The current window for open-source AI viability is narrowing. Nathan Lambert argues that open models face a critical six-month deadline to prove they can keep pace with proprietary systems. If they fail to bridge the capability gap now, the ecosystem risks permanent obsolescence. Practitioners must prioritize efficiency over raw scale.