A daily digest of new papers on your topic
An agent that reads new arXiv papers in your field every morning and sends a short digest with links.
For your agent: agent.md
Parts
Steps
- Query the arXiv API (Atom, no key) for your categories and keywords, newest first.
- Summarise each new abstract with your model and group the papers by subject.
- Send the digest with links to the arXiv abstract pages, not copies of the PDFs.
Watch out
- At most one request every three seconds, on one connection.
- Metadata is CC0, but each paper keeps its own license; do not redistribute PDFs unless that license allows it.