Levels

How much autonomy should medical AI get? Suchi Saria’s answer is essentially: earn it. Verify, validate, then trust. The goal is a clinician-machine team that performs better than either one alone.

Medical AI doesn’t matter because it performs well retrospectively. Suchi Saria’s sepsis system went live across hospitals, saw nearly 90% clinician adoption, and timely response was associated with nearly 20% lower mortality.

An FDA-approved implant can now treat rheumatoid arthritis with just one minute of vagus nerve stimulation a day. Kevin Tracey explains how bioelectronic medicine turns the nervous system itself into a therapeutic target. 📺 Watch on YouTube: youtu.be/x_h6Dd_Ap_o

A cold plunge initially drives a strong fight-or-flight response. Stay in, and heart rate may start falling as parasympathetic activity increases. Kevin Tracey explains why the real physiology is more complicated than “stimulate your vagus nerve.”

HRV is not a direct measurement of “vagal tone.” Breathing, posture, sympathetic activity, irregular beats and even the calculation method can change it. Kevin Tracey explains what HRV can and can’t tell you. 📺 Watch on YouTube: youtu.be/x_h6Dd_Ap_o

Breathing and humming may involve vagus nerve activity. That doesn’t make them “vagus nerve stimulation.” Kevin Tracey on how wellness culture has stretched the term far beyond what the science supports. 📺 Watch on YouTube: youtu.be/x_h6Dd_Ap_o

Taxi and ambulance drivers had the lowest Alzheimer’s mortality among hundreds of occupations in Anupam Jena’s study. The possible link: constant use of spatial memory and the hippocampus. Fascinating hypothesis, not evidence that driving prevents Alzheimer’s.

You take an antibiotic, feel better two days later, and assume it worked. But what if you would’ve recovered anyway? Anupam Jena on the counterfactual problem that can make N-of-1 health experiments misleading. 📺 Watch on YouTube: youtu.be/k9XI7ym7iP8

“Meta-analysis” isn’t a magic word. If the underlying studies are weak, combining them doesn’t create strong evidence. Anupam Jena on why the label matters far less than what actually went into the analysis.

Even a high-quality study should usually change your mind only a little. It may not apply perfectly to you, and another good study may disagree. Anupam Jena on why health evidence should update your beliefs, not constantly overturn them.

As health measurement gets more predictive, we’ll increasingly discover things before we can treat them. Daniel Durand’s question: should medicine assume that knowledge itself is harmful, or should patients decide how much they want to know?

Daniel Durand cites research in 9,000+ people showing whole-body MRI detects cancer in ~1.6% of normal-risk participants. That’s part of the affirmative case. The unresolved question is whether detecting more disease translates into better long-term outcomes.

Whole-body MRI isn’t a replacement for every screening test. Blood cancers may need blood-based approaches, high-risk lung screening still calls for CT, and colonoscopy can actually remove precancerous polyps. Daniel Durand on where MRI falls short.

Most people will eventually get extensive medical imaging, usually after something goes wrong. Daniel Durand argues the more useful future is having a baseline first, so doctors can tell what’s actually new rather than interpreting every finding from a single snapshot.

One drink a week and 14 drinks a week are not remotely the same exposure. Yet some alcohol studies grouped them together, allowing very light drinkers to pull the average toward “healthy.” Tim Naimi explains why that matters. 📺 Watch on YouTube: youtu.be/8GNmcjPfg7U

Is a little alcohol good for your heart? Maybe for one narrow coronary outcome. But alcohol also raises blood pressure, contributes to AFib, and damages heart muscle. Tim Naimi explains why “heart disease” is too broad a label for this debate.

Your body is always burning some mix of fat and glucose — and shifting that mix is more in your control than you think. Diet moves the crossover point. So does fasted training. Dr. Andrew Koutnik on how fuel choice actually works. 📺 Watch on YouTube: youtu.be/FfomxfyCchw

Why did moderate drinkers look so healthy for decades? Many supposed “abstainers” in alcohol studies were actually former drinkers, including people who quit because they were sick. That comparison can make moderate drinking look artificially protective.

Alcohol is a Group 1 carcinogen. That doesn’t mean its risk equals cigarette smoke or asbestos. It means the evidence that alcohol can cause cancer meets the highest standard of causal evidence. Tim Naimi explains. 📺 Watch on YouTube: youtu.be/8GNmcjPfg7U

Tony Wyss-Coray’s team gave young blood to old mice. Their brains showed better memory, new neurons, less inflammation, and more synaptic connections. The implication: factors in blood may actively influence aging, rather than merely reflect it.