Thanks to a loyal reader's invite, I attended an "unconference" in Boston last week. Instead of formal presentations, attendees rotated between focused small-group discussions, building relationships and trading views across disciplines. My fellow attendees work in fields far different from mine, and I spent much of the first day quietly calculating whether anyone noticed I was the group's least interesting person. At one point, someone mentioned they had changed their favorite color this year, a small and radical declaration. Adults do not generally announce to strangers that they are still a work in progress.
At the same time, my younger son was finishing his summer research and enduring two days of delays, detours, and cancellations on his way to visit friends and get home, including being ejected from two consecutive commuter trains for traveling with a bicycle during rush hour. Through phone calls and texts, his calm and his faith in travel success never wavered. His father is not such a resilient traveler when plans go awry. I had never seen my son as a Homeric archetype before, but his struggle to get home, ending at 1 a.m. on the morning of his 20th birthday, was Odyssean.
Last week was filled with intellectual yoga (without baby goats). Colleagues searching for answers, pushing my mind to stretch. My son, quietly becoming a heroic figure I don't fully recognize. Weeks that broaden perspective are exhausting.
Christopher Nolan's version of the Odyssey is clearly on my mind. It was fantastic, though I do not recall the original poem being about a man wrestling with PTSD. Claude helped me render a version of my son's journey in the dactylic hexameter of classical poetry. Link: A College Student's Odyssey (annotated, because of course it is.) Odysseus took ten years to get home. My son took two days, which, at nineteen, is the same thing.
P.S. I have permission from my younger son and the person referenced above to comment about their words and actions.
P.P.S. The line in the College Student Odyssey about the "sirens of Baltimore" probably deserves its own sociology thesis.
Gemini Notebook AI-generated podcast version of this week's newsletter.
https://wair.ajwein.com/podcast-2026-08-10-v1
Science and Technology Trends
A note before we start. This week's stories point more inward at scientists rather than outward at science's critics: a press release that oversold its findings, a research team that omitted a child's death, and a third that guardrailed its own technology before anyone asked. Skepticism of a claim is not blanket skepticism, but one can see how it is easy to generate fear, uncertainty, and doubt about what we think we know.
Most of us have viruses living in our bodies kept in check by our immune system. Epstein-Barr almost certainly, cytomegalovirus probably, and a family called Anelloviridae that lives in 80-90% of people. Almost nobody outside transplant patients ever gets tested for them.
A new Nature paper makes a strong case that severe COVID, independent of immunosuppression, can allow reactivation of these viruses. Researchers followed 1,154 patients hospitalized for COVID across 20 US hospitals between May 2020 and March 2021. Every patient was vaccine-naive, most of them because the vaccines did not yet exist. Investigators collected nasal swabs, peripheral blood cells, and, in ventilated patients, endotracheal aspirates, at up to ten visits across the following year. About half of the subjects had other chronic viruses actively transcribing their genomes (reproducing) in the first 40 days after their COVID diagnosis. Interestingly, each virus kept its own schedule. EBV peaked at hospital admission and faded; CMV and HSV1 showed up around week three.
Two findings stood out. First, immunosuppression does not fully explain this. Our long-held assumption is that viral reactivation means a suppressed immune system, but several viruses tracked in this study showed no association with immunosuppressive medication use. It is possible COVID exhausts the immune system, allowing these viruses an easier path to reproduction. Second, Anelloviridae was found months after the COVID infections, most commonly in samples from individuals reporting persistent physical disabilities, which becomes another possible explanation for long COVID syndrome.
None of this establishes causation, but the associations open many doors. It also sharpens the case for staying current on the COVID vaccine, since the vaccine decreases the risk of severe illness and hospitalization, the key events to the findings above.
Article
https://www.nature.com/articles/s41586-026-10740-z
AI-assisted summary
https://wair.ajwein.com/virus-reactivation-long-covid-2026-08-08-v1
I came upon a recent American College of Cardiology media release citing research with hyperbolic language. The press release reported on UC Riverside work examining the records of 246,822 patients with known heart disease, sorted by shingles vaccine status. Individuals who received the vaccine had 46 percent fewer major cardiac events over the following year. The lead author compared the benefit to quitting smoking. The same analysis found those patients were 66 percent less likely to die of anything at all (like car accidents or other seemingly unrelated causes of death). It turns out the source of these data is a conference poster, a claim that is not yet peer-reviewed.
When an intervention appears to prevent death from every cause at once, the data are describing the patients, not the intervention. This is the healthy user effect. In this instance, people who voluntarily take vaccines likely obtain other healthcare too.
None of this means the vaccine is ineffective. Wales rolled out its shingles vaccine for individuals under 80, based on date of birth. As such, they created an inherent intervention and control group separated only by their exact date of birth (79-year-olds and younger vs. 80 and older). Over the next seven years, the vaccinated group had about 20 percent fewer new dementia diagnoses, which is an absolute drop of 3.5 percentage points. When the investigators went looking for the healthy user effect, they couldn't find it. No effect on other causes of death. No increase in other preventive care. Just less dementia.
The vaccine is worth getting. However, the Welsh experiment ran on Zostavax, the live attenuated vaccine that has not been sold in the US since 2020. What Americans can get is Shingrix, and while it is the better choice for most patients, the dementia evidence for it is observational, which is the exact problem this whole section is about.
The problem is, press releases like this get transmuted by media and social media, essentially encouraging people to get vaccines for the wrong reasons; a setup that generates skepticism.
Analysis: https://wair.ajwein.com/shingles-vaccine-cardiac-2026-08-08-v2
(Sources referenced at the bottom of the linked analysis.)
Anti-Anti-Science
This week I found a science story that turned out to be an intellectual fractal, with layer upon layer of the same shapes.
In 2023, a software engineer and his wife learned that their four-year-old daughter was developmentally delayed, caused by a single mutated DNA base in a gene called CHD3. The condition is not fatal. Most people who have it live a normal lifespan, and she was on the less affected end of the range, needing to repeat a year of kindergarten. However, her parents were afraid she would never be able to live on her own.
Her father emailed a gene-editing researcher in Shanghai offering to pay for a treatment that did not yet exist. And, over the next two years, the family sent $860,000 to fund a gene-editing treatment's development, including $130,000 wired directly to a researcher's personal bank account. During those two years, the Shanghai researchers demonstrated that the gene-editing therapy successfully altered the genome of mice with an analogous genetic mutation and that when the gene-editing viral therapy was administered to monkeys, it could successfully enter the brain. (Gene editing involves attaching the gene-editing tools to 'neutered' viruses.)
The Shanghai hospital ethics committee approved a trial of the therapy on 2 January 2025. A required toxicology study involving four monkeys is dated 17 February 2025. All four treated monkeys had moderate to severe liver injury, and one also had kidney damage. The committee approved a first-in-human brain infusion six weeks before the safety data existed.
In March 2025, the girl, then six, received the therapy via an infusion into the fluid around her brain and spinal cord, trillions of viral particles carrying instructions to rewrite the mutation in her neurons. She developed thrombotic microangiopathy, a known complication of high-dose viral gene therapy in which the immune system attacks the lining of small blood vessels, and the kidneys fail. She died seven days after the infusion. Two days later, the hospital's own ethics board ruled the death clearly related to the treatment.
In late 2025, the research team published its mouse and other data using the same gene-editing therapy in Nature. The paper did not mention that the family had funded the work. It did not mention that the therapy had been given to a human child. It did not mention that she died. It also did not mention the liver and kidney injuries in the monkeys used for toxicology testing.
No institution involved either identified or called out the problems. The hospital knew in April 2025. The Shanghai district health commission fined it roughly $3,600 in September 2025 for the girl's death, and the responsible physician received verbal counseling. The university reviewed the parents' complaint in April 2026 and issued a written response declining to act. Nature published the study, unaware of the human use or funding. Everybody with the authority to intervene had the facts, opened a file, and closed it.
The girl's parents had recorded their meetings with the researchers because the science was over their heads. They gave the recordings, the bank records, and the messages to reporters at Science and Retraction Watch, who published everything the paper had left out.
The questions raised by this are deep:
That last question is harder than it looks, because the underlying research is good science. The team bred mice carrying the girl's exact mutation, and the DNA editing treatment corrected it. A monkey efficacy study, which used only two animals (and was different from the 4-monkey toxicology study), showed only that the virus reached the monkey's brain. The study could not test whether the virus led to editing the monkey's DNA. In other words, the mouse experiment showed the edit works. The efficacy study showed a delivery method reaches the brain. The girl was where both half-answers were put together for the first time.
The science was not junk, but the translation of that science to humans was. Those whose job it was to notice the gap between the animal data and the plan to treat the child did not call it out. This is why stepwise, gated research is critical. Not to slow down knowledge, but to make each unproven step visible and verified.
These kinds of events undermine health care and create skepticism in my patients (and throughout society). To be sure, this one is confusing, but exemplifies why these stories are so confusing to so many; I needed an LLM to get the sequence of events straight (and I read a lot of science stories). When the scientists themselves don't follow the proper scientific method, it makes the world of novel therapies more challenging for anyone to navigate.
Six days after the reporters published in July 2026, this CHD3 research trial's public registry was updated to state that the participant died of thrombotic microangiopathy, events known as of April 2025.
The Medscape article:
https://www.medscape.com/viewarticle/first-human-gene-editing-trial-what-went-wrong-2026a1000qfp
The Science Magazine article discussing many of these questions is here:
I had Claude put together a full review:
wair.ajwein.com/gene-editing-death-shanghai-2026-08-08-v1
AI Impact
I came across this blog post from Naomi Bashkansky, a Woman International Master in chess and an OpenAI alignment researcher, who left OpenAI in July for a startup called Conduit. Her new role is helping build thought-to-text technology, about which she is very, very bullish (see her idealized timeline for success). Fortunately, Claude detected my incredulity (it said, "Adam, I see you are WTF'ing!") and prepared an analysis of the topic.
Seriously, I was surprised to learn that this technology is more real than I imagined, with many layers of interesting complexity. Meta's Brain2QWERTY project reads the brain's motor activity while users type memorized sentences with a 61 percent word accuracy. The tech, a magnetoencephalograph, requires a magnetically shielded room and a scanner the size of a 1970s salon hair dryer.
A 2025 Cell article went further with implanted electrodes, decoding imagined speech in patients who could no longer move their mouths. It also picked up unprompted sentences, including participants silently counting shapes on a screen. The researchers declined to publish what they read, because they could not be sure it was what the participants were actually thinking. As a result, they also programmed the decoder only to go on when it detects an internally imagined keyword, which they set to "chittychittybangbang." Their software recognized this "on signal" password 98.75 percent of the time.
It's hard to know if the technology will evolve the way Bashkansky predicts. But reading thoughts with AI-driven electrical pattern analysis is not science fiction, and the decoding science is peer-reviewed. Several sources also highlight that the privacy practices of the companies collecting the brain data are not clear. Someone needs to start a fashionable Faraday cage hat and wig business. For the record, the words for this section exited my head through my fingertips.
AI-supported review of this topic
The J&E Random Kidney Facts of the Week (JERKFoW!)
Be edified.
My Coloradan friends live at altitude and may not appreciate the burden of us sea-level visitors. Upon arrival in Denver, or worse, the mountains, a flight of stairs makes us breathe like a 2-pack-a-day smoker. Even fewer of them appreciate what the kidneys quietly do to help.
The air at 5,000 feet gives you less oxygen per breath. Not because the air is different, but because there is less of it pressing on you. The atmosphere is ~21 percent oxygen at every altitude on Earth. What drops as you climb is the barometric pressure, and with it the partial pressure of the oxygen you actually absorb. The air at 5,000 feet feels like 17 percent oxygen at sea level, and at 13,000 feet like about 13 percent.
The body's immediate fix is to breathe faster. But breathing faster blows off carbon dioxide, and carbon dioxide is one of the things keeping your blood's acidity in balance. Lose too much of it, and your blood pH rises, which is to say it turns alkaline.
So the Colorado visitor's altitude experience is really a two-organ affair. The lungs overreact to the thin air, and the alkalosis that follows would throttle that overreaction back down if the kidneys did not start dumping bicarbonate, the body's main acid-base buffer, into the urine. The kidney is not cleaning up after the lungs. It is permitting them to keep going. This is why the first 24 hours at altitude come with copious urination, volume loss, dehydration risk, and, potentially, a less forgiving relationship with adult beverages.
Within 24 to 48 hours, the kidneys start a second and slower project, ramping up erythropoietin, the hormone that builds red blood cells. That raises the blood's oxygen-carrying capacity, which is a more durable fix rather than the emergency one, and it is why endurance athletes train at altitude. This topic deserves its own week, hint, hint.
The part my Colorado friends will never concede is that they went through all of this too. They just did it long enough ago to have forgotten the heavy breathing and the enhanced bathroom appreciation opportunities.
Things I learned this week
Babies and making babies edition
Two marine invertebrates taught me the difference between Kindchenschema, the baby-face features that make us find things cute, and pareidolia, seeing a face where there is none. The first is the sea pig, Scotoplanes, a plump translucent sea cucumber that trundles across the abyssal plain on inflatable legs. The second is the skeleton panda sea squirt, Clavelina ossipandae, a fingernail-sized tunicate whose front end wears what looks like a tiny panda mask. Sea pigs live 1 to 6 kilometers below the ocean surface. Clavelina is the focus of divers' clickbait photos off Japan, which is what turned it into a described species. Two different animals, two tricks played on the same primate wiring: the sea pig is baby schema, all-round body and no face, while the panda is pareidolia, a face assembled out of pigment spots that mean nothing.
Oddly, the one we mistook for a panda is the one actually related to us. Tunicates are the closest invertebrate relatives of vertebrates, and the larva has a notochord and a dorsal nerve cord, the same starter kit as a human embryo. The pig-shaped one is a sea cucumber, and phylogenetically it is nowhere near a pig.
See https://wair.ajwein.com/sea-pig-skeleton-panda-2026-08-08-v1
Other assorted resources to review:
https://www.sciencefocus.com/nature/underwater-panda-skeletons-sea-squirts
Scientific American's Today in Science email arrived Monday with the headline "female daddy longlegs will perform fellatio," leading to an article entitled, "Daddy Long Leg Sex is weirder than you might imagine." I found myself pondering the meeting in which somebody, working at a magazine founded in 1845, chose those words. And why does the author presume I have any imagined thoughts about daddy longlegs sex?
Then I stared at the pluralization. Daddy long legs. Daddy's long leg. Daddy Longs Leg. I may need to ask some attorneys general. In this case, the Venn diagram of good science, prurient interest, confusing grammar, and animal research is a perfect circle.
(As I re-read this article, it appears SciAm published it with two titles, depending on how you find it. The article page omits the word fellatio; the URL, the social card, and the copyright record still do. There were evidently two meetings, and the internet defied efforts to scrub the decisions of the first.)
The Article
and
AI-supported Review
https://wair.ajwein.com/daddy-longlegs-sex-2026-08-03-v1
If the daddy longlegs had an instinct to monetize, they'd learn from these marmots who now have an OnlyFans page. My conspiracy take: the marmots managed to lobby the federal government to cut off their research funding so that marmot scientists would start the OnlyFans page for the marmots. Or maybe the marmot researchers just got creative in the absence of ongoing funding. There is a sad and odd contrast between a family in Shanghai wiring $860,000 to a researcher's personal bank account and marmot biologists opening an OnlyFans for money. https://www.livescience.com/animals/land-mammals/were-trying-this-out-of-frustration-marmots-star-in-onlyfans-account-as-researchers-desperately-look-for-ways-to-raise-funds-amid-us-budget-cuts
One last observation: based on appearance alone, I think the red-lipped batfish should be part of the above sordid tales and inappropriate elementary humor.
https://www.sciencefocus.com/nature/what-is-a-red-lipped-batfish
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.
A 16th-century anatomical fugitive sheet, single large broadside, hand-tinted woodcut on aged laid paper with visible chain lines and foxing. Central standing anatomical figure, arms slightly away from the body, rendered in fine black woodcut outline with muted hand-coloring in ochre, oxblood, sage, and Prussian blue.
Three layered paper flaps over the figure's body, each shown partially lifted and curling forward with soft cast shadows and visible paper tabs, in the manner of a museum photograph of a working flap print. The uppermost chest flap is lifted to reveal, among the organs, a scattering of small ornamental seed-like forms rendered as specimens. A second flap over the abdomen is lifted to reveal two oversized kidneys, with a small engraved mountain range behind them and a fine stream of tiny glyphs flowing downward from the kidneys. A hinged flap over the cranium is closed and secured with a small brass padlock, a key hanging from a ribbon beside it.
A fourth flap over the lower left torso is sewn shut with red thread and sealed with red wax. It is not lifted, and nothing is visible beneath it.
Gallery and Prompt:
https://wair.ajwein.com/wair-art-fugitive-sheet-gallery-2026-08-10-v1
Fugitive sheets were printed broadsides sold to medical students and the curious in the 1500s: a standing anatomical figure with layered paper flaps you lifted to see the organs underneath. They were cheap, they were popular, and almost none of them survive, because the flaps easily fell off after repeated use.