I'd be curious to see how obfuscated code produced like this fares when analyzed with ghidra augmented with AI plugins.

Also, I'm surprised there seems to be no mention in the article of why standard decompilation techniques fail (I might have missed it).