开发范式转变：先写验证器，再让智能体生成代码
这条推文一句话点透了智能体编程时代的工作流变化：你写的不再是代码，而是判定对错的验证器。写 Cursor 或 Claude Code 工作流的开发者应该看看，做技术管理的也值得琢磨一下团队分工怎么变。
一条推文对比了两种开发范式：旧世界是先写代码再测试，新世界是先构建验证器（verifier），然后让智能体反复生成代码直到通过验证。
核心变化在于验证环节被前置：人负责定义可自动判定的验收标准，代码产出交给智能体迭代。这种做法依赖自动化验证的可靠性——只要验证器能准确判定对错，智能体的生成-验证循环就能收敛到可用结果。
对开发者来说，写好可执行的测试与规格说明，正在成为比直接写代码更重要的技能。
Old world: Write code, then test it.
New world: Build the verifier, then let agents generate code until it works.
This report is categorized under AITOP's "How-to" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include agents, coding assistants, prompt engineering, Claude Code.
Source attribution: curated from Yuchen Jin (@Yuchenj_UW). Cross-source corroboration: 11 independent sources have covered this development (including claudedevs, IT之家, rohanpaul_ai), confirming industry-wide signal strength.
