Show HN: Adaptive-K – Cut MoE inference costs 30-50% with entropy-guided routinggithub.com/Gabrobals1 pointGabrielebalsamo5 months ago