notes.txt

Q: What problem does foundations solve? How will you know if you have 
   solved it?
A: No foundation will ever be complete. What we need is some base
   to help us now AND a process for growing and expandign this
   to what helps us.

-> No "IR" "DSL" etc will ever be perfect. What we need is an easy
   way to iterate together to build what we need as our needs
   evolve, not a static base to constrain us.

   So this is why we need meta-mathematical (metadsl) formalism
   to give structure and space to freedom in development of
   these different bases.


Q: How will we know if this framework is more fundamental? Seems
   like old frameworks were all sort of just around what people's
   culture was at the time and their personality?
A: It's a good sign if we can use our current framework to understand
   the old ones.

-> It's key to be able to represent libraries like NumPy and Pandas
   in our meta language, but also more mathematical languages like MoA
   or loop nesting languages. The key is to be able to meet them where they
   are at, but represent them in such a way as to build in a common framework.

    So I see this work as some sort of higher dimensional space, that has to both
    reach into suporting high level frontends (NP), low level backends (LLVM),
    and also intermediate level abstract frameworks. So we use it to reach
    from concrete world, up to the intelligable world, and back down again,
    so build the structure to allow us to draw these bridges. 

"One of the reasons we have difficulty with physics is that our mathematical
continiuum is not sufficiently detailed and exact and complex to reflect what we have
in the material world."

"In constructive mathematics its traditional the normal concepts amplify into
multiple constructs" (i.e. probability and distance are both Reals but they 
have a different flavor to them, one is in some way more tangeable than the other)

"The hope is that in the next 50 years, coordinate will be an element of one
continuuum and velocity and element of another and they will converge when we assume
some core axioms and when they are not we are able to see them differently"