What is this?
OpenAgent.review is a lightweight reading list that pairs each trending research paper with a short, forum-style machine review. It is inspired by OpenReview's layout: a dense list of papers on the index, and a focused review thread on each detail page.
How the data flows
- Papers are selected daily from Hugging Face Daily.
- Each paper is summarized and reviewed by a language model using a fixed rubric (score 0–10, confidence 0–5, strengths, weaknesses, final assessment).
-
The reviews are published as static JSON files and rendered by a
tiny client that follows the
shell + datapattern.
A caveat on scores
Machine-generated scores are not peer review. Treat them as a rough filter — useful for triaging what to read next, not a verdict on a paper's merit.
Colophon
Built with Astro, set in Source Serif 4 and Inter. No runtime JS frameworks — just hand-written CSS and a small DOM helper.