The person I want to be was not sitting in seat 6C on my flight home from Chicago last Thursday afternoon. Two long days of meetings were behind me. Tired and stressed about organizing pages of notes and next steps, all I wanted was time to think behind the cover of my noise-canceling headphones.
As I sat down, the woman in 6B asked me to wrestle her elderly father's quad cane into the overhead bin. She explained that he'd had a stroke and that she was his full-time caregiver. She apologized in advance: he would likely need her help, and therefore mine, at several points in the flight, to urgently get to the bathroom. I was polite but curt. My headphones and Spotify were beckoning. But she continued. She and her father, in seat 6A, lived on the South Side, and she was taking him to her five-year-old nephew's birthday party in Baltimore. She asked if I lived in Maryland, and my minimal answers did not deter her friendly banter. She handed me her business card, exclaiming that she is the CEO and Commissioner of a girls' basketball league that serves underprivileged elementary-age girls, providing organized recreation and connecting them with mentors and other not-for-profits. I was silently grateful that the flight attendant interrupted us to take drink orders. I donned my headphones while she asked for a Diet Coke. While I was hoping for plane snacks and a small drink, flight 4616 served me a full-sized helping of shame. The flight was short, so upon landing, I handed 6B the cane, smiled, and quickly exited the plane. Did I mention my "contemplative plane music" Spotify list is really good?
At the risk of showing my age and taste, here is my contemplative plane mix: https://open.spotify.com/playlist/37i9dQZF1EIY9UYUhF7f1l
The Google Notebook LM AI-generated podcast version of this week's newsletter.
https://drive.google.com/file/d/1jOqPL2YIiaVWFyZ9Za57hmMdYdX5c4g3/view
Science and Technology Trends
Europe's heatwave pushed several articles into my feed emphasizing that humans have a "Goldilocks" zone of temperature for work, attention, and sleep, and that societies that engineer their way into it tend to outperform those that don't. The clearest case study is Singapore, whose founding prime minister, Lee Kuan Yew (LKY), credited air conditioning as one of the signal inventions of history and made installing it in the civil service his first act in office. My first-world comfort is LKY's strategic infrastructure.
Financial Times (FT) Article that led me down this path:
https://giftarticle.ft.com/giftarticle/actions/redeem/74520b3c-fdcd-416b-bb7c-e1f1b413b7e8
AI Summary on the science of human comfort and the call for AC as a tool for productivity (including in Europe)
https://claude.ai/public/artifacts/dedb9aaf-06fd-47a2-938d-a799c2affcf0
I often ponder the value of crowdsourcing strategies, especially how "fuzzy" clinical data might be resolved from the aggregate opinions of groups of clinicians or, increasingly, groups of large language models (LLMs). Last week, I found a not-yet-peer-reviewed economics paper exploring what happens when you ask people both what they believe and what they expect a group's average belief to be. It turns out this strategy yields more precise predictions than simply averaging respondents' opinions, which is a more typical type of crowdsourced data. Provided the group is genuinely diverse (challenging when using artificial or simulated groups, or groups drawn from a single narrow knowledge domain), the gap between a respondent's stated opinion and their prediction of where the rest of the group's opinion will land can be a valuable tool for sharpening a crowdsourced answer.
Some interesting takeaways: the notion of higher-order beliefs (my opinion, my opinion of others' beliefs, the delta between the two), and how it reinforces the need for carefully vetted samples in crowdsourced data, be it simulated or real human groups.
Paper:
https://arxiv.org/abs/2102.02666v3
AI-Assisted Summary (which covers some of the practical implications of this research):
https://claude.ai/public/artifacts/05e73a8e-acb0-4add-a6ad-7460ba8ab159
Anti-Anti-Science
While scanning social media for articles last week, I came across several posts declaring: "UK Biobank study proves that sunscreen causes cancer." The comments beneath them were a tide pool of logical fallacies, association mistaken for causation, anecdotal testimonials standing in for evidence, and the usual fear-mongering.
The study in question is a genetic study built on data from 472,672 UK Biobank participants, examining variation across 190 DNA-repair genes to identify variants associated with skin-cancer risk in people with documented sun-exposure histories. It was never designed to test whether sunscreen causes or prevents cancer. Buried in the data is a secondary observation: more frequent sunscreen use correlates with higher rates of skin cancer. The authors flag this as "paradoxical" and explain it. Sunscreen users tend to have greater UV exposure to begin with; people often apply too little or fail to reapply; and many adopt sun protection after a skin-cancer scare or because a family history already places them at risk. In other words, this is reverse causation. The high-risk population reaches for sunscreen, so the sunscreen is a marker of the risk, not the risk itself. The authors note that three other studies had already found the same correlation and the same explanation.
None of the posts I found offered any of that context; no mention of the study's design, its purpose, or whether it was built to examine this correlation at all. It wasn't. The paper is a retrospective analysis finding that people who carry certain variants of the FANCA gene (Fanconi Anemia complementation group A, part of the cell's DNA-repair machinery) and have a history of early-life sunburns are at elevated risk of skin cancer. That gene-plus-environment interaction is the real finding.
Here is how it played out on X:
Social media has a unique, hard-to-detect reality distortion field. Pulling this kind of thread takes time and attention. And in this case, a basic science study is weaponized into a conspiracy claiming the exact opposite of what is known about sunscreen.
The original paper:
https://pmc.ncbi.nlm.nih.gov/articles/PMC10840669/pdf/nihms-1929398.pdf
Grok helped me understand the history of this on X:
https://x.com/i/grok/share/2b214252881e49f1a79339ba9731fb8d
AI-assisted analysis of the paper: https://claude.ai/public/artifacts/5c2e04d8-654d-40a8-ac89-1c3ae7d3ac48
A succinct summary of all of the above:
https://claude.ai/public/artifacts/a126138a-8787-47cf-ad15-f40f7e52f1cb
I found a similar blog exploring association/causation fallacies this week. I did not realize there was a body of medical literature in the early 1980s asking the question, "Do hot drinks cause cancer?" (The short answer: they do not, but older people who drink hot drinks are more likely to get cancer). This blog post also explores the many questions that have arisen about the safety of coffee, which is one of my ongoing pet peeve topics. The review of coffee is again comforting:
1) Coffee is safe.
2) Caffeine is (for the most part) safe, and
3) There is data to suggest that coffee reduces the risk of mortality, cardiovascular disease, Type 2 diabetes, depression, dementia, liver fibrosis, various cancers, and chronic kidney disease.
Blog:
https://www.cremieux.xyz/p/do-hot-drinks-cause-cancer
Review of medical literature on the benefits and risks of coffee:
https://www.openevidence.com/ask/33e8e2d2-58f2-424c-b814-b6d8de91b2eb
NEW The J&E Random Kidney Facts of the Week (JERKFoW)
Two loyal readers suggested I salt the newsletter with random kidney facts for the nephrologic edification of the readers. Be edified, friends!
"Alcohol is hard on the kidneys," is an assertion usually delivered with great confidence by someone holding a drink, or telling me they "don't drink a lot and don't understand why they have kidney problems."
In fact, the kidneys are not where alcohol is broken down; the liver does the work. It starts with an enzyme called alcohol dehydrogenase, which converts ethanol (the chemical name for the alcohol we drink) into acetaldehyde; from there, the body breaks it down further into acetate and eventually carbon dioxide and water. The kidneys clear only a small fraction of alcohol unchanged, which is the same fraction a breathalyzer measures coming off your breath.
Alcohol is toxic to the liver precisely because the liver is where it is broken down: the first step of metabolism produces acetaldehyde, a reactive and toxic byproduct, and the liver is exposed to the highest dose of it. The kidneys, by contrast, are not directly injured by alcohol the way people assume. What alcohol does to your kidneys is mess with signals to the plumbing. Alcohol suppresses antidiuretic hormone (ADH), the hormone that tells the kidneys to hold onto water, so while you're drinking, you excrete more water than you take in and often wake up dehydrated.
In reality, the kidneys are not victimized by alcohol so much as made an accomplice to the next morning's hangover.
Things I learned this week
Not all truths are self-evident. I love finding quirky journal articles, like NYU professor Itai Yanai's recently published "How to Give a Bad Talk." Best of all, I learned that audience members who make eye contact, nod, and listen closely are called sunflowers. I love that.
Article:
https://www.nature.com/articles/s41568-026-00954-8
AI-assisted summary:
https://claude.ai/public/artifacts/4dee8039-ad4d-4396-aaf3-813bccf2f418
The Chinese company Yueban introduced its autonomous self-driving Xiaoban toilet to improve accessibility for people with mobility issues (it even has a bidet function). I wonder if their marketing team has explored a broader market amongst the unwilling (to get up from the couch) or those who want their toilet to follow them around, no matter where they are. If you take your robot toilet on a plane, would you need to buy it a seat?
https://www.theverge.com/tech/952441/yueban-xiaoban-self-driving-autonomous-toilet
This thing is literally a toilet masquerading as a robot vacuum, docking station, and all:
https://www.youtube.com/watch?v=7pslcbV5250
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.
Gemini: https://drive.google.com/file/d/1VPAQL2_l9eWscHgnrkfovMO4tIjMKCER/view
ChatGPT: https://drive.google.com/file/d/1IbFVGqTZyhbmH6oEmGimvYRlzUntYncp/view
Perplexity: https://drive.google.com/file/d/1ZnoP02wN41A-MNIAtKT1EavcJ2Q5XJW4/view
Create a single large illustration in the style of an ornate 19th-century statistical infographic plate, the engraved data-visualization tradition of Charles Minard's flow maps, Florence Nightingale's coxcomb rose diagram, and William Playfair's economic charts. Warm aged paper, fine copperplate engraving, hand-lettered serif labels, and a decorative cartouche title. Restrained scholarly palette: ink navy, sepia, oxblood red, and antique gold, with thin ruled grid lines throughout. The plate must look completely authoritative and rigorous, while the things it actually measures are gently absurd.
Full prompt:
https://claude.ai/public/artifacts/aeed3dd0-f687-4c12-997c-43b079fa42c3
Clean hands and sharp minds team,
Adam