frantically scans ai 2027 for the mid 2026 paragraph
Leopold Aschenbrenner's 2022 prediction of a US AGI Manhattan Project by 2027 draws fresh evaluation from frontier researchers
Samuel Albanie of Google DeepMind tracks mid-2026 capability milestones.
Many users praised Leopold Aschenbrenner's two-year-old AGI Manhattan Project prediction as remarkably accurate and prescient while some dismissed it as vague or an attempt to hyperstition dystopia.
Most Activity

context https://situational-awareness.ai/the-project/
If you haven't read the entire thing, you should - now more than ever

@coffeedev @leopoldasch I mean this is not really on point. If you just make vague generalized predictions about the government being scared about X thing you can post hoc rationalize it into being true no matter what happens lol

@coffeedev @leopoldasch

@coffeedev @leopoldasch Impressive indeed!
Check this fragment of something written in 1984

@coffeedev @leopoldasch he’s intentionally trying to hyperstition dystopia, it’s not a mere prediction

@coffeedev The simpler explanation is that the government is punishing Anthropic for standing up to it, especially ahead of its IPO. Reminding the firm that the feds can still destroy them even if they lose on some legal grounds. https://www.cato.org/legal-briefs/anthropic-v-department-war

@coffeedev @leopoldasch Yup

@coffeedev @leopoldasch What a great read. Thanks for sharing.

The military and political complex in DC cares about 2 parts of AI: #1: chips that power autonomous drones (navigation, target selection, etc.), and
#2: ASI. Super-Intelligence.
Government efforts could turbocharge the buildout of datacenters for ASI, but private $ are already building data centers, coding, and training AI at a high speed rate.

@coffeedev @leopoldasch legend he literally wrote 2026 two years ago, when's the movie

@biggestballerik @coffeedev @leopoldasch How is it not on point? If you read his essay pretty much everything has been accurate

@coffeedev @leopoldasch Straits still closed.

@coffeedev AI/ML Manhattan Project already happened in the 70s and 80s. Findings were operationalized in the 90s/00s. Google was how they indexed and expanded access to data. We won't see what this guy is talking about because it already happened.

@coffeedev @leopoldasch in crazy times we live for sure

@coffeedev @leopoldasch Almost as if he had insider knowledge

@coffeedev @leopoldasch

@SamuelAlbanie need to consult the prophecy

@coffeedev @leopoldasch The goat turned whale

@coffeedev @leopoldasch hey diem me right now

@coffeedev @leopoldasch That is some wild precog there dude