The Wings Were Made of AI

In June 2024 a young man who had just left OpenAI put a document on the open web and said, in plain English, that he could see the future.

The site is still there: situational-awareness.ai. A hundred and sixty-five pages. Charts. Timelines. A voice that does not hedge. Advanced AI that matches or beats every human skill by about 2027, then an intelligence explosion. Trillion-dollar computing clusters. A government project. “You can see the future first in San Francisco,” Leopold Aschenbrenner wrote. “Let me tell you what we see.” Only a few hundred people — mostly in that city and in the labs — had “situational awareness,” he said, and peculiar forces of fate had put him among them. “I can see it,” he wrote. “I can see how AGI will be built.”

That was the wow. Not a blog post. A map. It looked like the future had already been priced and printed.

Within days, people who mattered inside the AI world began treating the map as real. Scott Aaronson, a computer scientist who had overlapped with Aschenbrenner at OpenAI, called it one of the most extraordinary documents he had ever read. The world still had not "priced in" what was about to hit, he wrote — and anyone in national security who read it carefully would panic. On a long podcast the week of launch, Aschenbrenner talked for hours in the same register: he could see the path, feel it. If advanced AI were priced into the market tomorrow, he said, you could maybe make a hundred times your money. Michael Dell shared the essay. Ivanka Trump shared it. Thousands of others did too. Business writers framed his exit from OpenAI as courage — a young researcher warning that the labs were wide open to theft — so the document read less like a hot take and more like a whistle blown from inside the machine.

A year later the fawning had hardened into a job description. Fortune wrote that tech executives, investors, and policymakers treated him as a kind of prophet of the AI age. An Anthropic researcher told the magazine he had extremely high conviction about how the world would evolve, and was putting his money where his mouth was — literally. By June 2026 the fund that took its name from the essay, Situational Awareness, held more than twenty billion dollars, a profile of his June trading reported. Returns for the year through May were around two hundred and seventy percent after fees; since launch, more than a thousand percent. A famous trading firm that almost never backs outside managers was listed among the investors. When the fund's filings moved, smaller traders tried to copy the trades. Finance blogs called him this cycle's celebrity stock picker. The AI kid was having his moment.

Look closely at what carried him that high, and the wings are easy to find. The essay itself is AI's favorite genre: confident forecasting, smooth copy, a website that looks like a finished product the day it appears. The culture around it was already half-machine — feeds that amplify certainty, tools that draft the praise, charts that make a guess look like a measurement. AI wrote the weather. AI helped sell the weather. AI made the weather look like a place you could visit.

Then, in a few days in late July 2026, the weather changed. The stocks that needed his timeline to be true took a beating. The ones he had bet against rose. Lenders who had let him borrow three or four dollars for every dollar of his own asked for their money back. The fund had no other map. Most of the public positions went out in a fire sale — roughly thirty-five billion dollars of value gone in a handful of days, the same week he got married. The Atlantic reported that regulators began looking into the collapse. He has not been accused of wrongdoing.

A few days later he put hundreds of millions into a chip company. Even after the fire sale, people said he was still ahead for the year. The essay is still online. The sentence is still there. He said he could see how it would be built.

For a while the clear picture held. Then he flew too close to the sun.