But then, the really good Java (or Scala, or whatever GC-based lang) programmers are the ones that actually know and actively take into consideration how the GC works memory-wise and speed-wise. The same kind of thing can be said about interpreted langs and speed. Of course, that kind of thing also happens in C++, good C++ devs know the order of any instruction speed and memory-wise, but it is much more evident in C++/Rust and it's actually documented in the standard libraries, and the lang itself tackles the problem.