Research area · 01
Artificial Intelligence
Capability measurement, training economics, evaluation integrity and the trajectory of general-purpose systems.
Scope
We study artificial intelligence as an empirical subject rather than a narrative one. Our work here concentrates on what can be measured: training compute, data availability, benchmark validity, deployment scale and the gap between demonstrated and reported capability. We are deliberately conservative about extrapolation, and we publish the intervals around our estimates rather than the midpoints alone.
3 publications
Open in the archive- 01
The AI Frontier Report
A structural analysis of artificial intelligence at the close of its fourth compounding year — compute, capital, capability, labour, sovereignty and risk.
NewPDFFirst EditionAreaArtificial IntelligencePublished08 AUG 2026Report · 4 hr read - 02
Evaluation Integrity
A contamination and construct-validity audit of twenty-eight public benchmarks.
In reviewAreaArtificial IntelligencePublished22 JUL 2026Paper · 17 min read - 03
The Data Wall
Estimating the remaining stock of high-quality human text and what follows from its exhaustion.
PDFAreaArtificial IntelligencePublished19 NOV 2025Paper · 15 min read
Research notes
Verification cost explains most of the capability distribution
Where checking an answer is cheap, systems improve fast. Where it is expensive, they do not.
Artificial Intelligence30 June 2026 · 5 min readAgainst single-number capability claims
The most cited figures in AI are the least informative ones.
Artificial Intelligence30 April 2026 · 4 min read
Other areas