Event ·Multi-source

Lenny Rachitsky 推出进阶版 AI 评测系统指南

Lenny Rachitsky 联合 Hamel Husain 和 Shreya Shankar 撰写了《Building eval systems》的进阶续篇。文章提到他上周分享的 25 个 PM 职位中近一半要求会写 evals。案例包括 Ramp 将自动收据采集准确率从 35% 提升到 83%,Shopify 的 AI workflow builder 比原前沿模型方案快 2.2 倍、便宜 68%。Harvey 重建 AI 合同审查器后内部质量分接近翻倍,Cursor 的 Auto Balance 路由在成本降低 41% 的同时用户满意度提升。

Current read

做 AI 产品的话这篇挺实用:Ramp、Shopify、Cursor 这些公司靠写 evals 把指标提了多少,还附了个免费插件让编码 Agent 帮你干活。

6 sources73° AI heatLast updated 2026/9/22 23:51:23

Evidence

6 sources
Primary sourceLenny Rachitsky
View source
Related source 2Hamel Husain
View source
Related source 3shao__meng
View source
Related source 5eric zakariasson
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.