Chapter 23 — Key Takeaways

The one-paragraph version

"Nootropic" is not a claim; it is a mood, because it names no endpoint, instrument, or population. Semax and Selank are rationally designed peptides — an ACTH-related fragment and a tuftsin analog, each carrying the same protease-resistant Pro-Gly-Pro tail — registered as medicines in Russia, unapproved in the U.S. and E.U., typically given intranasally with all the uncertainty Chapter 22 attached to that route. Their supporting literature is real and largely inaccessible to most readers of this book, which produces an access-limited ⚠️: a statement about what can be evaluated, not a verdict on the research. Cerebrolysin, a porcine brain-derived peptide mixture, has the opposite problem — an accessible literature that genuinely disagrees with itself, producing a conflict-limited ⚠️. Dihexa and P21 have striking animal data and no completed human trials: ❌ on the Chapter 17 pattern. Noopept is not a peptide, and noticing that matters more than any opinion about whether it works. Underneath all of it sits the chapter's real subject: cognitive endpoints are unusually easy to fool, in five specific ways, and the standard that catches those failures is not a Western standard — it is a human one.

The claims and their ratings

Two ⚠️ symbols, two different findings. If your notes record only the symbol, you have discarded the distinction that matters most — and you will not recognize the news that would resolve it.

Ten things worth keeping

1. A claim that names a direction but not an endpoint cannot be wrong. And a claim that cannot be wrong is not a claim. "Nootropic" joins "anti-inflammatory" (Ch 10) and "immune modulation" (Ch 18) as the third instance of the same structural failure in a third domain.

2. Semax and Selank are one design idea executed twice. An active fragment plus a Pro-Gly-Pro tail that resists peptidase attack. The tail is an admission built into the molecule: the designers expected rapid degradation. It solves half-life and nothing else — not absorption, not distribution, not the blood-brain barrier.

3. "Not approved here" is usually "nobody filed here." A regulatory absence is most often a commercial decision about market size, patent position, and the cost of new trials — made by people who were not evaluating the science. "The FDA has not approved it" and "the FDA looked and said no" are different facts, and only the second is evidence about the compound.

4. The accessibility of a finding to you is not a property of the finding. You cannot conclude the research is poor from I cannot assess the research. You also cannot conclude the research is sound. Both directions, every time — the discipline is only real if it costs you something either way.

5. Rule 4 is under maximum pressure here. Downgrading because a literature is unfamiliar is downgrading on distaste wearing a lab coat. The test is mechanical: can I name the study-quality problem, or am I gesturing at a country?

6. The symmetric point is not politeness — it is accuracy. Publication bias, industry funding effects, selective outcome reporting, and the replication difficulties of the oxytocin literature (Ch 21) are documented features of the most accessible literatures in the world. The standard is the same six questions applied everywhere, plus honesty about which you could not apply.

7. "?" is a legitimate answer. It is not a zero and it is not a pass. Recording it honestly, with the barrier named, is the skill §23.3 teaches — and an access flag can clear without any new science being done, which is why it has to be written down to be noticed.

8. Mixed results after many trials are a measurement, not a blank. They usually mean the effect is small, or conditional, or absent with bias supplying the positives. What they almost never mean is that a large effect has been overlooked. Large effects are not shy.

9. Cognitive endpoints fool you in five specific ways. Practice effects (people improve by repeating tests), expectancy (subjective states follow expectation), instrument multiplicity (test enough things and something moves), baseline dependence (impaired populations show larger effects), and state-dependence (the same person is a different subject at four in the afternoon). All five point the same direction, and they stack.

10. For subjective enhancement, blinding is the study. A compound with noticeable sensations is functionally unblinded (§5.5). The fixes — an active comparator and a reported blinding-integrity check — are cheap and rare, and their absence is informative. And the placebo arm in an enhancement trial is not inert; it is a competing intervention working through motivation, which raises the bar rather than lowering it.

The transferable procedure

When you cannot read the primary literature: state the claim properly first, note your prior, search wider than your default, apply the six questions allowing "?", keep two separate ledgers (what you assessed / what you could not, with the barrier named), rate the assessable ledger only and label the rating's type, then write two triggers — what would clear the access flag, and what would change the rating itself.

When you meet an uncontrolled before/after cognitive result: decompose it. Practice effect, expectancy, regression to the mean, state differences, measurement noise, true drug effect. The first five are systematically positive and they accumulate. A control group subtracts all five at once, which is why it is not a formality but the measurement — and why adding participants to an uncontrolled study fixes nothing.

What would change the ratings

• Semax / Selank: an independently conducted, pre-registered, adequately powered randomized trial with one pre-specified endpoint, reported in full — in either direction. Separately, the access flag would clear on a broad-search systematic review or translations of the principal trials, with no new science required.
• Cerebrolysin: a large pre-registered independent trial in a defined population with one pre-specified primary functional endpoint and a defined treatment window; or an updated systematic review that resolves the current heterogeneity.
• Dihexa / P21: a completed and published phase 1 study, then a randomized trial with a pre-specified cognitive endpoint. One such trial moves either compound to ⚠️ or 🔬 immediately.
• "Nootropic" as a claim form: restate it with a population, an endpoint, and an instrument. It then becomes rateable, and might land anywhere.

Dossier

Field 5 — Human Evidence, filled for the hard case. When a literature is in another language or another regulatory culture, record the claim, the evidence you could assess, the evidence you could not and why (language / indexing / full text / era), whether anyone outside the originating system has replicated, the six questions with honest "?" entries, your rating with its type, and the two triggers. Demonstrated on Semax in §23's dossier section.

The rule, in one line: never let "I could not read it" silently become either "it does not count" or "it must be good."