Gary Marcus 引用数据：约 75% 的 AI 编程任务仍需人工修整
Gary Marcus 转了项调查，说 AI 编程任务里四分之三还得人手动改代码。如果你在团队里推 Cursor 或 Claude Code，这条数据能帮你算清真实的人力成本，别只看生成速度。
Gary Marcus 在推文中引用一项调查称，AI 编程智能体完成的任务中约四分之三（roughly three out of four）仍需要开发者做人工清理和修正。这一比例意味着当前编程智能体在生产环境中更多扮演辅助角色，而非可独立交付的替代者。
Marcus 的评论带有明显讽刺意味，他在推文中写道"要是当初有人能预料到这一点就好了"，暗指此前部分厂商和评论者对"AI 取代程序员"的预期过于乐观。原推附带引用链接，但目前公开信息中未披露该调查的具体样本量、任务类型和评测方法。
对开发者而言，这条数据点与近期其他关于编程智能体返工成本高的讨论一致：自动生成的代码可以加速起步，但审查、调试和重构环节的人力投入并没有消失。企业在评估 Cursor、Claude Code 这类工具的 ROI 时，需要把人工修整时间计入成本。
AI coding agents still need human cleanup in roughly three out of four tasks. If only someone could have foreseen this! Rajeev Kumar @KumarSamalkha @GaryMarcus Any thoughts on the same. Source --> https://t.co/hXoUBI2kcQ 🔗 View Quoted Tweet📊 3 ⚡
This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. 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 Gary Marcus, 编程助手, AI编程.
Source attribution: curated from Gary Marcus (@GaryMarcus). Cross-source corroboration: 11 independent sources have covered this development (including Sundar Pichai, Varun Mohan, TestingCatalog), confirming industry-wide signal strength.
