flickr and twitter
support near-identical
onebit 'nanoblogs' of faves
subscribeable via rss
hackable via api
hacking sequence:
download own faves' ids (count them)
save in a database
sort by author
rank by authorcount
for each author
download itemcount (save in db)
re-rank by fave frequency
detect densest faved authors
not already subscribed-to
download ids of faves of densest author
(save in db)
compare to own faves
calculate degree of taste-overlap
(twitter is different from flickr
because older tweets are rarely faved
so overlap won't extend back
past the more recent of 2 join-dates)
gradually extend database of others' faves
try to detect
still-unseen items
frequently faved by similar favers
.
