> That said, your broader point remains. I think we all appreciate the difference between "understanding" something and the ability to recite the call graph with your eyes closed after three beers.
> ...
> So I think the real question we ought to be asking, heavily motivated by the impact of LLM's on both the productivity function and the changing nature of offensive security, is how we can help more people like me safely contribute.
> Safely can mean a lot of things. Yes, there's the Rust conversation re: memory management. But more broadly, I think it's documentation, testing, processes, and - probably most controversially - a little more convergence of standards, styles, and even, *gasp*, code.

I always found debuggers to be fantastic code "explorers", including with the kernel. Nothing better to observe some call stacks "live" in their almost natural habitat. Unfortunately, debugging the kernel seems to almost always require some dark magic - whether it's some secret QEMU spell or something else, rarely practiced and even more rarely documented. Sad.