I am going to publish something controversial later...
It is not meant to be primarily programmed in, but to help communicating without other people first having to learn to parse sexps. I think this barrier is much bigger than Lispers want to believe. At least this is my experience during my PhD in an ML environment.
Feedback already welcome.
Try to publish ML papers with Lisp code in it, and you will have a hard time. Same in presentations, you cannot show any code, and if you do people think you are a little bit nuts. I had this many times.