Anthropic's New Evals Guide Draws Scrutiny from Hamel Husain

Anthropic released new Evals guidance for agent/LLM evaluation, which expert Hamel Husain is reviewing with peers; after testing the workflow he identified three main issues, most notably that evaluations should be written only after examining the data, not before.

2026-09-29 ~ 2026-09-29 · 3 related posts