Comment by kolinko

Output of LLM can be infallible* even if LLMs themselves make mistakes. Ditto with humans.

As much as anything can be infallible.