The Derivation
No finished formula on this site has been through a clinical trial — and the ingredients have been studied for decades. So what, exactly, follows from that? This page sets out the whole deduction: the three steps that follow from published data, the one step that does not, and what the valid steps do and do not license anyone to conclude.
How the combination is calculated →The equation
The engine solves one problem, and it is a counting problem:
maximize | ⋃f ∈ S ( P(f) ∩ Popen(G) ) | subject to S ⊆ F, |S| small
| G | the goals selected |
| Popen(G) | the pathways the published grading rules mark open for those goals |
| F | the catalog the engine ranks over — 68 formulas |
| P(f) | the pathways formula f is published as acting on |
| S | the combination returned |
The objective is the number of distinct open pathways that at least one selected formula acts on. Nothing else. There is no dose term, no pharmacokinetic term, no synergy coefficient and no outcome variable in it, because none of those is in the published method. The same goals always return the same S, because the graph and the grading rules are fixed before the question is asked.
Three steps that follow from the published data
| 1. Coverage | Goal G has published pathways P(G), each with its own page, its grade and its citations. Formula F covers the subset its ingredients carry a published edge to. That is a set intersection: it can be recomputed, not argued with. |
| 2. Strict improvement | If ingredient x covers A and ingredient y covers B with B not contained in A, then {x, y} covers A ∪ B, which is strictly larger than A. This is the answer to why many ingredients rather than one, and it is a theorem rather than a claim. It needs no synergy. |
| 3. Convergence | For each covered pathway, how many of the formula’s ingredients carry an edge to it, and at what grades. Where several do, several separate literatures name the same node. A structural fact about the graph — and it says nothing on its own about whether effects add. |
The step that does not follow, stated before anyone has to ask
From those three it does not follow that a formula produces any outcome in a person. Coverage of a mechanism set is not clinical efficacy. In the dominant evidence hierarchy, mechanism-based reasoning is the lowest tier for deciding whether an intervention helps. Every ingredient in a 200:1 concentrate sits below the dose used in the trial that earned its grade. No combination index, isobologram or factorial design exists for any formula here, and 0 of 68 have been trialed as a finished formula. Anyone who reads this page as proof of effect has read it wrong, and the counting above is not capable of carrying that weight.
Four properties that follow directly
| 1. Determinism | the graph, the mappings and the grading rules are fixed, so the same G always yields the same Popen and the same S* |
| 2. Coverage, not effect | the quantity maximized is the CARDINALITY OF A SET OF PATHWAYS. No dose, absorption, half-life, synergy coefficient or clinical endpoint appears in the expression |
| 3. Selection, not generation | S* is chosen from the pre-existing catalog F. No new formula is created and nothing is compounded per customer |
| 4. The grade constraint is external to the argmax | only pathways that already satisfy the published opening rule enter Popen, so the maximization cannot admit a pathway by finding more weak evidence for it |
One precision, because the fourth property is easy to over-read. a pathway is opened by the evidence base as a whole, not by the ingredients of the formula that later covers it, so a formula may cover an open pathway with a B driver of its own. MEASURED in the generator: no grade filter is applied inside the argmax. Which of a formula’s own covered pathways would meet the lane rule on its own drivers is published per formula rather than assumed.
Any stronger claim -- synergy, finished-product efficacy, pharmacokinetics -- requires additional premises the published method does not supply.
What it does establish
- that a formula acts on a counted subset of the pathways published for a goal
- which ingredient carries each of those pathways, at which graded evidence level, with the study identifiers on the ingredient’s own page
- which published pathways for that goal the formula does not act on, by name
- which of those no formula in the catalog could act on, by name
- that the selection is deterministic: the same goals return the same combination, because it is solved against a fixed graph rather than generated
What it does not establish
- that the formula produces any outcome in a person — coverage of a mechanism set is not clinical efficacy
- an effect size, a time to effect, or a dose
- synergy, in the pharmacological sense: no combination index, isobologram or factorial design exists for any formula here
- that the ingredients in a formula do not oppose one another. The published relation records that an ingredient acts on a pathway and carries no direction, so that check cannot be computed from this data in either direction, and no green answer to it is claimed anywhere on this site
- that an ingredient’s grade describes the finished formula. It describes that ingredient’s published science
The figures, measured
| Pathway-coverage rows across every formula and goal | 20,854 |
| … opened by at least one grade A driver | 9,815 (47.1%) |
| … a B+ driver with an independent corroborating driver | 4,783 (22.9%) |
| … a B+ driver standing alone, which the lane rule does not accept | 3,286 (15.8%) |
| … B or C only, which never opens a lane | 2,970 (14.2%) |
| Goals where the best formula covers every pathway no formula is barred from | 74 of 87 |
| Goal–pathway slots published in total | 587 |
| … of those, slots no formula in the catalog could cover | 35 |
| Graded ingredients with no pathway mapping, which add no coverage | 350 of 940 |
The fourth row is worth dwelling on, because it is the row a catalog would hide. 3,286 coverage rows rest on a single B+ driver, and the site’s own lane rule asks for a corroborating driver before a pathway opens on B+ evidence. Those rows are published, counted and named per formula in the engine rather than quietly folded into the total.
Both denominators, because one of them was wrong
A pathway with no graded ingredient behind it cannot be covered by anything that exists. Reporting a formula against a denominator that includes such pathways understates it; leaving them in the denominator unlabeled misstates the arithmetic. 35 of 587 goal–pathway slots are in that state, so every goal record now publishes the count of published pathways and the achievable maximum, and names the difference.
A worked example, with its own weak spots
Autonomous — Executive Command on the goal Supernormal Stimulus Protection: 14 of 23 published pathways, which is 14 of 17 once the pathways no formula could cover are set aside.
| pathway | drivers | grades | the lane rule, applied to this formula’s own drivers |
|---|---|---|---|
| Mood (serotonin) | 6 | A×1 B×1 B+×4 | a grade A driver opens this lane alone |
| Alertness | 4 | A×3 B+×1 | a grade A driver opens this lane alone |
| Supports BDNF signaling | 4 | A×1 B×1 B+×2 | a grade A driver opens this lane alone |
| Sleep | 3 | A×1 B+×2 | a grade A driver opens this lane alone |
| Social cognition | 3 | A×2 B×1 | a grade A driver opens this lane alone |
| Calm under stress | 2 | A×1 B+×1 | a grade A driver opens this lane alone |
| Executive function | 2 | A×1 B+×1 | a grade A driver opens this lane alone |
| Stress resilience | 2 | A×1 B+×1 | a grade A driver opens this lane alone |
| Calm | 1 | B+×1 | B+ alone — the lane rule asks for a corroborating driver |
| Default mode network | 1 | A×1 | a grade A driver opens this lane alone |
| Learning & memory | 1 | A×1 | a grade A driver opens this lane alone |
| Object recognition | 1 | B+×1 | B+ alone — the lane rule asks for a corroborating driver |
| Sleep hormone | 1 | A×1 | a grade A driver opens this lane alone |
| Stress & cortisol | 1 | B+×1 | B+ alone — the lane rule asks for a corroborating driver |
Not covered by this formula, named rather than omitted: Brain adaptability, Endocannabinoid tone, Sleep pressure. A further 6 are unreachable by any formula in the catalog: Dopamine breakdown, Dopamine clearance, Impulse control, Recognition memory, Social bonding hormone, Social engagement.
Every figure on this page is recomputed from the corpus by a build guard, and the build fails when one drifts. The method is the multi-target rationale described in the literature as network pharmacology (Hopkins AL, Nature Chemical Biology 2008;4(11):682–690, doi:10.1038/nchembio.118); nothing about it is particular to this company except that the data under it is published.
Run the falsification harness · the callable engine · The Limits Register · How evidence is graded