[danjl]

Also, the errors that an LLM makes are not errors that a typical human would make. This makes reviewing their code particularly challenging.

[kristiandupont]

I find the errors they make are often comically "human", like forgetting a premise.