Will there be an intelligence explosion in 2027?

What an intelligence explosion means here

In Situational Awareness, the intelligence explosion is the hinge of the whole argument: once AI reaches the level of a competent AI researcher (the AGI-2027 milestone), it can be turned on improving AI itself. Hundreds of thousands of automated researchers running in parallel would, Aschenbrenner argues, compress roughly a decade of algorithmic progress into about a year — vaulting from AGI to superintelligence by the late 2020s or early 2030s.

Why it's Pending, not Open or Wrong

This prediction is downstream of AGI-2027. Its precondition — a system that can autonomously conduct AI research end-to-end — has not been demonstrated as of mid-2026. Agentic coding is strong (~80% on SWE-Bench Pro), but that is assisting human researchers, not replacing the research loop. Because the trigger hasn't fired and the 2027–29 window hasn't arrived, there is simply not enough evidence to grade it either way. That's what Pending means on this scorecard.

What to watch

The leading indicator is the AGI-2027 verdict. If a system demonstrably automates a meaningful fraction of AI research engineering — the bar that would flip AGI-2027 from Open — the intelligence-explosion clock effectively starts. Until then, this remains the most speculative claim on the scorecard, and the one furthest from being testable.

Status check, 17 August 2026. At the dataset’s last revision (data.json, dateModified 2026-08-17) this prediction still reads Pending, with the recorded evidence line “Precondition (autonomous AI research) undemonstrated.” Worth noting what is not the blocker: the neighbouring compute-scaling prediction reads On track — “Effective-compute pace roughly held through mid-2026.” Hardware and effective compute kept their end of the bet; the unfired trigger is that no system has yet been shown to run the AI-research loop itself.

Frequently asked questions

Too early to tell. Aschenbrenner predicted one beginning 2027–29, triggered by AI automating AI research. As of mid-2026 that precondition is undemonstrated, so the prediction is graded Pending.

The idea that once AI can do AI research, it rapidly improves itself — compressing roughly a decade of progress into about a year and vaulting from AGI to superintelligence.

Because its trigger — a system that can autonomously conduct AI research end-to-end — hasn't been demonstrated, and the 2027–29 window hasn't arrived. There isn't yet evidence to grade it either way.