I somewhat agree. I have a stronger postulate: current proof assistants are worse for writing proofs than machine language.

There is a reason not to panic: it can only get better, not worse...