I agree with this, but you can still make assumptions. Again: your thinking works in the environment where the assumptions are valid and then you only argue that they are reasonable, but of course they are not true.
It's better than random blabbing, which might be true, but you cannot evaluate the situation in which it's true.
I wrote a book about thinking in uncertainty, hopefully I'll translate to English too and you can cover also unknown unknowns, that's not such a big problem.
Another good example is Wolfram's physics. He thinks about all the possible universes with all possible inference rules and he still can make useful predictions about reality, even though it's super general.