Google Reader does something similar to this when you subscribe to a site that doesn't have an RSS feed, although obviously it's generic. The general approach would be to periodically fetch the target page, parse it, regenerate the RSS feed for the [n] most recent items. It's not too hard, RSS...