Sources and citations
Rule 2 says every published note carries at least one source, and every claim that is not common knowledge is cited where it is made. This note is how to do that well.
Reliability is a spectrum
No source is always reliable or always unreliable. What you are looking for is a reputation for fact-checking and accuracy, in the specific area the claim belongs to. A physics textbook is an excellent source on physics and a poor one on nutrition.
Prefer secondary sources
- Secondary — review articles, textbooks, meta-analyses. These weigh the evidence for you, and they are what most notes should rest on.
- Tertiary — encyclopedias, summaries. Fine for orientation, weak as the only support for a claim.
- Primary — the original study, the standard, the raw document. Use with care: a single primary study is evidence that something was once measured, not that it is established.
Avoid building a whole note out of primary sources. That is how a note drifts into rule 5 territory without noticing.
Generally not acceptable
- Self-published blogs and personal sites
- User-generated content — forums, social posts, comment sections, other wikis
- Predatory journals that publish for a fee without meaningful review
- The output of a language model, which routinely invents citations that look real
Exceptional claims need exceptional sources
The more surprising the claim, the stronger the source has to be. A claim that contradicts the mainstream view in its field needs more than one good source, and the note should say that the mainstream view exists.
Paywalled, offline, and non-English sources are fine
The test is that a determined reader can check the source, not that checking is free or convenient. A cited book with a page number is a better source than a free blog post summarizing it.
Cite where you claim
Attach the source at the point the claim is made, not in a pile at the end. A reader should never have to guess which of five links supports the number in paragraph three.
When two credible sources disagree, cite both and say so in the note. Disagreement is information.