Intelligence explosion

AI improving AI fast enough to compress years of progress into months.

Facts

Lane: Models and thresholds
Wave: 10 of 12 (build order, not a date)
Needs: 22 boxes all the way back through hard links
Unlocks: 0 boxes downstream through hard links
Leverage: None
Cluster: Oversight and alignment
When: Researcher estimate: 2030 to 2038

Evidence as of October 2026

Aschenbrenner describes automated researchers compressing a decade of algorithmic progress into a year or less. Forethought argues that once AI automates AI research, software progress alone could accelerate if returns to research exceed 1.

Watch
Years of algorithmic progress per calendar year.

Timeline note
My estimate, after a superhuman AI researcher, and only if returns to research stay above 1.

01Forethought: Will AI R&D automation cause a software intelligence explosion?

Around it

Needs · 3

Superhuman AI researcherstage · Speculative
Compute supplycapability · Built today
Algorithmic efficiencycapability · Built today

Unlocks · 2

Superintelligencestage · Speculative · helps, not required
Self-replicating robot factoriesstage · Speculative · helps, not required

Brakes

Works against this box

Verified slowdownstage · Speculative

Loops and off-ramps

AI research automationreinforcing loop
Capex pullbackoff-ramp · cut off
Taiwan chip disruptionoff-ramp · cut off
Power shortfalloff-ramp · cut off
Capability plateauoff-ramp · cut off
Training data ruled outoff-ramp · slowed
Alignment failure in deploymentoff-ramp · cut off

Openings researchers' judgement, October 2026 scouting

Startup room 25/100 · rank 121 of 124 Research room 48/100 · rank 71 of 124

Crowding3 Lab overlap5 Small-team fit2 Buyer today1.5 Research openness4

Depth: light
Players: Recursive ($650m or $500m, conflicting); Mirendil ($200m seed); Autoscience ($14m)
Wedge: Whether a software-only explosion is compute-bottlenecked, and how to measure research taste. research