- "The Future of Programming" talk, by Bret Victor, 2013 - this is making its rounds again, idolizing failed experiments.
You can't grep or diff binary trees. You can't get a Smalltalk IDE on your iPad. You can't write an operating system or a big application in Scratch or the Mindstorms IDE. And even a small program in these won't work in the next version.
But you can edit plain text with any text editor, whether that's ed, nano, Vim, emacs, BBEdit, Atom, Textastic, Editorial, Eclipse, AppCode, whatever. You can save it safely and easily in any source control system. You can run an awk or sed script over your entire codebase and it just works.
If your language (or non-linguistic programming environment in some cases) is only usable from a single IDE, you've cut yourself off from every other analysis and editing tool in the world, you're dependent on that one tool to do everything you want.
I have old Mac and iPhone NIB files which can't be read with any current version of Xcode/Interface Builder, the file format was only supported by one dev tool and it's changed, and the old tools don't run on modern OS X. These NIB files "work", in that they deserialize into objects, but there's no way to edit them. Where possible, I now do most UI work in code; this isn't great fun, I end up with a ton of builder functions to avoid repetitive code blocks, but it'll still compile and work in 10 years.
This is also why I don't use a WYSIWYG word processor, I use MultiMarkdown. I've lost documents to proprietary WPs, and of course there's no way to run tools over them (except, sort of, MS Word with BASIC).
None of this has stopped people from making new non-text environments, or weird experiments. Experiments can be useful even when they fail, telling us what doesn't work. But they don't catch on because the tools aren't as good as text, and won't work in the future when the experimenter gets bored and quits.