I think cal.com is assuming LLMs are only good at hacking with the source code of the target, whether that's true I don't know