I think the “find file” section could use some clarification. Unless I missed something, as implemented it’s impossible to list paths within the filesystem (unl

I think the “find file” section could use some clarification. Unless I missed something, as implemented it’s impossible to list paths within the filesystem (unless the cortex stores the path? It’s not clear from the docs). At a minimum I’m curios about the costs associated with maintaining the cortex - there’s nothing about how the cost of metadata updates which is where the slowdown as disk fills up normally is since you have to do a sorted insertion and/or deletion or otherwise add indirection markers after a binary search. > The file's metadata in memory is updated to the new version. Which means this doesn’t work well for lots of (presumably small) files because of the bookkeeping overhead of needing to have all the metadata materialized in RAM? Have you tested how your filesystem scales as the number of files increases and how the RAM usage scales? Anyway, super interesting ideas. Congrats on achieving something difficult!