Every claim on this site is checked against the primary research before it is published, and every check leaves a trace you can follow. This page explains exactly how that is done and what the grades on the Ledger mean.
How we work
- Sources, not vibes. Every factual claim links to research or a primary source. No “studies show” without the study.
- Criticism included. Our Book Notes tell you what held up and what was refuted after publication — even for books we love.
- Field-tested. Methods get tried in reality — including ultramarathons and multi-year habit experiments — before we recommend them.
- Honest limits. Every guide includes when it will not work and when the right answer is a professional, not an article.
- Commercial hygiene. At most one clearly disclosed commercial element per article. Trust is the product; we do not spend it on banners.
What the grades mean
Every claim in the Ledger carries one of four grades and the date it was last checked.
- C — Confirmed. Replicated in independent samples, and the effect survives when the weakest studies are removed.
- P — Partial. Real, but smaller, narrower or more conditional than the popular version claims.
- H — Untested / hypothesis. Plausible and widely repeated, but the evidence that would settle it does not exist yet.
- R — Refuted. The original finding failed to replicate, or the popular claim misreports what the study measured.
A claim with no grade is queued: it is on the list, it has not been checked yet, and it is marked as such rather than quietly presented as settled.
How a claim gets checked
- The claim is written down as people actually state it — not softened, not strawmanned.
- The citation chain is traced back to the first source. Most popular claims lose their source somewhere in the retelling; that is usually where the error is.
- The study is read for what it measured — sample size, design, effect size — not for its abstract.
- Every identifier is resolved. DOIs are checked against Crossref and PMIDs against PubMed, with a deliberately fake identifier in every run: without it, “not found” is indistinguishable from “the request failed”.
- The current search results are counted. How many top-ranking pages still present the claim as settled fact, with no caveat.
- The grade is assigned and dated. If new evidence arrives, the grade changes and the date moves.
When we are wrong
Corrections are made in the article itself and logged, not quietly deleted. Spotted an error? Tell us.