The World Newsroom

What is included

The release catalogue contains 546 direct news RSS/Atom feeds, across 61 feed languages and 125 catalogued country/edition contexts. It also contains 472 curated public social/community feeds, expanded from102:263 Bluesky,179 YouTube,24 Mastodon,3 Lemmy,2 discussion-community and1 PeerTube feed. The normalized social catalogue groups395 publisher/account families; multiple platforms can belong to one publisher. These are endpoints, not counts of independent reporters, subscribers or complete platform coverage. The live Source atlas shows current availability for each configured endpoint.

Some publishers have multiple language or regional editions. Publisher-family grouping combines reviewed equivalents such as BBC editions, DW editions, UN News, and France Médias Monde services. Family grouping is a catalogue aid, not a comprehensive legal ownership registry or an assessment of editorial independence.

Country denotes publisher or edition context recorded in the catalogue, not article geolocation, a reporter's location, or comprehensive coverage of that country. Multi-country/global sources can have no country assigned. Feed language is a catalogue/declaration label; individual social posts can be in another language. Mixed-language domain-account feeds are labeled accordingly. Headlines and excerpts retain their original language; translation is not claimed.

Discovery and admission

Candidate feeds were researched from publisher/account pages, documented syndication formats and the Plenary community RSS directory, followed by bounded first-party endpoint checks from the production host. Domain-handle Bluesky discovery used domains already present in the reviewed publisher catalogue, not personal-account discovery. Collection stopped when a rate limit was encountered; remaining candidates were marked unprobed.

New admitted feeds returned parseable source-linked items with valid source dates and a publication within the preceding 90 days at review. A successful probe establishes transport/content availability at that time, not factual credibility, current publication frequency or future access rights. Blocked, missing, empty, old and unprobed candidates stay outside the configured ingestion count. Their dated research outcomes appear under Coverage opportunities; missing domain-handle RSS does not imply that an outlet has no account under another handle.

The October4 expansion reviewed public organizational profiles through Bluesky’s public API and social links on467 publisher homepages. Third-party mirrors/bridges, invalid handles and insufficiently scoped candidates were excluded. New Bluesky API feeds are bound to a reviewed account DID and retain only that account’s original non-reply posts; reposts and foreign authors are excluded. Publisher-linked YouTube and Mastodon channels retain the publisher-page identity reference. Domain-linked, platform-verified and publisher-linked identity bases are shown separately; they do not certify the truth of a post.

First-party feed/article hosts are allowlisted. Redirects remain within declared feed/publisher domains, private-address targets are rejected, DNS results are pinned for the connection, and requests have response-size and time budgets. No paywall, login or anti-bot challenge is bypassed. Complete X, LinkedIn, Reddit, private Telegram, Nostr relay or other platform-wide access is not represented by this catalogue; the existing advanced controls retain their separate source/access limits.

Publication dates and excerpts

• Publication timestamps come from RSS/Atom item fields or a Bluesky post’s record.createdAt, never its indexing clock. Atom update-only timestamps are labeled source-updated. Missing, invalid or future timestamps are not replaced by the retrieval clock.
• Undated records can be inspected explicitly through the date filter and are excluded from recent-publication windows. Retrieval and source dates remain separate.
• Bluesky and Mastodon RSS posts can legitimately omit a title. Their displayed heading is then a bounded excerpt of the original post and is labeled Post excerpt; it is not an AI-generated headline.
• Only source-supplied short excerpts and metadata are stored/displayed. Open the original link for the full report, current context, corrections and applicable publisher access. No full-text licence is implied.
• RSS/Atom markup is reduced to text, original scripts and bidirectional layout are supported, and declared character encodings are decoded. Seventeen locally hosted Noto font families provide lazy-loaded script coverage without sending reader font requests to Google. The font files retain their SIL Open Font License notices. Dates and language declarations can still contain publisher errors.

Reader, comparison and search

Topic navigation is inside every news and social reader, including fullscreen: All/Breaking, Economics, Health/Bio, Technology, World/Politics, Conflict/Defense, Energy/Commodities, Climate/Environment, Natural Disasters, Science/Research, Cybersecurity and Official/Regulatory. Search and Narratives/Related coverage remain available. Economics and Health subtopics are in the expanded filters. Topics combine stored source categories with bounded headline/excerpt keyword rules, primarily English with limited other-language cues; this is not equivalent multilingual semantic understanding.

Source types and political / editorial leanings

media-editorial-v1 keeps source type separate from perspective. Type filters include MSM/mainstream, alternative, wire/news agency, public-service broadcaster, state-linked, independent/investigative, specialist, research, official institution, advocacy, community and company/project. Types can overlap—for example a public-service broadcaster can also be mainstream.

Perspective filters cover conservative/right, progressive/left, libertarian, socialist, communist and other explicitly supported descriptions. Unknown is not centrist, and wire status is not a neutrality rating. Publisher self-descriptions, organizational profile descriptions and retained earlier Watchtower groupings have distinct basis labels. Source details show references, review dates and the separate type/leaning basis. A historical catalogue grouping is not presented as a newly researched independent bias rating.

Labels describe a publisher or organizational account, not every article, social post, reader, journalist or community member. National political terminology differs, multiple perspectives can coexist, and unassigned classifications remain visible. These fields do not change severity scores or establish credibility. Source type, leaning and platform filters work in both news and social, and are retained in saved views and exports.

Severity-first reading priority

headline-priority-v1 ranks news as Critical, High, Elevated, Unassessed, then Other coverage, with the newest publication first inside a tier and a stable record-ID tie-break. The same order is available for public social posts. Source details explains each label; exports include the assessment and method version. Related groups use the highest matched member priority, not a corroboration bonus.

• Critical cues include explicit immediate-warning/catastrophic-impact wording such as mandatory evacuation or a tsunami warning, and reports of large casualty tolls (20 or more, or wording such as dozens killed). These are source-text cues, not independently established counts.
• High cues include reported armed attacks, natural hazards, fatalities, major accidents, systemic banking/funding stress, public-health emergencies and major infrastructure disruption. Elevated cues include policy, elections and economic developments.
• Historical, hypothetical, forecast, negated, resolved and other contextual wording can reduce priority. Old attacks cited in current legal proceedings and detected cumulative or recurring death rates are not treated as new critical incidents. Limited sports/lifestyle/entertainment detection avoids some keyword false alarms. These rules can still miss context, irony, corrections or an event described indirectly.
• Missing/future publication times, or impact-cue reports older than seven days, are unassessed for current urgency. Unsupported language without a matched cue is unassessed, never assumed low impact. The normal three-day news window remains in effect until changed.
• There are limited phrase rules for English, Spanish, French, German, Portuguese, Russian, Ukrainian, Arabic, simplified/traditional Chinese, Japanese, Korean and Hindi. Coverage is uneven, not equivalent multilingual understanding. Other coverage means no high-impact cue matched; it does not establish safety or insignificance.

This is automated reading priority, not event verification, a credibility score, a prediction or an official emergency rating. Social posts remain unverified. Publisher identity, country and viewpoint do not increase severity. Latest and Global / publisher balance remain selectable.

Collection and freshness

The news worker runs on its existing background timer, normally every five minutes. Existing social feeds retain ten/fifteen-minute cadence; the370 newly added social feeds use30-minute polling. Host-aware concurrency, oldest-check-first scheduling, conditional HTTP validators, a210-second refresh budget and provider backoff reduce unnecessary traffic. A response of304 confirms unchanged cached content; it does not make its publication date newer.

A failing source can retain bounded last-good records for up to24 hours with original dates and an explicit retained-copy flag. Source failures remain visible. A broadly failed refresh cannot replace the accepted last-good cache. The news wire retains up to8,000 records and the social wire up to12,000; individual feeds,100-row reader pages and a96MiB snapshot publication budget are capped. This is a bounded working snapshot, not an exhaustive historical archive.

Snapshot freshness is based on collection age, with snapshots older than 30 minutes labeled stale. Source availability is shown as responding, cached within its poll interval, paused, unavailable, or not yet checked. Refresh in the reader checks the latest stored snapshot instead of forcing hundreds of upstream requests.

Saved views, privacy and replay

Reading lists and saved views are account-local browser storage, not encrypted server storage or cloud-synced accounts. Signed-out use keeps a session-only reading list. Account changes purge the active UI state, invalidate pending responses and close comparison/details dialogs before another owner's state is loaded. Storage failures retain the previous saved state.

The new public wire accepts only catalogue filters and pagination, never caller-supplied feed URLs, account handles, credentials or private source configuration. It uses a fixed reviewed registry. Existing custom-source monitoring remains in the classic/advanced controls and does not populate this public wire.

Live media requests and results are withheld during replay. Existing captured classic views remain available. Exports carry the selected view, original source metadata and snapshot time; they are not verification reports or automated predictions.

Account checks and recovery

Public headlines stay visible during account verification. Private reading lists, saved views, comparisons and prior search filters are withheld, and classic custom-source controls stay hidden until the check completes. A verified same-account return can restore the earlier view if you have not interacted with the public reader in the meantime. Account changes discard that pending state. Failed checks leave a usable public reader and an explicit account retry action. Failed media refreshes retain the accepted same-selection snapshot or show a retry panel; no all-hidden reader state is used.

Saved reading-list priority labels reflect the saved assessment time and are not silently updated to appear current.

Alternative and libertarian source expansion · October 2026

Added first-party feeds for Antiwar.com (news and original commentary), Activist Post, Freedoms Phoenix, Ron Paul Liberty Report, Ron Paul Institute, The Libertarian Institute, Future of Freedom Foundation articles, Foundation for Economic Education and The Corbett Report. Reason was already connected and remains labelled. Publisher self-descriptions and Watchtower editorial review are distinguished in Source details, with references and review dates. These labels describe source orientation, not factual accuracy or every contributor's beliefs. Activist Post and Corbett are labelled alternative/independent without a blanket political assignment.

Freedoms Phoenix includes curated links as well as contributed articles and carries a news-link aggregation label. Ron Paul Institute and Liberty Report are grouped as a linked editorial network for source-family context, not a legal ownership finding. Related links and platform cross-posts do not establish independent corroboration.

Social adds publisher-linked YouTube feeds, public podcast syndication and the publisher-linked Corbett Odysee RSS channel. Podcast episodes retain unique explicit source permalinks rather than collapsing repeated programme-page links. Only metadata and short excerpts are collected; audio/video is opened at its source, not recorded or re-hosted.

Rumble is available as explicitly labelled publisher-linked video references from the Ron Paul Institute RSS feed, plus publisher-linked Ron Paul and Corbett channel directory entries. Reference titles and dates belong to the referring article; video creator, native upload time and playback availability are not inferred. The production server received HTTP403 from a reviewed Rumble channel, so direct channel collection is not connected or represented as healthy. Channel-only references are excluded from connected-feed availability counts. Source atlas shows their identity and access qualification.

Libertarian and Alternative shortcuts apply a seven-day source window. Social also has Rumble, Odysee and Podcasts shortcuts; all ordinary topic/source/platform/date filters remain available. New source feeds join the existing bounded collector, source pauses, last-good retention and source-neutral reading-priority rules. This is not access to complete platform timelines. Source availability and publication activity are separate measurements.