Short answer

The pipeline is designed around one story identity. A story can have a short news page and a linked expanded article, but index pages and topic pages remain navigation surfaces. They do not become additional versions of the same news item.

Publishing stages

  1. Collect public AI signals from source feeds, posts, repositories, and channel inputs.
  2. Map a story identity, domain, vendor, source trail, media, and Telegram message relationship.
  3. Prepare a short brief and an expanded longform payload.
  4. Run quality gates for prompt residue, repeated context, source links, translation, media, and SEO.
  5. Build static news and article pages, update discovery files, and hand verified URLs back to Telegram.

Why this matters

Search engines and LLM agents need stable public URLs. Readers need a clear choice between quick context and a deeper explanation. The one-brief-one-longform model keeps both needs aligned without producing duplicate story bodies.