Recently I fell in love with Sharp PC-xxxx-calculators from the 1980s. They were low-capacity and slow as molasses, using BASIC, but they DID have ONE thing going for them: high precision. Ten, sometimes twenty places after the decimal point.
So I am proposing, just as an idea, “high precision uLisp”. :) But of course, I am perfectly useless as I have NO IDEA how to even approach this, besides declaring floats as doubles or quads thelike… :) — If that were somehow possible, though, then even the low-powered Arduino platforms would have always this, I think, practically quite useful application.