Last updated: May 26, 2026
Leopold Aschenbrenner
The clearest recent case for near-term AGI as a strategic mobilization problem.
This source has not been fully researched yet.
Target date
2027
Scoring lens
Blended score
40%
Capability score
45%
Deployment score
25%
The blended score is our overall estimate for the full prediction. The capability score asks, 'Could the technology exist by then?' The deployment score asks, 'Will it be widely used and visible in the real world by then?' For this prediction, the technology may arrive before the world is ready to deploy it safely or broadly.
Exponential lens
Capability can compound quickly. Deployment usually moves slower.
If frontier systems become useful AI researchers, capability gains could compound through automated experimentation, coding, model evaluation, and algorithmic progress.
Even strong capabilities may be slowed by agent reliability, lab security, compute and power bottlenecks, regulation, and the difficulty of deploying autonomous systems into high-stakes institutions.
The big thesis breaks down into visible milestones: timelines, institutional behavior, infrastructure, and capability signals.
| Prediction | Target | Status | Trackability |
|---|---|---|---|
| AGI-level AI researchers | 2027 | Too early | High |
| Models outpace many college graduates | 2025/26 | Actively track | Medium-high |
| Massive compute and power buildout | Late 2020s | Already visible directionally | High |
| Lab security becomes national-security-grade | Before AGI | Early signs only | Medium |
| Government AGI project | 2027/28 | Too early | High |
| Superintelligence before 2030 | Late 2020s | Highly uncertain | Medium |
Target
2027
Status
Too early
Trackability
High
Target
2025/26
Status
Actively track
Trackability
Medium-high
Target
Late 2020s
Status
Already visible directionally
Trackability
High
Target
Before AGI
Status
Early signs only
Trackability
Medium
Target
2027/28
Status
Too early
Trackability
High
Target
Late 2020s
Status
Highly uncertain
Trackability
Medium
Overview
Aschenbrenner argues that scaling, algorithmic progress, and national security dynamics point toward unusually compressed AI timelines.
The big prediction
AGI-level systems become plausible around 2027, followed by rapid superintelligence pressures.
Evidence for
This source has not been fully researched yet.
Keep tracking
