The problem there is that they try to parse voice as natural language -- if they instead made a simple programming language with a consistent format.. you'd be able to process the voice locally (speech-to-text has been mostly effective since like the 95s) and actually make it memorizable and semi-explorable.

And if you added a semicolon-keyword like "over" for radio, you'd avoid the issue where it waits a minute to see if you finished, making usage agonizingly slow.

But we somehow still haven't reached the point where people/corps finally realize that we simply can't parse natural language effectively