Every thread starts from source feeds, is normalized into a story record, and is published with direct receipts. The goal is not maximum volume. The goal is a cleaner operating picture.
What’s automated
- Collecting new items from RSS/Atom feeds.
- Scraping selected HTML sources when they publish without a feed and robots.txt allows it.
- Normalizing, ranking, and clustering related coverage into thread-like rows.
- Generating a static site and JSON indexes for search, browse, and saved state.
- Generating structured summaries through a configured Claude CLI summarizer when enabled.
- Publishing from a local cron job instead of GitHub Actions.
How the product is structured
- Homepage: a briefing surface with clear editorial hierarchy.
- Browse: the searchable archive and reference layer.
- Story pages: dossier pages with summary, trust context, and receipts side by side.
- Simple and Deep: one product, two levels of depth, not two disconnected experiences.
What we avoid
- Copying entire articles. We link out and summarize.
- Hype-first coverage. We down-rank fluff, customer-story wrappers, newsletter roundups, and low-context community noise.
- Unlabeled speculation. If we infer, we should say so.
Source health
Last run: Apr 12, 2026 at 8:36 PM UTC • Sources OK: 80 / 82 • Quiet: 2 • Failed: 0
- Google Gemma Releases: None
- Cohere Blog: None