I am always examining my supplement list based on perceived and measurable results.
I am not against large stacks per se, but there doesn’t seem to be much to recommend them, and frankly, I am getting sick of the timing and procedures involved in taking large stacks.
Our stacks will vary depending on such things as age and what we are trying to accomplish. At my age, I focus mainly on preventing sarcopenia and dementia.
So, I asked two of my favorite paid AIs, Perplexity Pro and Claude Opus 4.7. What some of the more prominent credentialed health and longevity specialists often mentioned in Rapamycin News are taking themselves. The following are excerpts from the responses. Some of the bold type is added by me.
Perplexity Pro:
Here’s a practical, source-based estimate: the better-known credentialed health influencers in this group tend to take roughly 5 to 15 supplements each, with many clustering around the 6 to 10 range, depending on whether you count only daily items or also occasional/considered items. The number is highest for Andrew Huberman, moderate for Matt Kaeberlein and Mike Lustgarten, and relatively low for Gil Carvalho based on publicly visible evidence.
Claude Opus 4.7:
The range is enormous — roughly 2 to 40+ — and it correlates inversely with how rigorously the individual demands human RCT evidence before acting. Lustgarten, Carvalho, and Kaeberlein are the most skeptical and take the fewest; Sinclair and Johnson the most. Kaeberlein has explicitly criticized large stacks on the grounds that combinatorial effects are more likely negative than synergistic, and he and Lustgarten both publicly reject several of Sinclair’s cornerstone compounds (NMN, resveratrol).
Norwitz is the awkward data point — his December 2025 full stack letter is paywalled on Substack, so his exact count isn’t public. From free content he’s clearly positive on creatine, phospholipid/lyso-DHA-bound omega-3 (krill oil) driven by his ApoE4/4 genotype, and nattokinase; full list would require a paid subscription.
(Why I know longer watch him unlike most of the others mentioned he wants direct payment for his knowledge)
A caveat on sources: most “daily stack” lists on aggregator sites (fastlifehacks, jinfiniti, etc.) are compiled from podcast mentions over many months, so they often inflate counts by including things the person took once or has since dropped. Kaeberlein’s end-of-2025 Optispan update and Huberman’s June 2024 Rhonda Patrick interview are the closest things to primary-source snapshots. Sinclair’s June 2025 Diamandis interview is the most recent direct statement for him.
Finally — and this is probably most relevant to how you’re thinking about your own regimen — the credentialed skeptics (Kaeberlein, Lustgarten, Carvalho) explicitly describe the 15–40 supplement stacks as a category error, and their own practice looks much more like the biomarker-driven, add-one-change-at-a-time approach you already use.
What “category error” means here
When Kaeberlein, Lustgarten, and Carvalho criticize large stacks, they’re not saying “15 is too many, 5 is fine.” They’re saying the entire framework of building a stack is wrong. Specifically:
The framework assumption is additive: each supplement has an expected benefit B_i, so total benefit = ΣB_i, minus some manageable interaction cost. Under this model, adding more supplements with positive EV is always good (or at worst neutral).
The critique is that this model is wrong in at least three ways that make counting the wrong activity entirely:
-
Effect sizes are mostly near zero in humans. Most supplements show benefit in mice or in vitro, modest signals in small human trials, and null or conflicting results in large RCTs. If the true B_i for most items is statistically indistinguishable from zero, then ΣB_i is also near zero, but Σ(side effects + interactions + cost + adherence burden) is emphatically not zero. The expected value is negative, not positive.
- Biological systems are nonlinear. mTOR, AMPK, NAD+, methylation cycles, and oxidative/antioxidant balance are regulated networks with feedback loops, not independent levers. Pushing multiple nodes simultaneously doesn’t produce the sum of individual effects — it produces an emergent response the organism may counter-regulate against, or worse, a genuinely harmful shift. Kaeberlein’s specific claim is that in complex regulated systems, random perturbations are far more likely to degrade function than improve it, for the same reason randomly editing source code is far more likely to break a program than optimize it.
- You can’t attribute effects. Even if you track biomarkers carefully, once you’re on 15 things you’ve lost the ability to say which caused what. Lustgarten’s whole public persona is built around the opposite: one intervention at a time, re-test, isolate the signal. The instant you stack, you’ve traded scientific knowledge for hope.
So the “category error” framing is: asking “how many supplements should I take?” is like asking "how many random prescription drugs should I take?" The number isn’t the relevant variable. The relevant variables are (a) do you have a measured deficiency or a biomarker-validated target, (b) is there human RCT evidence of benefit at the dose you’re using, and (c) are you isolating it enough to know if it’s working. If those three conditions are met, the correct number happens to be whatever number results — often small. If they’re not met, the correct number is zero, regardless of how compelling the mechanistic story sounds.
The single highest-yield correction
If someone wanted one rule that would catch most of the above, it’s the one Kaeberlein and Lustgarten both articulate: change one thing, retest, decide, then change the next one. The kitchen-sink user and the one-change-at-a-time user can end up on identical final stacks, but only the second one actually knows whether each item is doing anything, and can drop the ones that aren’t. The first one is just paying rent on pills.