Humans also make mistakes, but unlike LLMs, they are capable of learning from their mistake and will not repeat it once they have learned. That, not the capacity to make mistakes, is why you should not allow LLMs to do things.

Developers repeat the same mistakes all the time. Otherwise off by one wouldn’t be a thing.