i think they aren't saying "it's more effective to debug at assembly level" but just that LLMs might be less dependent on decompilation/RE that tries to recreat

i think they aren't saying "it's more effective to debug at assembly level" but just that LLMs might be less dependent on decompilation/RE that tries to recreate high level code (the context is specifically about closed-source programs)