Rebecca did a great job of collecting the comments to the 1.0 draft, and I think they are worthwhile to repost here.
Todd Huffman sent a separate PDF.
Perhaps the most important issue brought up (by Robert Freitas) IMHO is developing a good metric for our uncertainty ("to measure the degree to which we “don’t know what we’re doing”). We need a way of qualifying this on different levels, both in terms of model uncertainty and parameter uncertainty. Any ideas?