AI evaluation, data strategy, and AI-native organization design.

Yongshuai Li is a research-oriented builder working across agent evaluation, human-agent collaboration, data production systems, and the organizational forms that advanced AI makes possible.

thought
zh

数据生产不是标注工厂

数据生产更像知识生产:它需要任务设计、专家判断、反馈闭环和长期记忆。

#data#strategy
essay
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
project

Evaluation Field Notes

A working archive of evaluation patterns, failure modes, rubrics, and examples from agentic systems.

#evaluation#agents#notes
essay
en

Agent Work Needs Traces

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

#agents#evaluation#workflow
project

AI-Native Organization Map

A conceptual map for how teams change when AI participates in coordination, memory, and judgment.

#organization#strategy#research