I am currently using my regular RSS reader Feedbin. I eventually want to build something more context-specific. If you have any suggestions on how to do this better, please open an issue.
A preprint is included in the custom feed if its title, author list, or abstract contains any of the keywords. I realize that this is simplistic---if you need more logic please open an issue.
The latest feed is fetched from the ePrint RSS archive once every 2 hours and cached. The custom feeds rely exclusively on this cached version. This reduces the pressure on the upstream servers at the (imo reasonable) cost of the custom feeds being up to 2 hours stale.