It's the same algorithm, so no change, except it's being done in hivemind instead of hived (which is so far a bit more challenging to code in a performant way).