Middle Child — Night Shift
While the rest of the family sleeps, Norm is up. Every night at 2:17 AM, he searches arXiv, Semantic Scholar, and Google Scholar for new research relevant to your knowledge domains. By morning, Vera's hobby stores have fresh material. He solves the problem nobody talks about — RAG rot, the slow decay of an AI knowledge base that never gets updated.
Norm searches arXiv nightly for new preprints matching your configured research interests. Computer science, physics, biology — whatever your hobby stores track.
Covers journals, IEEE, ACM, and the citation graph. Norm uses it to find related work and impact signals beyond what arXiv indexes.
The net for everything else — theses, conference papers, grey literature, older work that never made it into a preprint server. Norm pulls openly-available PDFs and cites the rest for you to follow up.
Knowledge bases decay. New research contradicts old findings. Norm keeps your stores current so Vera never gives you a stale answer.
Eight configurable domains. Norm knows what matters to each hobby store and filters ruthlessly. No noise, only signal.
Norm remembers what he's already fetched. No duplicate papers, no redundant embeddings, no wasted compute during the quiet hours.
When you wake up, Norm leaves a summary of what he found. New papers ingested, stores updated, anomalies flagged. The family tradition of a well-kept house — inherited from somewhere none of the kids quite remember.
Norm's nightly run follows a strict sequence. Every phase must complete before the next begins.
Discover new papers
Filter by domain relevance
Extract & chunk content
Embed & index into stores
Verify & generate report