Editorial Policy
The rules our content is held to — and what happens when it falls short.
01. No unverified facts presented as verified
We do not publish claims as fact unless they come from the cited source. Most first-draft summaries are pipeline-assisted (see methodology); until editors have reviewed them we keep claims anchored to source material and label analysis as analysis. Numbers and dates must be traceable to the source or to our own pipeline runs.
02. Attribution requirements
Every news item and project page keeps a link to its original source (sourceUrl,
repository URL). Metadata derived from open-source repositories is attributed to the upstream project and authors, per the
licenses in use (Apache-2.0, MIT and similar) — details on
Source Attribution. We do not republish
copyrighted full-text; feed-derived coverage is re-derived into our own summaries.
03. Deindexing thin and duplicate content
Detail pages below minimum content quality ship noindex, follow and are excluded from
our sitemap. Duplicates (same underlying story/project re-discovered) are collapsed during pipeline deduplication; if
duplicates or thin pages slip through, the correction process deindexes them and, where useful, 301s them to the canonical
page. Search result pages are permanently noindex.
04. Reading-time rule
Reading times are computed from the real word count of the body content (roughly 200–220 words per minute for news,
matching the published readingTimeMinutes values). We do not round reading times
upward to look more substantial, and we do not omit it for long pieces.
05. Corrections
Errors are fixed transparently — see our corrections process.