But, more and more, it's usually not what you want.

That may be true in your world, but it certainly is not in mine.

And unilaterally declaring that the default math should be a slow, cumbersome, and very rarely actually needed, infinite precision implementation based upon your own personal requirements and other equally specious arguments, is...well, specious.

Further, it is far better that new programmers learn about the limits of machine precision early in their careers -- when what they are working on is likely to be non-critical and closely supervised -- than that they go through that early supervised period ignorant of those limitations and end up struggling to produce sufficiently performant code when it matters later.