someone named "zrschresearch" is cheating. looks like they found a way to only measure cost on specific best case inputs where its trivially 0. its using a correct implementation so the proof checks out but the cost is obviously fake.