Week of June 8, 2026
Last week, my Uber driver at BWI and I watched a woman screaming at the car she had just exited because she was at the wrong airport. Prince, my driver, told me wrong addresses had become a common feature of his work. Our conversation moved from there. He is from Nigeria; he loved meeting people, and he valued the flexibility of self-employment. The conversation was so good, I did not notice my AirPod case slip from my pocket as I exited his car.
For the next two days, I watched the case accompany Prince on his rides. Baltimore, Annapolis, Washington, D.C. Apple's Find My puts a small gray dot on a map, which is supposed to be the good kind of knowing. This information is not power. A lost object is mourned once and replaced; a tracked object is grieved on a refresh timer. Watching my case come asymptotically close to my house, then away again, is a livestream of my impotence.
I sent texts through Uber's app. I made the case chime, and I had it broadcast "I'm lost" to every nearby iPhone (presumably Prince's riders). Finally, after I told Uber's human customer service team I intended to show up at the apartment complex where my case had spent the night, they gave me Prince's number. One phone call later, we connected, and I learned Prince did not understand what he was looking for.
My case came home Saturday afternoon. Prince tried to refuse the thirty dollars I pressed into his hand, a fraction of Apple's replacement charge. My Nigerian Prince turned out to be the exact opposite of a scammer: a kind man hustling around Baltimore who simply didn't know what an AirPod case looked like.
The Google Notebook LM AI-generated podcast version of this week’s newsletter:
Science and Technology Trends
Sometimes the intellectual rabbit hole is evident, and sometimes the rabbit hole finds you. “Ultra High Specific Strength by Bayesian Optimization of Carbon Nanolattices” intrigued me enough to dive into the world of 3D-printed nanoscale carbon lattices. This is the challenge of building high-strength, low-weight materials by engineering and scaling more organic structural forms through iterative machine learning. In other words, while the headline is technically correct, machine learning can help design carbon scaffolding that has the compressive strength of steel and the weight of Styrofoam.
They've only done so at the scale of a grain of rice so far. Nevertheless, machine-learning-based iterative engineering will lead to greater advances in composite and restructured materials.
Article:
AI-assisted review of the article:
JAMA published a patient-focused FAQ on Ebola, a useful reminder of why it stays near the top of everyone's list of things to avoid. The handout makes me wonder which patients I would share this with. Each patient has their own comfort spot on the freak-out/inform spectrum.
In a moment of pareidolia, I was seeing a far happier, rave-going version of the virus. I question how the JAMA authors rotated and presented the Ebola image (credited to the CDC). ChatGPT helped me bring my bias to life:
CDC Update site with a horizontal (and less anthropomorphic-inducing) version of the image:
Anti-Anti-Science
Dr. Eric Topol interviewed Helen Pearson, a science journalist and an author who recently published the book "Beyond Belief: How Evidence Shows What Really Works". I have not yet read the book, but the interview addresses many of the points I often discuss in this section, including how one assesses the quality of medical evidence and considers how we know what we know about science. The interview hits some of the high points of the book, including a discussion of the trade-offs of meta-analyses. It is a succinct review of how we know what we know, the difficulties of interpreting data, and examples of how evidence can both amplify and overcome logical fallacies.
AI-assisted summary of the Interview
Related: loyal readers shared a recent story about individuals who cannot or will not weigh risk against reward, even when based on clinical evidence. The Idaho Department of Health and Welfare reports an outbreak of Campylobacter among 60 individuals who drank raw milk from two different dairies.
Here is what the medical evidence shows (paraphrased from an OpenEvidence review of raw vs. Pasteurized milk):
- In the U.S., unpasteurized dairy products cause 840 times as many illnesses and 45 times as many hospitalizations per serving compared to pasteurized dairy products.
- The most commonly implicated pathogens are Campylobacter, Salmonella, E. coli O157:H7, and Listeria monocytogenes. Some of these cause severe, organ- and life-threatening illness.
- The nutritional differences between raw and pasteurized milk are statistically detectable but clinically minimal for most vitamins, as milk is not a primary dietary source of the affected vitamins (B1, C, folate). The exception is vitamin B2 (riboflavin), for which milk is an important dietary source, and where a small reduction occurs with pasteurization.
- The available data suggest that raw milk consumption in early childhood is associated with some protection against allergies and asthma. However, this effect may be confounded by other farm-related exposures, and the responsible components appear to be heat-labile whey proteins and ω-3 fatty acids rather than raw milk itself.
Louis Pasteur (after whom Pasteurization is named) captured my sense of this story best, in a line often attributed to him: "It is the microbes who will have the last word.”
Review of the data on [why no one should be drinking] raw milk:
NEW The J&E Random Kidney Facts of the Week (JERKFoW!)
Two loyal readers recently suggested I salt the newsletter with random kidney facts for the nephrologic edification of all. They may have mentioned they were upset at me for “not telling them this before.” So, to avoid any further surprises about renal ignorance among my other readers, I will offer this section going forward.
Did you know I once cared for a patient with five kidneys? Most people who receive a kidney transplant keep their two native kidneys, since failed kidneys rarely cause trouble, and removing them only adds surgery time and risk. The new kidney is placed in the pelvis, classically on the right, leaving the left side open for next time. So the common picture is three kidneys: two originals and one graft. When a graft fails, surgeons usually leave it in and place the next one on the opposite side, making four. A third transplant, tucked in alongside an earlier one, makes five. Native kidneys only come out for specific reasons, such as polycystic kidneys that are too large to fit in the abdomen. Otherwise, the original kidneys stay, quietly retired, while the transplants do the work.
The National Kidney Foundation - 20 Questions about Kidney Transplants: https://www.kidney.org/kidney-topics/kidney-transplant#kidney-transplant-surgery-how-it-works
AI Impact
JAMA+AI published an interview with Dr. Nigam Shah (Chief Data Scientist at Stanford Health Care), who explores the epistemic and heuristic challenges of mining EHR data. Each medical specialty has its own data needs and uses. This interview captures a lot of what I spend my time thinking about: “The essence of the problem is how do I retrieve the relevant portion of the [patient timeline]? And right now we're pecking around.” He explores the struggle to organize medical information into a linear narrative, contextualized for each clinician - essentially, the right summary, in the right place in the workflow, at the right moment of care. Dr. Shah discusses ChatEHR - a Stanford-hosted LLM that is now reading FHIR-formatted medical data from the Stanford data systems. This system is still under testing, but it certainly speaks to the kinds of work that return more time and attention to clinicians and give patients a physician who has a better understanding of their data.
Interview
AI Analysis of the interview transcript:
Things I learned this week
I learned that in 1780, Benjamin Franklin wrote a satirical essay questioning the value of "theoretical academic discussions and questions" by submitting a proposal to the Royal Academy of Brussels, arguing that, among the many questions worth asking, the Academy might better study which foods would improve the smell of human flatulence. Essentially, he advocates exploring practical topics that matter to more people as a more valuable use of everyone's time and intellectual energy. Intellectual inquiry unmoored from practical reality has long been a source of frustration. Go Benji F.
The source letter from the National Archives
AI analysis of this proposal (I like where Claude points out B.F.’s potential conflict of interest - intellectual one-upmanship):
Headline of the week: “Cockroach Kingpin In Australia Caught With 100,000 Illegal Insects In Record Bust. The Madagascar hissing species is one of the world's biggest cockroaches, measuring 2 to 3 inches in length.” There is demand for exotic cockroaches? There is a “kingpin” breeding these to meet this “market demand”? Wanting to purchase large numbers of Madagascar hissing cockroaches seems like it should generate a whole bunch of life-choice questions.
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 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 cheerful, densely-populated picture-book town where well-dressed animals go earnestly about a week that happened to include a wandering AirPod case, a cockroach kingpin, and a microbe with the last word. Paste the prompt below into your image generator of choice.
Clean hands and sharp minds,
Adam
Comments
Post a Comment