Give credit at every step.
Every discovery item displays its source and links directly to the original. The news index stores permitted headlines and basic provenance, without article bodies or publisher images. Signal briefs use a separate cache of approved primary-record evidence, such as a CC0 research abstract or an agency-supplied summary. Each digest identifies what was used and the basis for that use.
An RSS feed, a public URL, or permission to crawl does not automatically provide republication rights. Publisher requests and source-specific restrictions take precedence over the convenience of ingestion. AI rewriting does not create permission to reuse someone else’s work.
The source directory records the current discovery reuse basis and whether ingestion is enabled. Pending sources contribute no stories. A source approved for headline discovery is not automatically approved for a substantive brief. We make no claim of publisher endorsement.
From the original to your index.
- A fixed source registry records permitted endpoints, coverage, and reuse basis. Only enabled, approved sources can be ingested.
- Production source checks are scheduled every 15 minutes, independently of visits. Reader requests can also refresh an index older than its 15-minute cache window. Requests have timeouts and size limits. Readers cannot supply arbitrary URLs.
- We validate source URLs, reject unsafe links and implausible dates, and keep the original publication date separate from the fetch date. Original publisher links stay intact when their terms require it.
- Duplicate canonical URLs are collapsed. Related coverage is grouped only when headlines substantially overlap or identify the same CVE within seven days. Every reference remains attributed; a cluster is not independent corroboration. Several category feeds from one publisher count as one organization.
- The browser checks the index every minute while visible. New items wait behind an “Update view” control to preserve your reading position.
The production index archives permitted metadata in Postgres. A scheduled Neon function runs the source checks, with a separate daily Vercel refresh as a backup. Schedule timing and source availability are best effort; a scheduled check does not guarantee a new story or an immediate browser update. The source directory shows the latest recorded checks and successful connections.
If a deployment has no configured database, it uses a temporary memory cache and the source directory identifies that limit. Feed failures keep the last available items and show source status. A successful fetch does not make an old publication new. “Latest” means latest within the available approved-source index, not complete coverage of the web.
A shorter read, with the evidence in view.
The Signal mark beside a news reference opens its brief. Signal briefs is a separate collection of available primary-source digests; OfRoot Insights contains OfRoot-authored guides and analysis. The original-publisher link remains available alongside the Signal action.
Briefs are assembled automatically from an exact, approved record: a matching DataCite DOI and research URL, Federal Register document number and URL, CISA CVE catalog entry, or FDA feed item with a matching release URL and title. We do not scrape commercial articles, review full papers, or infer a substantive summary from a headline. When the record is unsupported, missing, or temporarily unavailable, the page explains that coverage gap.
The reading view labels its extract as abridged primary-source text. It retains complete source sentences within a short reading budget, with structured source details where available. This is not AI-written prose, original reporting, independent verification, or comprehensive coverage of an event. OfRoot’s separately labelled context offers conditional considerations; it does not turn a source’s claims into established facts.
Discovery attribution and evidence have distinct citations. The discovery reference credits the publisher’s headline; the primary-record reference supports the excerpt and structured details. These roles may point to the same agency publication, so two numbered references do not necessarily mean two independent sources. Each brief shows its evidence scope, reuse basis, source-check time, and preparation time.
Available briefs normally use a one-hour cache window. A later request checks their evidence again after expiry. Unsupported or unsuccessful checks may be cached for 15 minutes. Brief preparation is separate from the news feed’s 15-minute refresh. A cached brief cannot make a removed reference public again: each detail view checks the current public index or the metadata archive and applies the current source and suppression rules. A headline leaving the latest-news window can still have a brief, while paused sources and removals take precedence over archived access.
For the applicable source policies, see arXiv’s metadata terms, DataCite’s CC0 policy, Federal Register’s site policy, the CISA KEV license, and FDA’s website policies. Full-text papers, third-party materials, and linked vendor pages have their own rights and are excluded from this digest process.
A useful selection, with the rules visible.
The main stream uses the source publication date, newest first. Items without a reliable date appear after dated items. Ticket sales, overt event promotions, and NIST event listings are excluded. Filters use the registered source beat, with conservative keyword routing. A category describes discovery coverage, not a fact-check.
The featured selection considers dated items from the last fourteen days. Recency, explicit technology relevance, and primary-record type guide the initial order; topic and publisher variety then balance up to four selections. One publisher receives at most two slots, even when it supplies several feeds. Each highlight states its source-based selection reason. These automated rules do not assert editorial verification, popularity, or comprehensive importance. There are no paid placements.
Saved references live in this browser on this device. They appear while the reference remains in the current index; this is a reading aid rather than a permanent personal archive.
Different sources, different kinds of evidence.
- Research: a research record or preprint. It may not be peer reviewed, independently verified, or ready for deployment.
- Release: an agency or product announcement. Attribute claims to the publishing organization.
- Policy update: an agency document, notice, proposed rule, or final rule. Read the original to understand its status, jurisdiction, and effective date. A proposal is not an enacted requirement.
- Security advisory: a source-issued alert or catalog entry. CISA catalog inclusion is evidence of known exploitation, not proof that a particular organization is compromised.
- Reporting: a publisher report, available only when its source has been approved.
Healthcare combines FDA releases, CMS health-technology announcements, and TechCrunch’s biotech and health reporting. Integrations includes interoperability research, NIST information-technology announcements, and relevant enterprise reporting. Coverage depends on the items supplied by these feeds; it is not comprehensive startup or product-launch coverage. New NIST and CMS permissions cover attributed metadata on the website and in email separately; article-body briefs require their own evidence adapter and reuse review.
Keep the record accountable.
For a correction, rights concern, removal request, or source suggestion, email communications@ofroot.technology with the source URL and relevant details. This opens your email application; it does not automatically send a message.
Operators can pause a source or suppress specific article IDs in the deployment configuration. Suppression is applied when the index is served as well as when it is refreshed, so stored metadata or a cached brief does not override a removal. Source terms should be reviewed again before broadening brief coverage or adding full article text, images, advertisements, or model-written summaries.