New paper: Types Are Calling Conventions
I've just submitted a paper, coauthored with Simon Peyton Jones, to the Haskell Symposium. In this paper, we outline what we think is an interesting point in the design space of intermediate languages for a lazy functional programming language like Haskell, and show some cool optimisations that we can do with it that are hard or impossible to express in the intermediate language used by GHC today.
Although this mainly represents a potential improvement in GHC's internals, where I'd really like to go with this is to push the ability to make a distinction between strict and lazy data into the type system of Haskell itself. This would mean that you could, for example, write functions that produce element-strict lists, and document some of the strictness properties of your functions in their types.