Andrew 讨论了编程代理的发展速度超出预期，以及更快的软件开发如何导致产品管理之外的瓶颈，从而催生出小型、赋能的通才团队。
Andrew discusses how coding agents have evolved faster than expected, and how faster software development creates bottlenecks beyond product management, leading to small, empowered teams of generalists.
要点 · TL;DR
编码智能体进展超预期，领先工具快速更迭。 Coding agents advance faster than expected, shifting tools rapidly.
企业 AI 需要自下而上的创新和自上而下的流程再造。 Enterprise AI needs bottom-up innovation and top-down workflow redesign.
构建速度加快加剧产品管理瓶颈，小团队更有效。 Product management bottleneck worsens as building speeds up; small teams win.
核心观点 · Key points
编码智能体的进展比预期更快，领先工具快速更迭。 Coding agents have advanced faster than expected, with rapid shifts in leading tools.
企业采用 AI 需要自下而上的创新和自上而下的工作流重新设计相结合。 AI adoption in enterprises requires both bottom-up innovation and top-down workflow redesign.
随着构建速度加快，产品管理瓶颈加剧；小型通才团队是关键。 The product management bottleneck worsens as building becomes faster; small generalist teams are key.
非结构化数据架构是主要痛点；使数据为 AI 就绪是未来的巨大工程。 Unstructured data architecture is a major pain point; making data AI-ready is a huge future project.
教育变革被过度炒作；互动课程虽有改进但尚未革命性突破。 Education transformation is overhyped; interactive courses are better but not revolutionary yet.
反共识 · Contrarian takes
关于工作末日的悲观论调被夸大；AI 不会导致大规模失业。 Doomsaying about jobpocalypse is overblown; AI will not cause mass unemployment.
自下而上的 AI 创新往往只带来增量收益，而非变革性投资回报。 Bottom-up AI innovation often yields only incremental gains, not transformative ROI.
推动 20%增长比 2%增长更容易，因为它迫使人们寻找创造性解决方案。 Driving 20% growth is easier than 2% growth because it forces creative solutions.
供应商锁定很危险；绝不签超过一年的合同以保持选择权。 Vendor lock-in is dangerous; never sign contracts longer than one year to preserve optionality.
对于 AI 原型开发，NoSQL 数据库比关系型数据库能实现更快的迭代。 NoSQL databases enable faster iteration than relational ones for AI prototyping.
开放权重模型仍落后前沿模型 6-9 个月，但对许多用例很有价值。 Open-weight models remain 6-9 months behind frontier models but are valuable for many use cases.
本期章节 · Chapters（共 13）
0. 引言与年度变化Introduction and Year-over-Year Changes
1. 软件工程的未来Future of Software Engineering
2. 背景与新人建议Background and Advice for New Software Engineers
3. 构建模块与智能编码Building blocks and agentic coding
4. 企业 AI 采用与 ROIEnterprise AI adoption and ROI
5. 分析数百个 AI 创意Analyzing Hundreds of AI Ideas
6. 前线工程师与供应商中立Forward-Deployed Engineers and Vendor Neutrality
7. 开源模型 vs 前沿模型Open Source Models vs Frontier Models
8. 构建智能体的数据策略Data Strategy for Building Agents
9. AI 的数据架构Data Architecture for AI
10. 现有数据架构的问题Issues with Existing Data Architecture