THE EDITORIAL DESK

Keep the source in sight.

Original practical guides

The four guides are original Practical Intelligence articles. Their examples are explicitly illustrative, not measured customer results. Primary references appear alongside the claims they support. Articles have publication dates; they are not silently rewritten from feed headlines.

The research digest

We display titles, author names, feed announcement dates and links from arXiv AI and language feeds, plus Creative Commons Open Source. Feeds are checked hourly on the server. This is a research reading desk, not a general news wire or a tool endorsement.

arXiv titles are grouped with simple topic keywords: agents, workflow and automation; code, software and program synthesis; retrieval, search and knowledge; evaluation, benchmarks, hallucinations and security; language, documents, translation and prompts. A paper can match several groups. Creative Commons entries are grouped under Open source. Non-matching research titles are omitted. Keyword matches are not AI judgments or evidence of research quality.

Reuse permission

• arXiv’s metadata policy provides metadata under CC0. Each paper’s text may have a different licence. We display metadata, not abstracts or full papers.
• Creative Commons Open Source states a CC BY 4.0 site licence except where otherwise noted. Each displayed entry includes publisher, author attribution, original URL and a licence link. We omit article content and images and normalize metadata presentation.

The imported licences permit commercial reuse under their applicable terms. RSS availability alone is not treated as permission to reproduce an article.

Source status

Original feeds: arXiv AI, arXiv language and Creative Commons Open Source.

Refreshes & retention

Cross-posted arXiv links are merged. The shared collection retains up to 35 days of research announcements and a small Creative Commons archive, with an overall 2,000-item cap. This digest selects the relevant metadata from that collection. Failed refreshes retain previously fetched entries; source status shows the last successful and attempted checks. Data is marked stale after three hours without a successful source refresh.

Our RSS export contains up to 100 latest signals with attribution. The normalized JSON snapshot is also public. Browser reading lists can retain metadata beyond the feed window.

Commercial content

No paid placements are currently active. If introduced, ads and sponsorships will be labelled and separated from guides and research entries. Sponsorship will not imply research validation or a benchmark endorsement. See advertising enquiries.