I think the Wolfram Language got 80–90% of it right, and that's why I want to reimplement and use it in the first place. It bothers me that I have to use inferior solutions, such as those implemented in Python and Julia, just because the better solution is prohibitively expensive.

Of course, there are a few things I think we can do better than the Wolfram Language, but given all the available legacy code, there's really no point in making it slightly different and losing the ability to use that code. However, we might build an alternative syntax and frontend in the future that surfaces the capabilities of the Wolfram Language with a better UX.