It was a short, but fun read. Congrats on the x64 improvement (=

I’ve got a few questions,

* Did you considered using a quality measure instead of the `effort` parameter to stop checking blocks? It seems to be too static, as different images might get good results with varying portions of the blocks reviewed (effort). If the metric turns to quality, then maybe the sorting can be dropped, or used as a second pass on fewer, low quality blocks.
 * Do you have stats on which proto-archetypes chains are descended the most and by what kind of blocks? Maybe there’s room for improvement adding more proto-chains