The open data behind this handbook

This handbook is generated from PubMed data, and the generation is reproducible and inspectable. That is the difference between it and a book.

How a page is built. For each topic, a query is run against PubMed's E-utilities with a publication-type filter that restricts the result to systematic reviews, meta-analyses and randomised controlled trials. Up to twelve records are fetched with their abstracts, parsed for publication type, journal, year and the conclusion paragraph, and stored as JSON.

How a quotation is verified. The sentence shown on a page under "What the studies found" is copied verbatim from the abstract of the record it is attributed to. Before the site is built, every one of those quotations is checked against the stored abstract — normalising whitespace, case and typography — and the build refuses to write a site containing a quotation that is not literally present in its abstract, a citation to a paper that is not in the evidence set, or a topic with no evidence. A failed build is a real failure, not a warning.

What that guarantees, and what it does not. It guarantees that a quotation is what the paper says. It does not guarantee that the paper is right, that the trial was well conducted, or that the effect generalises. A meta-analysis with very low certainty is quoted as such, in its own words, which is the most honest thing this format can do.

The chemistry is looked up, not remembered. Boiling points, molecular formulas and the physical facts that constrain the preparation chapters come from PubChem records pulled into the project. The distillation chapter's central claim — that a still at 78-100 C cannot carry minerals — is a statement about vapour pressure and boiling point, and it is the reason no essential oil is a mineral tonic.

No sponsors, no affiliate links, no advertising. The site loads no external resources, sets no cookies, and makes no requests off the page. There is nothing to monetise and nothing tracking the reader.

Licensing. Content under CC BY-SA 4.0; code under MIT; the underlying evidence is public-domain PubMed material and PubChem data. Anyone can fork it, re-run the harvest, change the topic set, and publish their own version — and if they disagree with a quotation, they can check it against the abstract in one click.

How to check a claim you doubt. Click the PubMed link on any evidence card. The abstract is free, and it is the same text the quotation was taken from. If the quotation is not in it, that is a bug and it is the kind that matters.