[NoGravitas]

The competence profile of any LLM-based AI is extremely spiky - whether it does a particular task well or not is pretty independent of the (subjective) difficulty of the task. This is very different from our experience with humans.

[shivaniShimpi_]

slow was the safety net for sure but then there were errors too, there's a sweet equilibrium spot where ai + human oversight reaches that efficient + almost perfect situation. ofc with the right methods