Scouts look around
Small research scripts read approved news feeds, papers, and official pages. They collect clues, not passwords.
COLLECTImagine a tiny newsroom, a careful librarian, and a fast printing press working together. This page explains the journey in plain language—without showing any private keys or secret settings.
Every daily report takes the same careful route, like a package moving through labeled stations.
Small research scripts read approved news feeds, papers, and official pages. They collect clues, not passwords.
COLLECTThe report writer turns the clues into a short story, keeping links beside the claims so readers can check them.
EXPLAINValidation looks for missing dates, missing sources, broken placeholders, provider errors, and duplicate report days.
VERIFYA build script turns Markdown into fast HTML pages for the latest report, archive, search, and each dated report.
PUBLISHThese are the jobs—not secret ingredients—that make the site tick.
A scheduled workflow starts the research run around the Hong Kong morning target.
A configuration list tells the scouts which official, research, and discovery sources to watch.
A plain-text recipe gives the writer its tone, sections, evidence rules, and continuity clues.
Tests and report checks catch mistakes before a new page is allowed onto the public shelves.
One static build makes the homepage, report pages, archive cards, and searchable index from saved Markdown.
Your browser downloads HTML, CSS, and JavaScript. There is no database server waiting behind each click.
The architecture guide explains the shape of the machine, not the contents of its locked drawer.
Secret values
Private provider credentials live in the automation platform’s secret storage.
No surprise overwrite
If a date already has a report, the generator stops instead of erasing it.
Safe failure
If research or writing fails, yesterday’s published pages are left alone.
The simple four-step journey contains smaller jobs. Here is the full relay race.
The clock is converted into the report’s calendar before any file gets its name.
Recent items are welcome; stale items are filtered away.
The scouts learn which official, research, and discovery shelves to visit.
Different feed formats become one tidy shape: title, link, time, source, summary, and topics.
Five articles about one launch become one evidence bundle instead of five pretend events.
Recent reports help separate a true update from yesterday’s repeated headline.
The writer creates the briefing. Incomplete or invalid work is edited and checked again.
A protected Markdown file becomes HTML pages, a search record, and a deployable site artifact.
Each station repacks the same evidence for the next job.
Research, writing, repair, validation, Markdown saving, page building, tests, and packaging happen here.
Your browser receives ready-made files and handles small optional interactions locally.
A failure should leave the last good public site standing.
Other usable sources can continue. No response is treated as missing evidence, not permission to invent.
Retries stay within limits, then a conservative evidence-only edition can preserve the day.
The exact problems are sent back for editing. The edited draft must pass validation again.
No successful site artifact is published, so visitors keep seeing the previous good version.
The official announcement, a technical paper, and a discovery article enter through three doors. Clustering recognizes one event. The official page and paper become strong evidence; the discovery article becomes context. The writer explains what is demonstrated, labels what is only claimed, and avoids inventing a price. Validation checks the links and required sections. One careful story then appears in Markdown, HTML, the archive, and search.
Three doors in → one verified story out.Read a real briefing, explore the archive, or search the signal.