I think the main difference, and the angle that makes your approach much more feasible, is that what users of your solution build fully lives „outside“ the core application that they build against, right?

While my approach is that users would modify the core application itself (even if initially in a sandboxed copy), which would be a lot more powerful, but also a lot more complex, with many more hard questions that would need to be answered, like how to deal with data and security and so on.