The viability of open source AI is currently facing its most serious test to date. Nathan Lambert argues that the window for open models to remain competitive is closing rapidly. This pressure forces a reckoning over whether open-weight research can keep pace with closed labs. Practitioners must now weigh the cost of independence against raw performance.