ALPHAZETA / RESEARCH INDEX

AI & Tools

Experiments with AI, agents and the tools of research.

AI evaluation

Generating better answers is easier than recognising them

Producing more candidate answers helps only if you can choose well among them. Judge that ability on the actual task: another model’s agreement is not a guarantee, and using the same model is not an automatic failure.

Published 26 September 2026

Read the article

AI & ToolsTooling case study

Evaluating Jev

How we used agents to evaluate Jev—and changed our minds about its best use

A decline became growth. A promising score missed the error. What paired tests and source checks taught us about using a small model in research.

Research through 24 September 2026

Read the article

Research is dated. Conclusions are conditional.Join the research discussion