Jensen Huang: Give the Same Question to Multiple Large Models and Let Them Criticize Each Other

X's new head of product, Peter Sellis, is truly one of the top product managers in the world today. With one sentence, he laid bare the top strategic mindset of Silicon Valley: once you understand why SpaceX dares to proactively eliminate its most profitable Falcon 9, you will understand why the plans of the vast majority of companies are all slow-motion suicide. Peter has actually laid out an extremely brutal product logic for us: most mediocre companies die at the hands of competitors, while all top companies die because they dare not shoot themselves. When many teams do business planning, they are essentially doing self-deceiving linear extrapolation. They look at how much money they made this year, look at what resources they have on hand, and then cautiously patch things up on their existing turf, terrified of touching the cake of vested interests. What this kind of forward-inference thinking produces is often just a cheaper defective product from the old era. But the backward-inference method Sellis talks about directly reverses the timeline. It is absolutely not about guessing tomorrow from today; rather, it forces you to stand at the future endgame and look back: if the standard for human-computer interaction ten years from now is lightweight smart glasses, then no matter how good we make phone screens today, it is meaningless; every line of code now must serve that endgame. What is most ruthless is that he directly benchmarks this method against SpaceX's crazy, almost suicidal decision back then. Falcon 9 was clearly the most profitable and highest-success-rate ace rocket in the history of human commercial spaceflight, with cash flow so good it could coast for twenty years; yet Musk turned around and wanted to eliminate it completely, betting his entire fortune and life on Starship, which had exploded several times. This is like Cortés burning his retreat ships with his own hands after landing, leaving himself not a shred of room to merely survive. Sellis marveled that almost no large company in the world can reach SpaceX's level. Because human nature is deeply attached to the comfort zone, and people would rather count money on a ship doomed to sink than jump into icy water to build the next nuclear submarine. In fact, there is an extremely cold, counterintuitive truth here: the most dangerous state in the business world is never when you face desperate straits, but when you are holding a highly profitable golden rice bowl. Daring to smash with your own hands the cash cow that earns you the most money is often the only way to save your whole life when the future wave comes crashing in.

View original

X 新任产品负责人 Peter Sellis 不愧是当世最顶级的产品经理，他一句话就把硅谷最顶级的战略心法挑明了：看懂 SpaceX 为什么敢主动淘汰最赚钱的猎鹰 9 号，你就会明白为什么绝大多数公司的规划全是在慢性自杀。 Peter其实给我们掏出了一套极其残酷的产品逻辑：绝大多数平庸公司死于被对手干掉，而顶级公司全部死于不敢对自己开枪。 很多团队做业务规划，本质上是在做自欺欺人的线性推演。 看今年赚了多少钱，看手头有什么资源，然后小心翼翼地在现有的地盘上修修补补，生怕动了既得利益者的蛋糕。 这种顺推思维做出来的东西，往往就是更便宜的旧时代残次品。 但 Sellis 讲的倒推法，直接把时间轴倒了过来。 它绝不是从今天去猜明天，而是强行逼自己站在未来的终局往回看： 如果未来十年后人机交互的标配是轻量化智能眼镜，那我们今天就算把手机屏幕做烂也毫无意义，现在每一行代码都必须全部为那个终局服务。 最狠的是他把这套方法，直接对标到了 SpaceX 当年那个近乎自杀的疯狂决定上。 猎鹰 9 号明明是全人类商业航天史上最赚钱、发射成功率最高的王牌火箭，现金流好到可以躺着吃二十年； 但马斯克反手就要把它彻底淘汰，把全部身家性命押注在连炸了几次的 Starship 星舰上。 这就像当年科尔特斯率军登陆后亲手烧掉退路的战船一样，不给自己留一丝苟活的余地。 Sellis 感叹说，全世界几乎没有哪家大公司能做到 SpaceX 这种程度。 因为人性骨子里都在恋战舒适区，宁愿在一条注定沉没的大船上数钞票，也不愿意跳进冰水里去造下一艘核潜艇。 其实这里有一个极度冷酷的反直觉真相： 商业世界最危险的状态，从来不是你面临绝境，而是你手里正捧着一只利润丰厚的金饭碗。 敢亲手砸碎自己最赚钱的现金牛， 往往是在未来大浪打过来时，唯一能保住整条命的活路。