Side project: a second HiveD implementation in go — replays up to 30M blocks before failing. Curious whether a fresh AI pass with full repo context could outperform the original. Plans to maybe run a top-100 witness on it for the experiment.
This sounds very interesting! Code public by chance?
Code not public yet as it's still missing a lot of critical features (right now it's only replaying, can't do P2P) :)