I am actually building something in this space right now. Basically Spotlight but searches in < 10ms, half a dozen ML models used for indexing and a UI optimis

I am actually building something in this space right now. Basically Spotlight but searches in < 10ms, half a dozen ML models used for indexing and a UI optimised for finding and opening your file as quickly as possible. All local. And so I added a ChatGPT screen just to see if it was a good fit and what I found was that with LLMs it's more exploratory i.e. lots more back/forth and the latency is so much higher. So not sure if the future is some hybrid like Dash or more like Spotlight/Glean where you have two separate apps.