One of the most insightful posts I’ve read recently. I especially enjoy the rationale behind the options you chose to reduce costs and going into detail on wher

One of the most insightful posts I’ve read recently. I especially enjoy the rationale behind the options you chose to reduce costs and going into detail on where you find the most savings. I know the post primarily focuses on neural search, but I’m wondering you tried integrating hybrid BM-25 + embeddings search and if this led to any improvements. Also, what reranking models did you find most useful and cost efficient?