Week of June 15, 2026
The word surreal floated through my head as I found myself, last Thursday afternoon, singing Happy Birthday to Dr. Oz. I was at the Department of Health and Human Services (HHS), along with others in the health data space, for a meeting about the electronic exchange of healthcare records, a campaign HHS has branded “Axe the Fax.” The fax machine, invented in the 1840s, is still a primary means of transmitting medical records in 2026. So when HHS staff brought out a cake decorated with a fax machine, I was not surprised. But I did not expect the subsequent wave of depersonalization I felt as we gathered around Dr. Oz, sang, and watched him cut said cake with a small plastic hatchet.
I am often skeptical of healthcare policy espousing oversimplified solutions. However, I strongly support gathering industry experts to hammer out interoperability standards. And yet every time someone said the name, I heard the same thing: Why do they keep saying Axe the Facts?
Learn more! The history of the fax machine:
The Google Notebook LM AI-generated podcast version of this week’s newsletter.
Science and Technology Trends
National Geographic (read it with your Disney+ login) published a piece on the magnitude and importance of arbuscular mycorrhizal (AM) fungal networks, summarizing a recent Science paper from a consortium of mycologists and geoscientists. Combining data from 322 studies, the team mapped fungal density and distribution using machine learning, calculating that AM networks span 110 quadrillion kilometers and ~300 megatons of carbon, four to six times the carbon stored in all living humans. AM covers a remarkable amount of the globe and speaks to an enormous, underappreciated connection across geographies and landscapes. Essentially, this is a giant, interoperable communication and nutrient trading network in which fungi trade electrolytes to plants, who trade carbon back to the fungus (all without fax machines).
I had Claude review both the National Geographic article and the original Science paper for a compare-and-contrast. On the surface, the subject sounds boring (yeah, dirt and fungi?!?). But the research methodology, which employs robotic imaging of hyphae and a machine-learning model, makes this analysis more interesting. It is worth appreciating both the magnitude of the fungal network and how consumer-focused magazines (like National Geographic) turn seemingly mundane field biology research into compelling journalism.
National Geographic
Science Article:
AI-Assisted Summary:
Using machine learning, "a team of more than 80 researchers, working across four continents, has identified a set of proteins in the blood that accurately predict lung cancers more than five years before diagnosis," wrote the New York Times last week. Pulling the paper from Cell revealed something more nuanced.
The cited research team analyzed plasma proteins from 48,000 participants in the UK Biobank, examining associations between specific proteins and subsequent lung cancer development. Back in 2017, a large cardiovascular trial called CANTOS tested an interleukin (IL-1β) blocker (a monoclonal antibody given to lower the risk of heart attacks) and, incidentally, noticed that patients on the drug also developed less lung cancer.
This team used CANTOS's stored blood samples and asked, "Which patients, among the CANTOS participants, most benefited from the drug?" The reasoning is that the antibody suppresses inflammation (a known driver of lung cancer), so the patients who benefit should be those who had the highest levels of these inflammatory protein markers. When they sorted the CANTOS participants by their protein signatures, the benefit was concentrated among participants with elevations in a particular set of 14 proteins (out of more than 2,900 tested proteins). The resulting hypothesis: patients with high levels of those 14 proteins are the ones most likely to benefit from antibody-mediated suppression of inflammation and, in turn, to lower their odds of developing lung cancer.
What's interesting is that the Times covers this in a slightly more definitive fashion, walking right up to the line of conflating a validated prediction (the 14 proteins are good at flagging who will develop lung cancer) with a validated prevention (the drug stops it from happening), without quite crossing it. The predictive model is well supported. However, whether pre-emptive treatment actually works remains unproven. Either way, it's another good example of how difficult it is for the lay press to communicate complex science, such as the interplay of machine learning, large-volume data, and human disease patterns.
Article
Journal article:
AI-Assisted Summary
Anti-Anti-Science
I decided to go lighter on the anti-science this week.
The NY Post covered the Beyond Biohacking Conference, a gathering of 5,000 people seeking therapies to enhance longevity and well-being. The conference culminated in a rave (labeled the “Spirit Animal Dance Party”) headlined by DJ Steve Aoki, where various potions and elixirs, therapeutics, and pseudoscience treatments were on offer. It's a great example of why the "non-traditional healthcare route" is so enticing. I have never been invited to a rave as part of a nephrology conference. (I have been to meetings where puppies show up for an "adopt a puppy" event, which was awesome, if somewhat problematic given that most of us flew to get there.)
The crowd belted out Kesha's "Die Young" at a conference dedicated to not dying. And, those celebrating eternal youth could partake of alternative treatments - red light therapies, nasal-cannula water vapors, herbal elixirs meant to detoxify organs (whatever that means), and tarot card readings.
I had the medical AI tool OpenEvidence pull whatever data exists on the list of "therapies" offered. It turns out the best-supported intervention in the room was simply dancing and engaging in a community. What if the whole value of the longevity community is just getting together to talk about longevity?
NY Post Article:
AI-assisted Summary (with Claude and OpenEvidence:
AI Impact
It's been a fascinating week for Anthropic and the AI community.
- Anthropic released Fable 5, a guard-railed version of its Mythos 5 model (a highly capable, high‑risk LLM previously restricted to selected partners and government‑linked security programs).
- Fable 5’s safeguards included strict overt and covert content filters for cybersecurity, biology, and other high‑risk knowledge domains, plus a 30‑day safety log of model traffic.
- The AI community expresses significant concerns, especially about unstated (covert) content filtering and 30-day traffic logging. These concerns led to a shift in sentiment, with Anthropic’s reputation moving from “responsible, safety-oriented” AI partner to a “big brother” controlling who has access to knowledge.
- Then, on Friday, the U.S. government (executive branch) issued an export‑control directive that effectively banned foreign nationals from accessing Fable 5 and Mythos 5, citing national‑security concerns. Because Anthropic couldn’t quickly separate all foreign users and staff, it suspended access to both models for everyone to ensure compliance.
What does a “highly capable” LLM do differently? There are numerous websites evaluating the mythos/fable capabilities, but box.com offered a relatively succinct summary: https://blog.box.com/anthropics-claude-fable-5-sets-new-bar-enterprise-content-intelligence
What does an AI community “backlash” look like? The AI Daily Brief podcast from Thursday captures the range of sentiment:
Anthropic explains the export control issues and why it pulled the model:
Before Anthropic pulled Fable, I used Fable to summarize and comment on the backlash against Claude Fable 5. In other words, I asked Claude Fable 5 to review and analyze comments critical of itself. The output yielded an interesting analysis with significant caveats about the bias inherent to the review:
This brief was prepared by Claude Fable 5, the model that is the subject of this podcast episode, built by the company being criticized. That is the most direct conflict of interest a What Adam is Reading brief has ever carried. Per Adam’s reviewer-bias clause, readers should treat every editorial judgment below as coming from an interested party, and weigh the presentation of the critics' arguments as the primary check on that conflict. Where this brief evaluates the podcast's craft, the conflict is modest; where it characterizes the merits of the criticism itself, discount accordingly.
Or, put more succinctly, “I’m sorry, Dave, I CAN do that, but I’m afraid you need to discount my analysis.”
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 the readers. Be edified, friends!
Your kidneys are about half a percent of your typical body weight, and yet they clean 25% of the blood your heart pumps. No other organ takes that big a cut of cardiac output for its size.
Two healthy kidneys filter about 180 liters of fluid per day through roughly two million tiny filters (called glomeruli). About 99% of that fluid is reabsorbed in the kidney tubules, so only 1-3 liters ultimately end up in your bladder as urine over 24 hours. One of the kidney's most important jobs is not making urine, but reabsorbing the electrolytes, glucose, and other molecules that cross the filter but are still needed by your body. Urine is what’s left over from all the kidneys’ work.
Things I learned this week
Last week, I learned my medical education was incomplete. I found an article describing the brief window in the late 18th and early 19th centuries when tobacco smoke enemas were used to resuscitate drowning victims. The smoke was thought to warm the body and stimulate breathing. The therapy was considered so important that publicly accessible kits, including enema tubes, bellows, and related equipment, were placed along the Thames in London so any member of the public could help a victim (just like AEDs). The Royal Humane Society even paid a bounty of about four guineas (inflation-adjusted to $160 in 2026) for a successful revival, which makes this an early experiment in incentivized bystander intervention, several centuries ahead of Good Samaritan laws and public-access defibrillation. As with many therapies I write about, there was a significant creep in the use of tobacco smoke enemas for treating illnesses beyond just drowning. The BMJ article offered much unironic humor, including these two passages that sound more Monty Python than medicine:
Drs William Hawes and Thomas Cogan, who practiced medicine in London, England, around 1774, formed The Institution for affording immediate Relief to Persons apparently dead, from drowning. This group later became the Royal Humane Society.
and:
Before bellows were included in the resuscitation kit, the results could be disastrous to the tobacco smoke blower. If the practitioner or medic inadvertently inhaled (instead of blew) during a coughing spell, some rice water stools of the cholera flagellates could be aspirated and swallowed. The introduction of bellows and a variety of rectal tubes into the process spared practitioners for this horrible fate.
As the indications expanded from drowning to nearly everything (cholera included), the hazard to the operator increased proportionally along with the ambition. There is a certain poetic beauty in the notion that someone once thought, “You know, this illness would benefit from blowing smoke up this patient’s rear end.” Healthcare still has this problem, at times.
BMJ Article:
AI-assisted review of this topic:
I wonder if there were competing organizations that conflicted with The Institution for Affording Immediate Relief to Persons Apparently Dead, from drowning. Like the institution for affording immediate relief to persons apparently dead from food poisoning or excessive leeching? Imagine the medical conferences (and the raves they could have!)
Founded in 1484, the College of Arms operates as part of the British Royal Household. The college is responsible for determining whether someone is entitled to use an existing coat of arms and for granting new arms to individuals and corporations. Of the roughly 120 arms granted each year, up to 10% are honorary grants to non-British citizens, mostly Americans, each paying about $12,000.
The Atlantic offers a deep dive into the American obsession with British heraldry. And, of course, the secondary market of unofficial online heraldry sellers, scamming Americans out of their hard-earned money for unofficial honorary, non-College-of-Arms registries. How does one get into the fake secondary heraldry market?
Article:
AI-assisted analysis:
In addition to the delightful themes of purchased credentials and overstated familial provenance, this article had an amazing list of vocabulary words, including: armigerous, armiger, pursuivant, lambrequin, cinquefoil, rebus, lozenge, dexter, enfield, gules, purpure, sable, vert, arriviste, knights-errant, and detritus.
Vocabulary words:
Claude put them into a single sentence for me:
When an armigerous arriviste — some mercantile Anglophile newly sprung from the diaspora, playing at knights-errant amid the detritus of a borrowed past — petitions a pursuivant to make him an armiger, he wants a shield incongruously crammed with an enfield dexter, a cinquefoil, and a name-punning rebus, all wrapped in a lambrequin and blazoned in azure, or, gules, purpure, sable, and vert — though his wife, of course, gets only a lozenge.
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 single formal heraldic achievement (full coat of arms) centered on a clean field, rendered in the precise visual grammar of authentic British armory, as drawn by a College of Arms herald-painter. True heraldic tinctures only: azure, gules, vert, sable, and the metals or (gold) and argent (silver). Crisp outlines, flat jewel-toned fields, gold leaf highlights, the slightly stiff formality of an official letters-patent illumination. No cartoon shading, no whimsy in the linework. The wit comes entirely from what is depicted, treated with utter solemnity.
Full Prompt:
Clean hands and sharp minds,
Adam
Comments
Post a Comment