Just weeks ago, I had to pay out a $5000 bounty on my Cuckoo Cycle proof of work scheme [1], because I had wrongly assumed that hash sets were faster, given tha

Just weeks ago, I had to pay out a $5000 bounty on my Cuckoo Cycle proof of work scheme [1], because I had wrongly assumed that hash sets were faster, given that the hash set reduces to a simple bitmap. Where the article considers up to 10M elements, Cuckoo Cycle deals with about a billion elements, thus answering the question of what happens when cranking up the data size. It turns out that despite using 32x to 64x more memory than the bitmap, sorting is about 4x faster. Blog entry [2] explains how Cuckoo Cycle reduces to a counting problem. [1] https://github.com/tromp/cuckoo https://github.com/tromp/cuckoo [2] http://cryptorials.io/beyond-hashcash-proof-work-theres-mining-hashing http://cryptorials.io/beyond-hashcash-proof-work-theres-mini...