How this is made
The weekly run
Once a week the pipeline requests only the feeds on the source list, reads each item’s headline, summary, publication time, and link, files it under its ISO week, removes duplicates, and publishes. Full article text is never republished — the site keeps the source’s own summary and links back.
Where the Chinese summaries come from
Chinese headlines and summaries are produced by a language model and go live without per-item human review. A model can misread or omit the point, so the original headline, the source name, and the link stay next to it — any item can be checked against the original.
How sources are chosen
Editorial control happens at the source level rather than per item: the list is hand-picked and verified, the pipeline may fetch only those addresses, and even redirects must stay within the same official domain. Changing what gets covered means changing the source list.
What this site does not do
No full-text republishing, no scoring, no ranking, no advertising or affiliate links, no reader tracking. A source that fails to fetch is simply absent from that issue; nothing is substituted for it.