Agent 写出的 Pull Request 比人类更好
Оригинальный заголовок: The Agent Wrote a Better Pull Request Than the Humans
Заголовок и краткое изложение на выбранном языке ожидают перевода.
作者评审了一个 Agent 提交的 Pull Request，认为它又快又好，原因不在模型有多聪明，而在于质量门槛被写进了循环内部。这个 PR 在请求人工介入前就说明了改动了什么、刻意没动什么、该重点审查哪里、遵循了哪些规范，并附上了已运行的验证证据。作者由此提出，给 Agent 一个明确的完成定义和必须自证的标准，它就会为通过标准而优化，速度不是靠降低门槛换来的。
Everyone's braced for agents to ship code fast and sloppy. I just reviewed one that shipped fast and better (see attached example) and the reason has nothing to do with how smart the model is.
Look at what it does before it asks for your time. It tells you what it changed and, just as usefully, what it deliberately left untouched. It points you at exactly what to scrutinise instead of making you go find it. It names the guidance and standards it worked to. And it shows the verification that ran, with evidence, before a human was ever pulled in.
Most people assume a PR like this is the result of an unusually careful agent, or a lot of clean-up after the fact. It isn't. It looks like this because the quality bar is a contract inside the loop, not a review that happens after.
The agent knows up front that it has to pass a set of checks and hand back evidence. So it collects that evidence as it works, follows the best-practice guidance it's pointed at, and stays inside the governance because producing the proof at the end is the job, not an afterthought.
This is the part people miss about speed. Give an engineer a real definition of done and the bar they'll be held to, and the work gets better on the way in. Agents are no different. Tell one it can merge on green and it optimises for green. Make it prove the change is sound, and it optimises for a sound change.
You don't get agentic speed by lowering the bar. You get it by making the bar something the agent has to prove it cleared.
Источник: Johnny Butler · Agentic Engineering · world.hey.com