As far as I understand it, the general scientific consensus is that linguistic relativity does not hold true. More anecdotally, I don't often find myself constructing sentences before having an intuitive (non-linguistic) understanding of what I am trying to express, and the process of putting ideas into words doesn't feel like it would be advantaged by high information density, but rather that the majority of filler-words are useful for smoothing out the process and providing information at a more controlled pace, which can be useful for things like adding emphasis to particular phrases and concepts.

From what I've seen, there is evidence for the weak linguistic relativity hypothesis. Is that wrong?

Not per se, but that would allow for any inefficiencies to be moved around, which is not contrary to what I'm arguing. It's quite a bold claim to say that people will think faster or more efficiently in general, but not as bold to say they will think faster or more efficiently about some things. You might be able to make a language more efficient for technical communication while making it less efficient for poetic communication, but I don't believe you can make a language more efficient at everything people need to communicate.