Anthropic 的 Fiona Fung 探讨 AI 如何改变软件工程，工程师代码产出提升 8 倍，并分享关于主动性、责任感和编程未来的见解。
Anthropic's Fiona Fung discusses how AI has transformed software engineering, with engineers shipping 8x more code, and shares insights on agency, accountability, and the future of coding.
要点 · TL;DR
AI 提升了工程师的能力上限，编码不再是瓶颈。 AI lifts the ceiling on what engineers can do, making coding no longer the bottleneck.
高自主权与高责任感是 AI 时代团队的关键。 High agency with high accountability is key for AI-era teams.
随着代码速度提升，验证和质量框架变得至关重要。 Verification and quality frameworks become critical as code velocity increases.
核心观点 · Key points
编码不再是瓶颈；它提升了任何人所能做到的上限。 Coding is no longer the bottleneck; it's lifted the ceiling of what anyone can do.
高自主性与高责任感是 AI 时代团队的关键。 High agency with high accountability is key for teams in the AI era.
成长心态和积极拥抱 AI 是成功的关键。 Growth mindset and leaning into AI are essential for thriving.
随着代码速度提升，验证和质量框架变得更加重要。 Verification and quality frameworks become more important as code velocity increases.
管理者应亲自使用产品并保持动手，以保持对产品的感知。 Managers should dogfood their product and stay hands-on to keep pulse.
由于变化迅速，规划应做到即时和轻量级。 Planning should be just-in-time and lightweight due to rapid change.
反共识 · Contrarian takes
工程师与智能体协作可能感到孤独；结对编程午餐会有所帮助。 Engineers may become lonely working with agents; pair programming lunches help.
编码中的心流状态乐趣已失；满足感现在来自产品影响。 The joy of flow state in coding is lost; satisfaction now comes from product impact.
代码行数等指标可能误导；应关注结果而非产出。 Metrics like lines of code can mislead; focus on outcomes over output.
非工程师如产品经理现在也能提交代码，角色边界变得模糊。 Non-engineers like PMs now ship code, blurring role boundaries.
异步智能体增加了上下文切换负担；需要预留专注时间。 Context switching load increases with async agents; need to block focus time.
培养下一代工程师可能需要学徒模式，而非传统路径。 Growing next-gen engineers may require apprenticeship models, not traditional paths.
本期章节 · Chapters（共 34）
引言与 Fiona 背景Introduction and Fiona's background
从 CD 转向在线发货Shift from CD to online shipping
赞助商插播：WorkOSSponsor break: WorkOS
2026 年 AI 驱动的软件团队AI-pilled software team in 2026
紧跟反馈与交付Staying on top of feedback and shipping
代码审查与测试驱动开发Code review and test-driven development
招聘画像：创意构建者与系统专家Hiring profiles: creative builders and deep systems experts
产品主人翁心态Product ownership mindset
卓越与受挫工程师Thriving vs frustrated engineers
成长心态与直面恐惧Growth mindset and facing fear
AI 采用者与非采用者差距The gap between AI adopters and non-adopters
如何推广 AI 采用How to spread AI adoption
分享 AI 用例作为话题Sharing AI use cases as conversation starters
工程的下一个前沿The next frontier in engineering
例行公事与异步工作Routines and Asynchronous Work
倾听之旅与指标陷阱Listening Tour and Metric Pitfalls
平衡速度、质量与影响Balancing Speed, Quality, and Impact
脏话仪表盘与评估Swear Word Dashboard and Evals
Mercury 广告Mercury Ad
管理者即 IC 与球员教练Manager as IC and Player-Coach Approach
从 Meta 到 Anthropic 转型Transition from Meta to Anthropic