How this publication decides what to believe

Two marks per claim, a rubric that has been revised three times because real compounds broke it, and a review date on every grade.

About

The Compound Codex reads the published research on peptides, SARMs, hormones, nootropics, longevity compounds and supplements, and reports what it actually says. Every claim gets a mark. Every mark is assigned by a written rule you can read, check, and argue with.


Two marks, not one

Most coverage of these compounds collapses into a verdict — it works, it doesn't, it's a scam. That hides the more useful question, which is how much anyone actually knows. A compound can have excellent evidence that it does nothing, and terrible evidence that it does something wonderful. Those are opposite situations, and one number cannot describe both.

So every claim here carries two marks, assigned independently.

How strong the evidence is
Preliminary
Emerging
Moderate
Strong

Set by counting trials, independent research groups, total participants, and whether a meta-analysis exists. Not by how convincing the findings feel.

What that evidence found Supported Mixed Not supported Untested in humans

The direction of the result, kept separate from its weight. A Strong · Not supported means the question was settled, and the answer was no.

Since version 2.3 a third mark appears where warranted — a safety flag recording where a harm signal came from, not how bad it is: ⚑ Identified in trials ⚑ Documented in case reports ⚑ Restricted by a regulator. Provenance is checkable; severity is a judgement call, and judgement calls are what this system exists to avoid.

The full rule — every threshold, every definition, the four-step procedure for when trials disagree — is published at How we grade the evidence.

The rubric has been wrong three times

A grading system nobody ever revises is not rigorous; it is just unexamined. This one has changed three times, each because a real compound broke it, and each revision is logged with the article that caused it.

Revision history
v2.1 Creatine exposed a scale problem. The original rule could award the top grade on three small trials. Creatine has dozens of trials and thousands of participants; a compound with three studies of twenty people should not sit beside it. A scale requirement was added.
v2.2 Resveratrol exposed a category error. Its central mechanism claim — that it activates SIRT1 — was refuted in a test tube, with no human participants to count. The scale had no way to express that. Mechanism claims are now handled separately instead of being forced onto a clinical scale.
v2.3 Cardarine exposed a safety gap. Its development was halted over cancer findings in rodents — a fact of overwhelming importance that the two existing axes could not carry, since it concerned neither evidence strength nor whether the claim was true. The safety flag was added.

Each of those revisions changed marks on articles already published. Those articles were re-graded and say so. A rubric that only ever ratifies what was already written isn't doing any work.

Every grade has a review date

Evidence moves. A claim graded Emerging today may be Moderate after the next trial reports, and one graded Supported may not survive a larger replication. A publication that grades once and never looks back is producing an archive of things that used to be true.

So every breakdown carries a next review date, and the working file behind the whole site — every claim, with the trial counts, participant totals and independent-group counts that produced its mark — is maintained as a single ledger rather than scattered through the prose.

That ledger also records, for each claim not yet at the top of the scale, the exact criterion holding it there: 442 more participants, or one more independent research group, or replication in a second population. Which means the moment a qualifying trial publishes, it is already known which marks it moves.

What this publication will not do

  • Sell anything it grades. No supplements, no affiliate links, no sponsored breakdowns, no discount codes. The only revenue is reader membership, which is the entire reason the grades can be unfavourable.
  • Recommend doses. Where a dose appears, it describes the protocol a cited trial used. That is a report, not advice, and the distinction is not decorative — several compounds covered here are prescription-only or controlled substances.
  • Grade on mechanism. A plausible pathway is not a result. Compounds with elegant mechanisms and no human trials are graded Preliminary, however good the story is.
  • Repeat a figure it cannot trace. If a number cannot be followed to the paper it came from, it does not appear — and the article says openly that it was left out.

That last one comes up more than it should. Three examples from published breakdowns:

LGD-4033 · "+1.21 kg lean mass"

Quoted everywhere, including by sellers. It could not be traced to a figure reported in the trial it is attributed to, so the breakdown reports what the paper does say and notes the discrepancy.

RAD-140 · "90:1 anabolic-to-androgenic ratio"

A precise-sounding number with no locatable source in the published literature. Omitted, and flagged as omitted.

Caffeine · CYP1A2 genotype percentages

Widely circulated splits of the population into fast and slow metabolisers, at odds across sources. The breakdown grades the underlying claim Mixed and declines to pick a figure.

Who writes it

Breakdowns are published under The Archivist. The byline is deliberate: the argument here rests on the rule and the citations, both of which are published in full, and neither of which should be more or less persuasive depending on who assembled them. Every figure traces to a named paper with a PMID. Check the sources rather than the author.

Corrections are made in the open. If a mark is wrong, the correction and its reasoning are published alongside the original grade, not quietly swapped in.

Membership

The breakdowns are free, and stay free. All of them, permanently — the archive is not a teaser for something else.

What membership funds is the part that costs the most and gets published least: keeping the grades current. Watching for the trials that move a mark, re-running the rubric when they land, and telling you which published verdicts changed and why.

Free archive, paid maintenance

Reading everything costs nothing. Knowing the moment a grade moves is the part worth paying for.

Always free
  • Every breakdown, in full
  • The complete grading rubric and its revision history
  • All sources, with PMIDs

The Compound Codex — plain-language breakdowns of published research, graded under a public rubric. Informational only; nothing here is medical advice, and nothing here endorses the use of any compound discussed. Several compounds covered are prescription medicines or controlled substances.

© 2026 The Compound Codex · Every claim cited, none of them prescriptive.