AI Agents Can Now Ship Real Code. The Problem Is Knowing When They Shouldn’t.
中文摘要
AI智能体已能编写代码，但验证其输出的成本正不断上升，如何判断何时不应依赖它们正成为工程领域的一大难题。
English Summary
AI agents can now ship real code, but the rising cost of verifying their output is creating a significant "verification gap" in engineering.
原文节选
As agentic tools take on more of the engineering workload, the verification gap is becoming the most expensive part of the job. Continue reading on Medium »