About AI Radar

AI Radar automatically aggregates official AI blogs, GitHub releases, trusted media and community discussion, clusters the sources that describe the same happening into one event, labels its verification status, and keeps tracking follow-ups on a timeline.

Events first, not articles

Several pages, posts, releases or news stories may describe the same thing. This site groups them into a single event with multiple sources, one timeline and an explicit verification status. We do not rewrite, ghostwrite or pile up articles.

What the statuses mean

• Confirmed Direct primary evidence exists (official blog, official repository, paper, announcement).
• Corroborated At least two mutually independent trusted outlets report it, without direct official confirmation yet.
• Limited sources Fewer than two independent trusted sources. Community reposts never upgrade the status.
• Rumor Only unverifiable sources.
• Disputed Trusted sources conflict (human editorial decision only).
• Debunked Sufficient evidence proves it wrong (human editorial decision only).

Source tiers

• Primary: official blogs, docs, official GitHub, papers, government documents, company announcements.
• Trusted media: professional tech media and research organizations.
• Community: Hacker News, Reddit, developer blogs with identifiable authorship.
• Unverified: anonymous accounts, screenshots, unverifiable claims.

Where the data comes from

The pipeline fetches configured RSS/Atom feeds hourly (official blogs, GitHub releases, trusted media, community), then applies deterministic deduplication and conservative clustering. Every event page preserves the original titles, links and timestamps — every fact is click-traceable, and no fact is ever filled in from an AI model's memory. Summaries are excerpts of source-provided text (marked EXTRACT), never generated content.

About translations

On the Chinese edition, English event titles are shown as labeled machine translations for convenience only; detail pages keep the original title, and every evidence entry keeps its untouched original title and link. Translation never alters facts and plays no part in verification.

Honest limitations

• Clustering is conservative: better to under-merge than to merge wrongly. One happening may temporarily appear as several entries.
• “Limited sources” does not mean false — only not independently verified yet.
• Times use the source-declared publication time; when unparseable, the discovery time is used and labeled as such.

Open data

All raw items, events and run records live as JSON/NDJSON in the repository, auditable through git history. Machine-readable outputs: RSS · health.json · search-index.json.