Mike Krieger 谈 Anthropic 如何使用 Claude Fable 5Mike Krieger on How Anthropic Uses Claude Fable 5
核心要点：Claude Fable 5 标志着 AI 作为可靠队友的飞跃，能够处理复杂、长周期任务且几乎无需监督，从根本上重塑了软件开发工作流。
Anthropic Labs 负责人、Instagram 联合创始人 Mike Krieger 分享了长期使用 Fable 5 如何改变了他的构建者思维。他现在将大量架构规划和 overnight 执行委托给模型，信任它处理中断、搭建解决方案并恢复工作。这将生产力从持续提示转变为高层意图设定和验证。
Krieger 强调周末用 agent-native 架构构建了一个个人媒体追踪器，包括应用中长按聊天实现自我修改代码。他重视验证循环，如截图、视频捕获和动态工作流以获得可靠结果。软件工程并未终结，而是演变——所有权、品味和生产理解仍是人类优势。
“现在感觉它就是一个我可以委托大量工作的队友。”
Anthropic Labs 负责人、Instagram 联合创始人 Mike Krieger 分享了长期使用 Fable 5 如何改变了他的构建者思维。他现在将大量架构规划和 overnight 执行委托给模型，信任它处理中断、搭建解决方案并恢复工作。这将生产力从持续提示转变为高层意图设定和验证。
Krieger 强调周末用 agent-native 架构构建了一个个人媒体追踪器，包括应用中长按聊天实现自我修改代码。他重视验证循环，如截图、视频捕获和动态工作流以获得可靠结果。软件工程并未终结，而是演变——所有权、品味和生产理解仍是人类优势。
“现在感觉它就是一个我可以委托大量工作的队友。”
The Takeaway: Claude Fable 5 represents a leap in AI as reliable teammates capable of handling complex, long-horizon tasks with minimal supervision, fundamentally reshaping software development workflows.
Mike Krieger, Head of Anthropic Labs and Instagram cofounder, shares how extended use of Fable 5 has transformed his builder mindset. He now delegates substantial architectural planning and overnight execution to the model, trusting it to handle interruptions, scaffold solutions, and resume work. This shifts productivity from constant prompting to high-level intent setting and verification.
Krieger highlights building a personal media tracker in a weekend using agent-native architectures, including self-modifying code via long-press chat in the app. He emphasizes verification loops like screenshots, video captures, and dynamic workflows for reliable results. Software engineering isn't over but has evolved—ownership, taste, and production understanding remain human strengths.
"It really feels like now it's a teammate that I can delegate a lot of work to."
查看原文 →
Mike Krieger, Head of Anthropic Labs and Instagram cofounder, shares how extended use of Fable 5 has transformed his builder mindset. He now delegates substantial architectural planning and overnight execution to the model, trusting it to handle interruptions, scaffold solutions, and resume work. This shifts productivity from constant prompting to high-level intent setting and verification.
Krieger highlights building a personal media tracker in a weekend using agent-native architectures, including self-modifying code via long-press chat in the app. He emphasizes verification loops like screenshots, video captures, and dynamic workflows for reliable results. Software engineering isn't over but has evolved—ownership, taste, and production understanding remain human strengths.
"It really feels like now it's a teammate that I can delegate a lot of work to."