Consider a service that tracks new papers in a domain; then applies this and sends a weekly digest or podcast.

True. tried manually reviewing all papers on arxiv under cs.cl for a year and a half and really felt the need to automate this https://corca.substack.com/

Did you automate it?

Tried to with vector embeddings and stuff and worked out okay-ish but sadly failed to get a solid mvp. Some retro tldr: 1. Fomo of losing papers not looked over by a proper researcher 2. What is a good, novel paper - research undoubtedly has its trends and cold start problem, inevitably fueling papers with overused topics to be tagged noteworthy, negating the purpose of doing daily arxiv digest