Sourcing policy
Which sources count, what each level is worth, and the failure each one is prone to.
The hierarchy
Primary literature. Peer-reviewed papers, located through ADS, cited with a DOI or bibcode where one exists. The strongest thing an article can rest on, and still not certainty: review checks that a result is competently argued, not that it is true.
Missions, agencies and instrument teams. NASA, ESA, JAXA, ESO, STScI, NOIRLab, the IAU, and the institutes that build and operate instruments. Authoritative on their own hardware, schedules and data. An agency announcement is accurate about what its scientists believe; it is not a reviewed result, and an article that treats it as one has promoted it a level.
Catalogues and archives. SIMBAD, VizieR, Gaia, the NASA Exoplanet Archive, the Minor Planet Centre, HEASARC, NED, the JPL Small-Body Database. The reference layer. A catalogue value is a best current estimate from a specific data release and it changes, so articles name the catalogue and, where it matters, the release.
Preprints. arXiv, before review. Genuinely valuable — most significant results appear there first — and clearly labelled PREPRINT. Preprints get withdrawn, and results more often weaken than strengthen between preprint and paper.
What does not count
Other outlets' coverage. Content farms and SEO pages. Social posts without a traceable source. AI-generated summaries, including summaries of good sources — a summary is where a qualifier goes to die.
The objection to each is the same: every rewrite is a chance for a hedge to be dropped by someone who did not know it was load-bearing.
How references are recorded
Each reference carries its title, its publisher or author list, the publication where there is one, its date, its URL, its DOI or bibcode where available, its source type, and the date it was actually consulted. That last field is never estimated and never backfilled — a URL without a date read is a claim about a page that may no longer say what it said.
The build enforces the mechanics. An article that cites a reference it does not declare, or declares one it never cites, does not build. Reference lists that drift out of step with their text are the normal decay mode of hand-maintained footnotes, and they are invisible to a reader who trusts them.
Uncertainty
A measurement is quoted with its uncertainty or it is not quoted. "1.1 times the radius of Earth" and "1.1 ± 0.4" describe different states of knowledge, and the second is frequently the whole story.
Where a source gives no uncertainty, the article says the source gives none.
Links to sources
External citations are ordinary links. They are not marked nofollow: they point at the evidence an article rests on, and refusing to pass a signal to NASA or to a journal would be a statement about those sources that this publication does not intend to make.