I think metaprogramming is a bit of an antipattern in most use cases(The more power the programmer has, the less the compiler/linter can follow it, just because... there's more possibilities).
But for the stuff it that needs it, it seems to be unbeatable, and this seems to have type annotations, which fixes one of traditional lisp's ickiest things.
Plus, having access to Python's libraries should make it possible to really evaluate what LISP can do by as a language, rather than having the choice mostly made by ecosystem size for a lot of projects.
I prefer a much more predefined structure that fits in better with autocomplete, and focuses more on reuse than new ideas, but I could see some pretty amazing stuff being done in this, and it being a benefit to the Python community as well.