An article explores the AI safety debate between advocates of pausing AI research to solve alignment problems beforehand versus those who believe alignment must be solved through iterative engineering during actual AI development, using an asteroid impact analogy to frame the stakes of competing approaches to existential risk.