I would like to see extended floating point accuracy for outputs of smaller numbers that don't even 52 bit values. When working mostly less-than-zero values such as ratios and percentages, we are limited to only 6 digits of precision unless we arbitrarily multiply all of our values by 1 million or 1 billion in order to achieve a greater (useful) precision of the final product.
It would be nice to automatically have access to 12 digits of precision when working with smaller values, especially when they have to be multiplied with a previous product many hundreds or thousands of times. The errors do add up significantly, and it's rather painstaking to avoid.
Otherwise, if there's a library out there (in C++) that supports "big math" with huge numbers and infinite precision, I think many of us would really enjoy using that. $bigcalc() or something.
Well. At least I won lunch.
Good philosophy, see good in bad, I like!