Are there any theoretical limitation that prevents us from achieving this ideal unicorn programming language for parallel computation? Or is this a commonly held belief amongst the language designer?

@sirah I'm guessing that there are some inherent tradeoffs between these three traits. For instance, productivity boosts are generally accomplished by either abstracting away performance fine-tuning or specializing for a particular domain. I doubt that such a language is theoretically possible.