Geometric algebra also affects another "in-process" goal.
I have BLAS and LAPACK in the Axiom sources (volume 10.5).
I've spent some time on the question of changing BLAS to use
John Gustafson's UNUM representation, which eliminates a lot
of code because various "standard errors" cannot occur.
I'm in the process of re-architecting Axiom, of course.
The primary research effort, as you know, is incorporating
proof technology.
But in the process of re-architecting there are more things
to consider. Two of them are "front and center" at the moment.
One concern is "Geometric Algebra". See