Situational Awareness

Leopold Aschenbrenner wrote the most important thing I've read this year. The core argument: we are on track for AGI by 2027 and superintelligence by 2030, and almost nobody in a position of power is taking this seriously.

What hit me hardest wasn't the timeline predictions. It was the national security framing. The idea that whoever builds superintelligence first effectively wins — and that this is currently being treated like a startup competition rather than a Manhattan Project.

Whether you agree with the exact timelines or not, the directional argument is hard to dismiss. The compute trendlines are real. The algorithmic progress is real. And the gap between "what researchers believe privately" and "what institutions are planning for" is enormous.

Read the full thing. It's long. It's worth it.