I think Jev and LLM+BAML can be complementary. My "on-deck" experiment with Jev is to pair eval. Sort of council. LLM+BAML to give structured output akin to Jev's pritives and see if they agree or disagree. If they disagree, then flag for human review.