[amelius]

The problem with LLMs as I see it is that they make logic errors all the time, and somehow LLMs themselves are not smart enough to reason about their own reasoning.