Of course you may ask! Exactly that, I want to see if I can monetize in one way or the other, so that's why I keep the compiler code to myself for now.

I'm planning to support generation of CUDA code and maybe also Triton for GPU inference. Would that be helpful for you?