Writing

Essays and longer notes, often drafted from Chinese-first thinking and shaped for English-facing reach.

zh

评测是一种组织能力

Evaluation Is an Organizational Capability

AI 评测不只是模型上线前的一道门,而是组织持续学习、判断与校准的系统。

#evaluation#organization#data
en

Agent Work Needs Traces

Agent evaluation should treat intermediate traces as first-class evidence, not incidental logs.

#agents#evaluation#workflow