I have been turning one piece of this over for days. Everyone is watching the loud story. Secretary Rubio saying of Ukraine that this is not our war, it is on a different continent. Trump, asked what he would do if Xi Jinping moved on Taiwan the way Washington moved on Venezuela, answering that it was Xi's decision and he knows I'll be unhappy. For Taiwan that landed like a thunderbolt. But the loud story is not the important one, and I think we are all going to look back and realize we were reading the wrong headline.

Start with what the retreat actually is, because it isn't a collapse. It's a consolidation. Canada as the 51st state, Greenland, the Panama Canal, the Gulf of Mexico renamed the Gulf of America — this is a country that knows it overextended and is pulling back into its own hemisphere, and a hemisphere with three oceans and enormous resources is not a consolation prize. Put Canada, Greenland and Venezuelan oil on the balance sheet and even a $38 trillion national debt starts to look manageable. Europe has understood what this means. It will build its own army, calculate more selfishly capital by capital, and in the end live with Russia and a partitioned Ukraine.

And nobody moves militarily for about a decade, because of seventeen elements in the periodic table. Deng Xiaoping said the quiet part out loud back in 1992: the Middle East has oil, China has rare earths. The light ones are everywhere and a country can build processing in five to seven years. The heavy ones — dysprosium, terbium, the magnets that go into F-35s and bunker busters and missiles — are essentially only in China and the sliver of Myanmar beside it. China never wanted to play that card, because a card starts losing value the moment you use it. It got pushed hard enough to play it anyway, and what it bought was stability.

Here is the part I think will matter in thirty years. America's AI has to charge you. The Magnificent Seven is about a third of the entire US stock market, and a valuation like that is a promise of revenue, so the models have to be proprietary — that isn't a strategy, it's an obligation to shareholders. China went the other way and made its large language models open. DeepSeek, then Alibaba's Qwen, given away. Remember Sam Altman telling Indian founders not to waste their time, that there was no way to catch up? That was true only for as long as intelligence was a toll road. Now Singapore, Brunei, Laos and India are all in the game without financing a frontier lab, and they can build the proprietary thing that actually earns — the mining application, the model that predicts which building is about to fail — on top of a floor that cost them nothing.

Watch what that does when it compounds. A toy factory in June already fitting dolls that recognize your face. Exoskeletons getting lighter and cheaper and smarter every quarter, which for a lot of us is the difference between walking at eighty and not. Humanoid robots that went from stumbling to landing perfect backflips onstage beside human dancers in about six months. The robots aren't the point. The point is that this is diffusing into every industry in the country at once, in a way it cannot where the floor is metered. So here is the question I'd leave my nieces with: thirty years from now, which one aged better — the empire that consolidated three oceans, or the country that gave away the thing everyone needed?