I've been thinking about this and similar issues myself lately.
There have been two aspects that were raised in this discussion and I would like to address both. First, the website for reviewing papers, noting errors in papers etc. A few months back Scott Morrison, Jessee Johnson and myself were chatting and decided to open such a website. The result, a wiki, is wikademic.org and it has very few submission and even fewer users...I am the main user and I use it as a sketch pad for scribbling my notes about papers that I read. The current idea for wikademic.org is that it will be a place to discuss papers: the difficulties therein, the mistakes, the missing references, a summary that isn't just what the authors want it to be, the nice parts, the parts that are an obvious copy from a different paper...you get the point.
So far it has failed to attract people because it is small, and vice versa. I have recently discovered another, slightly larger website that has a similar goal (but it leans more towards the humanities), acawiki.org and was thinking about integrating the two websites. This has yet to happen, and I'm not sure it will. I think that some sort of institutional backing is very important so that the people who donate their time and effort to writing in the site feel that whatever they write will be available to them later if they want. Perhaps this can be solved technically by providing an easy way to email a review page or a set of pages to yourself or something like that.
If you are interested in wikademic.org please drop me a note.
The second thing that was mentioned in this discussion is the possibility of finding papers that are interesting to you. I was just thinking about how such a program would be a fantastic time saver. Imagine having a program/webiste know the people and papers that interest you and by looking at the arxiv and other various (free or locked) sources of titles/abstracts/authors/fulltext could alert you whenever a paper that you "should" read comes up. This could happen because the paper is new, or because your interests have shifted/expanded (the program should find this out by trivial actions on your part such as downloading a new paper to your computer, or adding a contact to your email program, or by starting to write a new paper yourself, and not by explicit actions such as you creating a filter or somesuch).
If such a program exists, I'd love to hear about it. And if someone wants to write it, I would applaude them, and possibly help too!
