The supporting literature.

A conclusion is only as strong as the documents that refer to it.

The coordination frontier

An interactive exploration of compute growth, human oversight, and the economic contribution of scheduling.

Read supplement ↗

Methods and assumptions

How we selected the evidence, estimated the uncertainty, and decided which uncertainties to leave out.

Read supplement ↗

Reassurance as an alignment metric

A system that says it is safe is easier to evaluate than a system whose safety must be established.

Read supplement ↗

AI 2069: the complete report

The full scenario, both policy endings, research notes, and methodology in a downloadable paper.

Read supplement ↗

Research standards

Our papers undergo internal peer review. The peers are selected for their familiarity with our previous conclusions. A description of the review process is available.