Since the Prismatic team is reading, I'll post a couple requests/annoyances.<p>Context:<p>I'm on the obsessive side when it comes to news reading. I'm trying to accurately predict the future, which mostly works. I clear out my full Prismatic feed twice a day supplemented with first page HN twice daily and my google reader subscription covering ~300 low-volume blogs and 5 planets. Out of the 700-800 feed items I scan in a day, usually about 15 are interesting. Prismatic sources 2-3 of these that I don't get in other sources and usually has 10-11 of the 15 covered. I consider this to be excellent performance for a single source and recommend it to pretty much everybody.<p>Issues:<p>* I get a lot of (exact) duplicate articles. It seems like my feed tries to pull ~350 articles. If I cover them all, then refresh, I'll frequently but not always get ~200 unread articles, 90% of which I've just skimmed. I've also had this happen a couple hours later with the same 90% previously read behavior. On a more infrequent basis, I get the same article repeated for months. As an example, I've been getting 'Journey through the MVC Jungle' once every two weeks or so since it was published last summer.<p>* Topic clustering for medium volume topics. I know you're doing this for higher volume topics (though I'd prefer slightly more pruning) but I get a lot of repeat articles for lower volume topics. My annoyance at the moment is the rash of "no Firefox on iOS" articles. It's a stupid story (no actual change in Moz policy) but I've had like 40 articles on the topic all from different sources over the last 3 days or so. Don't know if there's a great solution but I'd like to be able to blacklist `Firefox AND iOS` for the next two weeks.<p>* Integration of low volume sources. Personal blogs by clever people are the absolute best source material and I do not want to miss any articles in these feeds. Looks like you're on top of this.<p>* Topic bucketing. I find it easier to partition the news by topic and then do my skim. For example, I want to do all my Clojure articles in one pass, then Mozilla articles in a second pass, then Web Design articles in a pass, etc. This seems to be an unusual pattern but it helps me filter out duplicates faster which is important due to volume. I do it on Google Reader by grouping feeds into folders and 'gu[foldername]'.<p>* Can't read on the subway (Android). I emailed back and forth with Jacob last October and said that I'd write an offline capable mobile version of the site but I still haven't gotten around to it. Happily, my framework is coming together and with the closing of Reader I actually have a deadline to finish this.<p>Regards.