This seems to be another attempt at a grand, unified packaging and IPC / linking method. There have been hundreds of these, and there is no attempt at all to discuss where all the others went wrong, and how this will improve upon them.
I would love instantly dig into any software component. I can do that with website: I can right-click on button and click "Inspect" and I'm very close to discover what code executes. Imagine the entire computing platform built this way. Few navigations and you are reading TCP/IP code. Click here and there and you are inspecting USB packet from your mouse.
Also live code update is so rare, yet so useful. Set breakpoint on that line, stop there. Add some code and it'll be executed immediately.
Java invented Javadocs, which is awesome. I can read almost 0 documentation on some simple library yet I can use it with autocomplete. JavaScript isn't quite there yet, but it could be.
There's a lot of improvements to be made in current software stack.
"One library to rule them all" isn't as compelling as the UNIX principle. Because when one of these systems is proven to be inefficient or buggy, it doesn't discredit the entire library.
I want some of these parts in some of my apps. But not all, always.
Self-describing components in .NET
Visual Basic 6's component method.
Windows has had various of these, including ODE and COM.
Obviously, any linux distribution, and plain old C linking, probably comes closest (although that tends not to be self-describing.
Lots of RPC systems I would say fall into this list, just google for "self-describing RPC" to find a bunch of versions.
Also, it's not quite correct to say that these failed. They did what they were supposed to do; they simply were not the magic bullet some proponents made them out to be.