So, I really do think the Input Hypothesis is correct when it comes to language acquisition,
BUT,
since learning more foundational linguistics stuff I'm convinced that it can give you a bit of a speed multiplier with your learning if you know how to apply it alongside the input-based methods. You still can't actually move forward without input, but grammar and general linguistics knowledge will help you move forward if you apply them in a way that's compatible with the Input Hypothesis. For example, you could target specific features you want to acquire and make sure to get more input for them, and figure out related structures that may even be necessary to acquire first.
It's like, vocabulary input is the bulk of the structure, grammar is the scaffolding/support, and general linguistics knowledge is the degree in architecture. You can't actually do anything with just a degree in architecture besides imagine making something, and if all you have is the support or scaffolding you don't actually have a building. But for a lot of types of buildings you don't need either of those and can just brute force your way through piling the bulk material in different ways, and even make your own theory of structure-making through your experience.