat, but I do want to use "hostile grappling operation" in daily life. At dinner, for example, "Pass the pepper, or I will be forced to engage in hostile grappling."
Paper
https://www.csis.org/analysis/could-russia-be-poised-use-orbit-weapon-against-commercial-satellite
Contrast the journalistic verbiage above with another space-related headline I found this week: “Invisible jet of radioactive goo is heading for Earth – should we be worried?” The goo is plasma, fired at about three-quarters the speed of light by a black hole 12,000 light-years away. I am sure the satellite movements above aren't related to its impending arrival, but I love the mix of pop sci-fi, astronomy, and clickbait. Asked whether we should worry, the astronomer behind the finding answered, "No."
Article
https://metro.co.uk/2026/10/17/invisible-jet-radioactive-goo-heading-earth-worried-29702099/
Anti-Anti-Science
On October 5, 2026, the National Institutes of Health opened a clinic to care for and study patients with post-vaccination health problems. I support honest research into vaccine harms. I worry that current HHS leadership (which replaced the entire vaccine advisory committee last year and has a long track record of vaccine skepticism) will shape which questions get asked and how the answers are reported.
I also found a Nature Microbiology editorial arguing that we measure vaccines too narrowly. We ask whether a vaccine prevents infection, hospitalization, and death from the illness it targets. We rarely ask whether it prevents the chronic conditions that can follow. For example, people who got a flu shot have fewer cardiovascular events and, from my other readings, shingles vaccination is associated with lower dementia risk.
https://www.nature.com/articles/s41564-026-02506-0
I had Claude put together an analysis that adds context on the past six weeks of vaccine news from HHS.
https://wair.ajwein.com/vaccine-second-axis-2026-10-17-v1
My take: the long-term, large-scale analyses needed to find rare vaccine harm use the same tools as those needed to find delayed vaccine benefit. I hope the NIH clinic is an honest effort to establish causal relationships for both good and bad outcomes. We will see whether NIH invests the time, rigor, and statistics needed to capture rare, long-term, hard-to-measure events. If it looks as hard for the benefits as it does for the harms, it will find both.
AI Impact
I will offer only one article this week. Science covered Harvard physicist Matthew Schwartz who, as a visiting researcher at Anthropic, used Claude to explore open questions across a wide range of academic fields. He developed BootLoops, a set of tools that finds problems in one field that a technique from mathematics, physics, or computer science can solve. So far, the result is 36 cross-field manuscripts. Accelerating cross-disciplinary work is exactly what AI is suited to do (says the nephrologist who writes mahjong apps).
It is a clear, readable piece about how AI lets people work well beyond their training. Read it, or listen to it.
Articles
and
https://bootloops.ai/bootloops.html
What caught my attention: among the 36 papers, Bootloops found that CMS ratings programs (star ratings) for Medicare Advantage (MA) plans and dialysis facilities may have some mathematical problems.
https://wair.ajwein.com/claude-shaped-science-2026-10-18-v2
CMS sets cut points in its star rating programs by sorting scores on a measure and splitting the sorted list into five groups (1 to 5 star groups). The federal regulation says the split should make each group as tight as possible. The clustering method CMS uses (Ward's method) merges nearby scores one step at a time and never revisits a merge so it can miss the tightest split. For scores on a single line, the exact best split has been computable since a 1958 paper by the statistician Walter Fisher. Using it would move about one in five Medicare Advantage program ratings and about one in four stars on the dialysis patient experience survey (ICH CAHPS).
In other words, a statistical method older than Medicare suggests Medicare has been grading on the wrong curve. The preprint, which its author labels preliminary, estimates that 34 Medicare Advantage contract ratings would cross the four-star bonus line. The fix was published in 1958. A physicist and a chatbot noted this.
The J&E Random Kidney Facts of the Week (JERKFoW!)
Kidney function is most commonly measured by creatinine, a molecule that is the end product of creatine metabolism in the muscles. Because creatine degrades into creatinine at a steady rate, daily creatinine production is proportional to muscle mass. (To be clear, CREATINE is the energy-rich molecule used by muscles; CREATININE is the muscle waste product.)
The kidneys filter creatinine into the urine. Because muscles release creatinine into the bloodstream at a fairly constant rate, and healthy kidneys clear it at a constant rate, blood creatinine levels are a widely used marker of how well your kidneys are cleaning your blood. When the kidneys aren’t filtering well, creatinine rises not because muscles produce more creatinine, but because the kidneys remove less creatinine from the bloodstream.
It’s like using the water level in a sink to tell how clogged or unclogged the drain is. If you have a constant flow of water from a faucet, you can tell if your drain is clogged by watching how high the water level rises as the clog worsens.
A few caveats:
Clearly, blood creatinine is imperfect, but it is cheap and widely available. As such, we kidney folks use several other tests to understand a person’s kidney filtering ability more precisely.
Ultimately, no single test perfectly measures kidney function, but you can estimate it in many ways. Either way, the greatest value is tracking these data over time. Just be sure to let your doctors know if you are taking creatine supplements.
https://wair.ajwein.com/creatinine
Things I learned this week
Last week, I found data further supporting the unsettling sense that I am living in a Stanley Kubrick film. On Sept. 18, influencer Frankie LaPenna stepped into a San Francisco cage with three humanoid robots. Five days later, the California State Athletic Commission sent a cease-and-desist order to the promoter, Robot Entertainment Kombat, and its CEO, Cix Liv.
I think the real question is, which film? Maybe a mashup entitled “Full Metal A.I. or: A Space Odessey on How I Learned to Keep My Eyes Wide Shut and Love the Shining Clockwork Orange.”
https://ground.news/article/california-officials-challenge-startups-human-robot-cage-fighting-events
and
https://www.nytimes.com/2026/10/08/technology/human-robot-cage-fights-california-rek.html
I learned about Osedax, colloquially known as the zombie worm. This worm has no mouth or intestine. Instead, it secretes an acid that dissolves bone, letting bacteria do the eating. It lives on the bones of dead animals and degrades their skeletons, which annoys paleontologists, since those skeletons are the fossil record. The article I found discussed relatively new research that identified these worms, or at least the holes they leave behind, in the roots of prehistoric shark teeth. The research paper is real, the science press coverage is a bit hyperbolic, and the zombie worm backstory is far more entertaining.
Amusing Zombie Worm Facts:
The paper adds a new line to the worms' résumé: the first fossil evidence of Osedax boring into shark teeth, in the roots of five-million-year-old requiem shark teeth near Pienza, Italy. The worms are far older; their borings appear in 100-million-year-old plesiosaur bones. The press made it sound as though they summered in Tuscany. (“Italian zombie worms say ciao to Pliocene shark teeth.”)
What I found (with a great AI image):
https://scienmag.com/zombie-worms-caught-in-the-fossil-record-boring-into-pliocene-shark-teeth/
The paper
https://fr.pensoft.net/article/167615/
AI-supported analysis
https://wair.ajwein.com/zombie-worms-tuscan-shark-teeth-2026-10-18-v1
Art of the week
A visual mashup of newsletter topics in an exercise to see how different LLMs interpret the prompt. I use an LLM to summarize the newsletter, suggest prompts, and generate images with different LLMs.
This week, a first-person mahjong game on 1920s bone-and-bamboo tiles against an Italian zombie worm, a cage-fighting robot, and a grappling satellite.
https://wair.ajwein.com/wair-art-bone-hand-gallery-2026-10-19-v1
Clean worm tubes and sharp minds,
Adam