The state of the art has definitely improved. Zama are doing something called 'programmable bootstrapping' which can enable efficient (e.g., faster) Homomorphic

The state of the art has definitely improved. Zama are doing something called 'programmable bootstrapping' which can enable efficient (e.g., faster) Homomorphic Inference on deep neural networks -- https://whitepaper.zama.ai/whitepaper.pdf?_ga=2.218937936.879794490.1623664533-158940722.1622744631 https://whitepaper.zama.ai/whitepaper.pdf?_ga=2.218937936.87... Good question about database indexes; that could be an interesting use case study.