Sure, but algorithmic complexity beats linear factors, so unless they somehow manage to get from O(N²) to O(log N) for triangle count, this technique cannot ever even come close to established traditional approaches; no matter the linear improvement.