In addition to Bug's rational, having the peltier directly
contact the carboy would limit the system to 1 carboy.
Right now, i'm trying to have 2 carboys filled, one in
primary ferm. and one post-primary but once in a while
Real Life(TM) gets in the way

Eventually i'd like to get
into longer fermenting brews like russian imp. stouts,
ciders, and meads. having that extra cooling space will
be invaluable. sooner or later, i'll pick up a garbage bag's
worth of packing peanuts to help eat the dead air space
when there's 1 carboy.
As ahuntertx and bub suggested, having the cold side of
the radiator directly in contact with water did cross my
mind a few times.
For a few reasons, i decided against it.
first, by forking the MoFC design, it would allow me to
easily fall back to its original design in case mine didn't
work. there's still a possibilty that i may use some type
of hybrid to help reach pitching temps or maintain lager temps.
Also, IMO the location of the thermometer probe would
eventually cause temps to deviate from your desired
range. lets say you manage to insulate the probe to
monitor the wort/beer. the resevoir of water, which
would most likely be less than the volume of wort/beer,
would get colder than the wort/beer.
once the thermometer shuts off, the resevoir would
continue to drop the wort/beer's temp. The flip side is
pretty much reversed.
On top of that, there's dealing with keeping the water
sanitary. Bleach/Iodpohor loses its ooomph after a while.
Though Star San probably could be managable.
On a side note, using those cans of spray foam to create
a customized insulating shell for each carboy crossed my
mind a few days ago. but this is way to long already