Harness是什么？

ref: 于超老师的对谈 https://mp.weixin.qq.com/s/KSXKy7nVAQ8eshrBRtGnUA Lili Weng’s blog: https://lilianweng.github.io/posts/2026-07-04-harness/

Harness 这一概念产生于智能体的迭代自我进化（RSI, recursive self-improvement): an AI uses its current intelligence to improve the cognitive machinery that produces its intelligence.

放到如今的AI语境中，就是AI在一个反馈loop中更新自己的权重，甚至改进自己的训练pipeline，部署系统，以达到更好的效果。

I explicitly mention “deployment system” because the layer between the raw model and the real-world context seems to be as important as the model’s raw intelligence (i.e. the evals right after pretraining). Harnesses are important components of AI deployment, as shown by successful coding agent products such as Claude Code and Codex. A harness is the system surrounding a base model that orchestrates execution and decides how the model thinks and plans, calls tools and acts, perceives and manages context, stores artifacts, and evaluates results.

Harness包括AI外围的一些infra，比如其能够调用的工具，context, 存储，测评工具，等。