Situational Awareness: The Decade Ahead is a June 2024 essay series by 📝Leopold Aschenbrenner arguing that 📝AGI by 📝2027 is plausible and that 📝superintelligence will arrive before the decade ends.
The series is structured as five parts plus an introduction, published as a single website and a 165-page PDF. Part I, "From GPT-4 to AGI: Counting the OOMs," traces compute, algorithmic efficiency, and "unhobbling" trendlines to argue that the GPT-2-to-GPT-4 jump will repeat by 2027. Part II describes a post-AGI intelligence explosion in which hundreds of millions of automated AI researchers compress a decade of progress into a year. The remaining sections cover the trillion-dollar compute buildout, lab security against state actors, the unsolved superalignment problem, the geopolitical race against the CCP, and an anticipated US government "Project" by 2027–28. The essay is dedicated to 📝Ilya Sutskever, and was released alongside a same-day Dwarkesh Patel podcast appearance and the launch of Situational Awareness LP, the AGI-thesis hedge fund built around the essay's investment claims.
The series became one of the most-circulated AI policy documents of 2024 and the reference text for the "AGI by 2027" position now common in frontier AI discourse. The fund subsequently scaled from roughly $225 million in seed capital to approximately $5.5 billion in reported US equity exposure by Q1 2026, making the essay one of the more consequential pieces of investor-thesis writing in recent memory — a manifesto that became a balance sheet.
Worth reading not because the timelines are right but because the document compresses the SF-insider AGI worldview into a single legible artifact. Whether the 2027 call lands or not, this is the canonical reference for what the people building the systems actually believe.