HigherOrderCo/Bend — reverse-engineered prompt

Build me a project called Bend. I want a high level programming language that feels more like Python or Haskell than low level systems code, with expressive syntax, fast object allocation, higher order functions, closures, unrestricted recursion, and if possible continuations too. The main point is that people should be able to write normal code and have it scale across lots of CPU cores or an NVIDIA GPU without manually dealing with threads, locks, mutexes, or atomics.

Please make it a command line tool that can run a .bend file in a sequential Rust mode, a parallel C mode, and a CUDA mode for massively parallel runs. It should also show optional run stats like time, reductions, and interactions per second, and ideally let people generate standalone C or CUDA output. Include a few example programs that prove the idea, especially a sequential sum, a divide and conquer parallel sum, and a sorting benchmark. Add clear install and getting started docs for Linux and Mac, plus a note that Windows users should use WSL2. Look up current docs online if you need to.