This is such a dangerous take these days. Something can appear to "work" in the main case but be grossly or subtly broken for common edge cases. Malicious contributors (see recent xz exploit) are skilled at making code that works but also has subtle security issues. You can't just say "it doesn't really matter if you understand the code" in one breath and then "of course you have to watch for security and safety issues" in the next. How can you watch for those issues if you don't understand the code? And suggesting LLMs makes the deep mistake of thinking that an LLM understands anything at all.
Specifically you said "it doesn’t really matter if you understand the code"...the literal code under review. Not the underlying mechanism of an abstraction you're using, which is useful, sure, even if you don't understand how it works under the hood. If you're reviewing the code that implements an abstraction, you better damn well understand how it works under the hood.
