Our Research

How AI products are measured, and what the measurements miss.

Notes from our research and evaluation work. Each one sets out a method in enough detail to be argued with: what was measured, how it was isolated, and what the result does and does not support.

Research noteJuly 202611 min read

Assessing AI, when you cannot see inside it.

The number of ways to build an AI product has outrun the market's ability to tell them apart. This note sets out how to isolate the four components of an AI product — the unit, its data and knowledge, the connections and logic that aim it, and the surface a user touches — and read each one independently.

AI assessmentProduct measurementEvaluation methodology

Caliper Lab is independent and is not affiliated with any AI vendor. Research notes describe our own methods and findings; where a note draws on a measurement we ran, the conditions of that measurement are stated in the note itself.