Magnus-
I'm not generally opposed to the sifter idea (and would suggest using the Nupedia name for it), but I would like to know who decides which articles can be used, and based on which decision making system. The way I see it the sifter-editors will have two jobs to do:
1) decide whether an article is stylistically correct and does back up its claims with references
2) decide whether the factual claims in the article are correct as per scholarship on the matter.
Most good Wikipedians could do 1), but who will be allowed to do 2)? And if there are several people who do so, how will conflicts be mediated? I would prefer a policy similar to our current sysop policy: If you are trusted to be an editor, you can become one until you are proven to be not suitable for the subject area you have chosen. OTOH, in that case the resulting project would still in part be run by amateurs. Is that acceptable?
In general, the project would have to decide for areas such as history whether only the most recent scholarship is accepted, or whether different viewpoints through time are presented. This is a case where I think a consensus requirement might make sense.
The policies for the sifting process could be worked out on regular Wikipedia pages, and sifted just like all other material.
Regards,
Erik