Chapter 42 — Key Takeaways

The one sentence

Somebody is paid at every step between a video about a peptide and a vial in a refrigerator — and the most important thing that payment structure does is not to make anyone lie, but to determine which true things never get said.

The structure

1. The funnel has five steps: attention → interest → low-friction consultation → prescription → subscription. Each step is individually legal and defensible. The assembly is a sales channel with a physician in it.

2. Persuasion is concentrated at the top; money is concentrated at the bottom. Enormous creative effort goes into steps 1 and 2. Step 3 — the one with clinical consequences — is engineered to consume as little of your attention as possible.

3. In commerce, friction is a defect. In clinical care, some friction is the safety system. The scheduled visit, the slow history, the pharmacist's second look, the follow-up that catches a side effect at week six. A conversion metric has no term for a harm that a delay prevented, so it will optimize the delay away without anyone deciding safety mattered less.

4. Chapter 6 §6.4's epistemic laundering has a commercial twin. The claim is laundered as it descends the funnel, and the prescription at step 4 retroactively validates everything above it — a physician was involved. The physician's involvement is real; it evaluated a form, not the claim in the video.

Telehealth: both halves, held at once

5. The access benefits are real. Specialist deserts, travel and time barriers, shame as a barrier to first contact, and consistency of screening. Remote prescribing reaches people who were not being reached, and proposals to eliminate it are proposals to re-impose barriers on the people least able to absorb them.

6. And the structural pressures are real. Revenue tracks fills; the encounter is often brief and asynchronous; nobody owns month three; and the entity advertising, prescribing, and dispensing may be one entity. Vertical integration does not corrupt anyone — it removes the checks that existed because three parties had different interests.

7. Conventional medicine has volume incentives too. The claim is about integration, not about screens. Anyone who reads this chapter as an attack on telehealth has read it wrong.

8. The diagnostic question for any remote service: what would have caused this service to decline to prescribe? If the intake you completed cannot produce a "no," it was not screening.

Affiliates and disclosure

9. Know which behavior the payment is attached to. Per click buys curiosity. Per signup buys the completed intake. Per fill buys purchase. Revenue share buys persistence — your continuing to take something.

10. Disclosure informs. It does not fix three things: it does not change the incentive; it does not undo the selection effect; and it does not touch trust that was built on unrelated content — nor does it survive a clip.

11. The selection effect is the big one. A creator who finds a product disappointing usually makes no content about it. The visible distribution of opinion is not the distribution of opinion. Every individual video can be disclosed, sincere, and accurate while the aggregate picture is badly wrong, because the distortion lives in content that does not exist.

Content and platform economics

12. A qualified claim is a worse piece of content than an unqualified one. Hedges cost seconds in a format priced by the second; uncertainty does not retain; the caveat placed at the end reaches the smallest possible audience.

13. A transformation narrative never has to state a causal claim. The viewer assembles it, which means the viewer cannot be told they were misled.

14. Engagement optimization cannot rank by evidential quality — and this is architectural, not a matter of will. Watch time, completion, shares, and transcript text are features of a file. Evidential support is a relationship between a claim and a literature, and it is not in the file.

15. So the selection pressure is toward confidence, novelty, and personal narrative — precisely the three properties Chapter 5's hierarchy ranks lowest. Over this feature set, the optimization target and the epistemic target are close to inverted.

16. The moderation layer is not an evidence layer. Keyword-based suppression penalizes the precise noun, so what circulates is a vaguer claim about a vaguer object — less checkable, less rebuttable, less linkable to the literature. A system built to reduce harm exerts pressure toward imprecision.

Credentials, regulation, and the compounding case

17. A credential displayed in a monetized channel is doing marketing work its holder may not intend. Authority transfers across topics; scope does not. There is no visual difference in a thumbnail between a clinician speaking within their competence and one speaking well outside it.

18. "Clinicians should not make content" is the wrong conclusion. Silence cedes the field to people with no obligations at all. The right conclusion is narrower: note that the credential is performing two functions at once, and the second happens whether or not it is intended.

19. The absence of visible complaints is close to zero evidence of safety. A referral relationship has a comment section, not a pharmacovigilance system, and the people it went badly for stop posting.

20. Advertising rules attach to what an entity is legally selling. A manufacturer sells a regulated product, so its promotional speech is regulated. A pharmacy sells a service. A supplement seller sells a food-category product. A creator sells attention. The gap is not an oversight; it follows from the structure.

21. The inversion worth memorizing: the absence of an approval made advertising for compounded preparations less constrained, not more — because promotional rules attach to a label, and there was no label to attach them to.

22. A marketing layer built on a regulatory opening migrates rather than dissolving when the opening closes. It is an asset: channels, audiences, creative, affiliates, staff. That gives you a predictive tool for the next gap.

23. And the demand was real. People who used compounded preparations were often responding to a pricing and coverage failure somebody else created. An account that calls them gullible has misidentified the problem.

The skill

24. The five-question read. What is being sold? By whom? At which step of the funnel — including retention, the invisible category? How is the speaker paid, and does the payment depend on my action or only on my attention? And: what would this person have said if the evidence were bad?

25. If the answer to the last one is "nothing — they'd have made a different video," you have found a selection effect, not a liar.

26. A business model is evidence about the process that produced a claim, not about the claim. A conflict is not a refutation (that is the genetic fallacy), and an absence of conflict is not a warrant (unpaid people are wrong constantly). The read narrows your search; Chapter 5 concludes it.

27. Run it on content you agree with. Otherwise you have built a rationalization engine, not a method. This is Rule 4 — never downgrade with distaste — applied to sources instead of molecules.

The four ratings

None of these is a claim about a molecule. They are claims about structures, and they take a population, an endpoint, a reason, and a falsifier like everything else in this book. A rating system that only worked on pharmacology would be a laboratory instrument rather than a thinking tool.

What this chapter deliberately did not do

It named nobody. Not a creator, not a company, not a clinic, not a service, not a platform. Everything here is structural, because a structure you can recognize generalizes to the next one and a villain you can name does not.

And it did not treat incentive as proof of dishonesty. A vendor with a financial interest may sell an excellent product. A creator earning a commission may believe every word. Sincerity is cheap and abundant, and the interesting effect was never lying anyway.