Source
The a16z Show — Mark Hyman’s Guide to Living Longer and Staying Strong (November 7, 2026) ·
at 14:38 ▸
When this applies
when optimizing for one visible metric (e.g., weight, a KPI) without tracking a correlated but invisible cost (e.g., quality/composition), risking a worse long-term outcome than before the intervention
The takeaway
When pulling a powerful lever toward one target metric, explicitly check which supporting structures that metric depends on, and protect those supporting structures in parallel.
Why it works
A powerful lever that suppresses appetite reduces total intake, but without a complementary practice (protein + resistance training) to protect muscle, the lever degrades a hidden but critical variable (lean mass → metabolic rate), producing a rebound problem once the lever is removed.
One level up
Any optimization that ignores its own supporting infrastructure risks a worse state than the one it started from once the intervention stops.
Where it stops holding
This stops working when the supporting structure (here, muscle) doesn't actually degrade under the intervention, or when there's no rebound risk once the intervention ends.
Detail
approach: Identify what structural support a desired outcome metric depends on, then protect that support explicitly during any aggressive intervention.
signals:- Rapid loss of a visible metric (weight) without attention to a hidden structural cost (muscle)
- Eventual rebound or worse outcome after the intervention ends or is stopped
rationale: People on GLP-1s aren't hungry, eat less protein, don't strength train, lose significant muscle, end up with a slower metabolism, and regain weight faster once they stop.
In their words
They're not hungry. They're not eating as much. They don't have much protein. They're not strength training. And they're losing significant amounts of muscle and they're not replacing it. And that means they're going to have a slower metabolism eventually.