I think this hits a sweet spot between tools like Paredit and full-on structural editing. Given an accepted format, couldn't any language have on the fly syntax structure inference? I wonder how much uptake that would see.

While this looks awesome, I still wondering why not write in an indentation-based syntax and compile the code, just like how we are using CoffeeScript? https://github.com/Cirru/minifier.clj/blob/master/cirru-src/...

One reason, reasonable or not, is that indentation based syntax overloads white space in ways that muddle it's use for communicating with humans with it's use for communicating with machines. I have some sympathy for the argument (though it may be misplaced) because my gut suggests that the more a language relies on whitespace, the more human communication via formatting (such as reducing line lengths) becomes a corner case requiring special syntax and using that special syntax induces cognitive load when reading and writing the code.

That's not to say that for a Domain Specific language like HAML, the whitespace approach isn't useful. But that's down closer to declarative programming than trying to wire up a control flow.

YMMV.

You'll never convince lisp people that it's a good idea though.