
◆ 先别急着上Agent
◆ "AI写了90%的代码"和"产品经理一句话上线"背后
◆ Anthropic真正在建的东西:失败管理系统
这才是 Anthropic 的工程文化:不是乐观地追求 agent 全能,是悲观地假设 agent 一定会出错,然后系统性地管理出错的后果。
◆ 验证是分水岭
◆ 连Anthropic自己都没做到非编码场景的全自动闭环
连 Anthropic 这样一个拥有最强模型、最深Agent基建、最成熟工程文化的公司,在自己的非编码业务场景中,都还没做到全自动闭环。那些认为只要搭好agent研发流水线就能解决业务问题的公司,需要认真想想自己的预期是不是建立在一个根本不存在的参照物上。
◆ 被忽略的那一半
◆ 校准你的预期
◆ 真正的机会在哪里
◆ 参考资料
Anthropic官方工程博客与文档
[1] Building effective agentshttps://www.anthropic.com/engineering/building-effective-agents
[2] Effective context engineering for AI agentshttps://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents
[3] Writing effective tools for agents — with agentshttps://www.anthropic.com/engineering/writing-tools-for-agents
[4] Effective harnesses for long-running agentshttps://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents
[5] Best practices for Claude Codehttps://code.claude.com/docs/en/best-practices
[7] How we built our multi-agent research systemhttps://www.anthropic.com/engineering/multi-agent-research-system
[8] How we contain Claude across productshttps://www.anthropic.com/engineering/how-we-contain-claude-across-products
[9] Scaling Managed Agentshttps://www.anthropic.com/engineering/managed-agents
[10] Demystifying evals for AI agentshttps://www.anthropic.com/engineering/demystifying-evals
[11] Quantifying infrastructure noise in agentic codinghttps://www.anthropic.com/engineering/infrastructure-noise
[18] How Anthropic teams use Claude Codehttps://www.anthropic.com/news/how-anthropic-teams-use-claude-code
[19] How AI Is Transforming Work at Anthropichttps://www.anthropic.com/research/how-ai-is-transforming-work-at-anthropic
[25] An update on recent Claude Code quality reports (April 23 postmortem)https://www.anthropic.com/engineering/april-23-postmortem
[28] When AI builds itselfhttps://www.anthropic.com/institute/recursive-self-improvement
Anthropic安全与合规
[20] Constitutional Classifiers: Defending against universal jailbreakshttps://www.anthropic.com/research/constitutional-classifiers
[21] Next-generation Constitutional Classifiershttps://www.anthropic.com/research/next-generation-constitutional-classifiers
[22] Detecting and countering malicious uses of Claude: March 2025https://www.anthropic.com/news/detecting-and-countering-malicious-uses-of-claude-march-2025
[23] Responsible Scaling Policyhttps://www.anthropic.com/responsible-scaling-policy
[24] Building safeguards for Claudehttps://www.anthropic.com/news/building-safeguards-for-claude
Anthropic产品与医疗
[13] Introducing Anthropic Labshttps://www.anthropic.com/news/introducing-anthropic-labs
[17] Anthropic Careershttps://www.anthropic.com/careers
[26] Advancing Claude in healthcare and the life scienceshttps://www.anthropic.com/news/healthcare-life-sciences
[27] Qualified Health case studyhttps://claude.com/customers/qualified-health
Anthropic产品与生态
Introducing the Model Context Protocolhttps://www.anthropic.com/news/model-context-protocol
New capabilities for building agents on the Anthropic APIhttps://www.anthropic.com/news/agent-capabilities-api
Claude's Constitutionhttps://www.anthropic.com/constitution
高管访谈
[6] Dario Amodei, CEO Speaker Series, Council on Foreign Relationshttps://www.cfr.org/event/ceo-speaker-series-dario-amodei-anthropic
[12] Mike Krieger on Lenny's Podcast, "Anthropic's CPO on what comes next"https://www.lennysnewsletter.com/p/anthropics-cpo-heres-what-comes-next
[14] Dario Amodei on Dwarkesh Podcast, "We are near the end of the exponential"https://www.dwarkesh.com/p/dario-amodei-2
[15] Daniela Amodei on Vanta Frameworks for Growthhttps://www.vanta.com/resources/hiring-10x-engineers-at-anthropic
[16] Daniela Amodei on Sixth Street, It's Not Magichttps://sixthstreet.com/podcasts/a-conversation-with-daniela-amodei-co-founder-and-president-of-anthropic/
[29] Boris Cherny & Cat Wu, "Reflecting on a year of Claude Code"https://www.youtube.com/watch?v=Hth_tLaC2j8
夜雨聆风