Week of January 26, 2026
Both our kids are back at college. Between semester breaks and weekend visits, empty nesting continues in fits and starts. This week, the weather, our calendars, and a childless house aligned for our first real empty-nest flex: we last-minute rebooked a trip to avoid the snow. So this week's flex was really an ironic flexodus - a successful run to Denver to avoid winter weather. But now I'm writing in a hotel lobby and missing my espresso machine.
One follow-up - Last week's opening (Sephora is in my future?) generated an unexpected flood of makeup recommendations. Now I need to figure out what makeup the makeup artist used in this picture. (Her artistic talents were very much put to the test.)
Note: The next newsletter will likely be on 2-8-26. I am travelling next weekend.
The Google Notebook LM AI-generated podcast version of this week's newsletter.
Science and Technology Trends
Using data from two large, 30+-year studies involving over 111,000 healthcare professionals, Harvard researchers found that most types of exercise (walking, jogging, running, weight training, etc.) improved longevity. However, individuals who engaged in multiple activities lived 19% longer than those who stuck to just one or two, even when both groups exercised the same total amount. Bearing in mind that the data are self-reported and intensity is hard to quantify, it appears exercise variety matters almost as much as exercise volume.
Article
AI-supported analysis:
National Geographic reviewed the health benefits of reading. They cited Yale research indicating that reading extends life (by an average of 23 months!) and protects against cognitive decline. The study tracked 3600 adults for more than a decade, controlling for education, income, and baseline health. The proposed mechanism is via stress reduction, similar to meditation. Both physical books and audiobooks engage the same brain pathways, and just 10-30 minutes daily is enough to see benefits, making this one of the most accessible longevity interventions available. There is no indication whether working out (with a mix of activities) is synergistic with reading (see above).
Article:
AI-supported Analysis:
The data from this study are not statistically significant, but interesting nonetheless. Scientists trained people to increase activity in their brain's reward center using real-time MRI feedback, then gave everyone a hepatitis B vaccine to see if "thinking positively" could boost immune response. Participants who learned to use mental strategies focused on positive expectations showed greater activity in one brain region (the VTA) and stronger antibody responses to the vaccine. The findings suggest "positive thoughts" could influence how well one's immune system responds to vaccination and offer a potential scientific basis for the power of positive thinking and the placebo effect.
Article:
AI-Supported Analysis:
Anti-Anti-Science
The US Government cancelled and froze many federal research grants over the last year. Nature published an interactive article highlighting how the cuts are impacting various agencies and the contrasting views of Congress and the President.
Article:
AI-Supported Summary
The All-In Podcast interviewed Dr. Oz last week. Despite being a bit of a love-fest, the podcast is a fantastic example of how difficult it is to have a clear, data-driven conversation about healthcare. The hosts and Dr. Oz highlight many of the US healthcare system's common and real problems: rural and economic access issues, unsustainable costs, workforce shortages, and fraud, waste, and abuse. Their solutions are overly simple and highly dependent on technology - wide-scale data sharing (which falsely assumes a minimum of data quality), AI-supported care (which depends on high-quality data and the patient's ability to contextualize their symptoms), and specific medications (like GLP-1s). The frustrating part is that none of their suggestions are bad ideas. However, shaping human behavior is messy and inconsistent, healthcare systems (especially for the chronically ill) are Byzantine, and the most expensive healthcare costs are the end-points of diseases that take decades to develop and often have no "magic bullets" for treatment. Over-simplification, confirmation bias, and omission fallacies are all forms of anti-science rhetoric.
Here is the episode:
I had Perplexity watch the YouTube version of the podcast to offer some analysis:
The Unbiased Science podcast captured why consumer tech driving long-term health choices is so difficult. Their review of food-advising apps, like Yuka, highlights the challenge of context and complexity:
Apps like Yuka aim to make decision-making easier. People use them because they want a simple answer. Scan, score, done. That's the seduction. The problem is that simplicity comes at the cost of accuracy.[...] This oversimplification might feel comforting in an age of information overload, but it often leads to inaccurate or unfair assessments. Understanding why requires a look at what a real cosmetic and food safety evaluation involves and how formal scientific review differs from the snapshot judgments made by apps.
You can read the transcript of the episode here:
Related: Robert Kennedy publicly linked his vocal problems (spasmodic dysphonia) to doses of the flu vaccine he received in the 1990s. There is little evidence linking the flu vaccine to this disorder. Read the article and pay attention to the following fallacies:
- Post hoc ergo propter hoc – inferring causation from temporal sequence alone.
- Argument from anecdote – basing causal claims on personal experience and litigation side-effect lists rather than population-level data.
- Confirmation bias – stopping flu shots based on one plausible (but unproven) match to his condition.
- Appeal to ignorance – treating the absence of disconfirming data as evidence, despite billions of administered flu vaccine doses over decades.
Contrast: Healthcare is noisy because, often, uncertainties outnumber the certainties, and people are desperate for better answers. Occasionally, someone with resources attacks the problem with the intensity of a startup. Sid Sijbrandij, the billionaire founder of GitLab, used his wealth and obsessive engineering mindset to treat his own cancer. This biotech blog details his journey—it's worth the read, if only to marvel at what unequal resources can achieve when pushing scientists and research to solve problems with evidence.
AI-Summary of the events:
AI Impact
Though AI models can pass medical licensing exams with near-perfect scores, the full complexity of real medical work (note-taking, data summarization, billing, and coding) remains challenging. Stanford researchers created MedHELM, a comprehensive evaluation framework. They tested 9 major AI models (like ChatGPT, Claude, and Gemini) on 37 real-world medical tasks. Claude performed best overall at 66% accuracy, suggesting a need for strong governance, task-specific monitoring, and caution when scaling.
Article:
AI-supported Summary:
Nothing makes a wintery weekend more comforting than reading about the risks of malicious AI agent "swarms." This editorial by a Nobel laureate and AI experts describes how autonomous AI agents can now create coordinated "swarms" of fake online personas that manipulate public opinion at scale and sophistication, mimicking human behavior that evades detection, fabricates online consensus, and potentially corrodes democratic discourse. (Imagine if we redirected that swarming intelligence toward sourdough recipes or nuclear fusion.)
Article
AI-Supported Summary:
Things I learned this week
The Ig Nobel prizes are awarded yearly for humorous, silly, or obvious academic research. The 2025 psychology Ig Nobel went to researchers who discovered that positive reinforcement made people feel special (perhaps to a fault) and negative feedback generated the opposite feeling. In their published study, researchers told 364 people they scored above or below average on an IQ test (regardless of actual performance) and found that positive feedback increased not just self-assessed intelligence but also the feeling of being uniquely special. Conversely, negative feedback about participants' IQ decreased both self-assessed intelligence and narcissistic feelings (though narcissistic individuals were better at protecting their self-image even after negative feedback). I think they (inadvertently?) backed into most of what is happening on Facebook.
Ig Nobel Prizes (check out the 2025 Literature winner for documenting his fingernail growth).
Here is the (peer-reviewed?) paper:
A certain juvenile delight drew me to this article; the (rendered) image of prototaxites is...evocative. These 25-foot-tall, branchless, tree-shaped organisms inhabited Earth 400 million years ago. Since their discovery 165 years ago, their biological identity has remained mysterious; the leading theory held that they were massive fungi. This new study (using chemical fingerprinting and microscopy on Scottish fossils) found that Prototaxites lacks fungal molecular signatures and has internal structures unlike those of any known organism. They appear to be an extinct lineage of complex, multicellular life with no modern descendants.
Article with striking image (I think they knew what they were doing):
The scientific paper:
AI-supported analysis. (You can giggle like a first grader and still be educated!)
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 painting in the unmistakable style of a 1980s Bob Ross PBS tutorial. The scene depicts a snowy mountain valley rendered in soft, blended oil paint with visible brushwork. In the background, titanium white snow-capped mountains fade into a hazy sky.
The central focus is a cluster of ancient 25-foot-tall prototaxites—mysterious, branchless, tree-shaped organisms with textured bark-like surfaces in muted browns and grays—towering over the landscape like alien monoliths.
Scattered throughout the valley are Bob Ross's signature "happy little trees," but on closer inspection, the trees are a swarm of AI agents: their foliage contains faint, barely visible humanoid avatar faces with neutral expressions, dozens of them clustered together, blending into the forest. The effect is subtle and slightly unsettling beneath the cozy aesthetic.
Full Prompt with all the details: https://claude.ai/public/artifacts/e6cadd44-efb8-4a61-b2bb-a2f5d6a36652
Grok is my choice this week.
Epidemiologic Realities
This newsletter began during the pandemic. I leave these links here as a marker of 1) how to find resources on incidence and prevalence of various diseases and 2) to remind myself why I wear masks in my clinic and, often, on planes.
Flu, COVID, and RSV are highly prevalent in wastewater samples. Forecasted COVID rates are trending down.
Wastewater Scan offers a multi-organism wastewater dashboard with an excellent visual display of individual treatment plant-level data.
https://data.wastewaterscan.org/
https://data.wastewaterscan.org/
The Pandemic Mitigation Collaborative (PMC) uses wastewater viral RNA levels to forecast COVID-19 rates over the next 4 weeks.
https://pmc19.com/data/
https://pmc19.com/data/
Clean hands and sharp minds,
Adam
Comments
Post a Comment