However, I think a piece of software that evolves is what most people intend when they start with a minimum viable product. What you don't see too often are benchmarks to "grade" users and automatically move them up to more advanced features. I'd have to think more about it to see how it could be implemented and how beneficial it would be but I think its interesting none the less. It would be similar to a game in which you "unlock" certain maps/levels/etc.
Thats a fine objective for _some_ software. Twitter clients should be like that, the majority of what email clients or web browsers can be like that.
It's possible (probable?) that fully featured project management software can't be like that. There's too much external system level understanding required to expose everything the user needs to know about how and why to use the system in optimal ways to be able to encode it all in beautiful UX/UI.
Try and work out, as an exercise, how you design a UX/UI for a git client that'll allow a cvs or subversion user to "pick it up with ease". Some things "just don't work that way", and I'm reasonably sure project management, at least at medium and large scale, require user training (whether formal or self-trained-via-google, or through many iterations of trial and error).
_The Humane Interface_ discusses the balance between delivering something with "optimal usability" from an objective and scientifically validated standpoint vs. delivering a product the target audience can quickly familiarize themselves with and transfer their existing domain knowledge to.
So, while the Canon Cat may be a revolution in word processors, the average user will do better with something similar to Microsoft Word because they know it and can apply what they already know rather than learn a whole new way of word processing.
If you're performing a complicated task, it's reasonable to ask that the software not make the task harder, but it's also reasonable to expect to have to learn how to use a new tool.
I suspect that project management (at least for some large projects) may be such a problem domain, where the tools are unavoidably complex.