Beyond product lists

Research that helps you see what the claims leave out.

Our approach to benchmarks, product intelligence, evidence, and explainers, and an honest account of its limits.

What we do

Repeatable tests with published tasks, conditions, sample sizes, and limitations. None have been published yet.

Tracking version changes, pricing, deployment options, and documentation, each with a dated source.

Source links, verification dates, and claim scope for every important claim.

Plain-language explainers that separate current software from research on, and speculation about, advanced machine intelligence.

Why the word 'superintelligence' in a brand name, a headline, or a policy document does not make a current product superintelligent.

A checklist of questions that determine whether a benchmark number tells you anything about the tool you plan to use.

We have not run any benchmark or product test yet. Everything published so far is methodology and explanation.
Results from tests of AI products can vary between runs and change with product versions. We will record version and date, and results will expire.
Where we summarise third-party research we have not reproduced it, and we label it that way.