一个技术探索者关于'也许可行'的笔记 | 跨界技术探索者的成长手册 2025-11-04T04:05:37.806Z https://www.o0o0o.sbs/ MaybeItWorks Hexo https://www.o0o0o.sbs/2025/11/04/%E9%97%AEAI%E5%B0%B1%E5%A5%BD%E4%BA%86-%E5%85%B3%E4%BA%8E%E5%A4%A7%E6%A8%A1%E5%9E%8B%E7%9A%84%E9%82%A3%E4%BA%9B%E5%9F%BA%E7%A1%80%E8%AE%A4%E7%9F%A5/ 2025-11-04T01:00:00.000Z 2025-11-04T04:05:37.806Z

当团队中有人说"问AI就好了"，有人嘲讽"上下文工程"，还有人对基础概念一知半解时，我们真正需要的是什么？

https://www.o0o0o.sbs/2025/10/21/Agent%E9%A9%B1%E5%8A%A8%E7%9A%84%E6%B5%8B%E8%AF%95%E5%8F%98%E6%9B%B4%E7%AE%A1%E7%90%86%E7%B3%BB%E7%BB%9F%E8%AE%BE%E8%AE%A1%E5%AE%9E%E5%BD%95/ 2025-10-21T06:30:00.000Z 2025-11-04T04:05:37.758Z

Agent自动化测试开发时，最大的挑战不是生成代码，而是如何管理变更。从10节点工作流到4阶段设计，从索引文档到规范体系，这是一次关于如何在规范性与灵活性之间寻找平衡点的探索。When we tried to automate test development with AI Agent, the biggest challenge was not generating code, but managing changes. From a 10-node workflow to a 4-phase design, from index documents to a specification system, this is an exploration of finding the balance between standardization and flexibility.

https://www.o0o0o.sbs/2025/08/10/%E4%BB%8EReAct%E5%88%B0%E6%B7%B7%E5%90%88%E6%9E%B6%E6%9E%84%E7%9A%84%E6%8A%80%E6%9C%AF%E9%80%89%E6%8B%A9%E8%89%BA%E6%9C%AF/ 2025-08-10T06:30:00.000Z 2025-11-04T04:05:37.770Z

当我们面对Agent技术栈的多种选择时，每一种都声称能解决我们的问题。但在软件在环测试这个特定场景下，ReAct的观察-推理-行动循环真的比Function Calling的确定性工具调用更优吗？经过深入的技术对比和实践验证，我发现答案远比想象中复杂。When facing multiple choices in the Agent technology stack, each claims to solve our problems. But in the specific scenario of software-in-the-loop testing, is ReAct's observe-reason-act cycle really better than Function Calling's deterministic tool invocation?

https://www.o0o0o.sbs/2025/08/10/%E7%8E%B0%E6%9C%89%E6%B5%8B%E8%AF%95%E8%87%AA%E5%8A%A8%E5%8C%96%E6%96%B9%E6%A1%88%E4%B8%BA%E4%BD%95%E5%A4%B1%E6%95%88%EF%BC%9A%E4%BB%8E%E5%B7%A5%E5%85%B7%E5%B1%80%E9%99%90%E5%88%B0%E8%AE%A4%E7%9F%A5%E8%B4%9F%E8%8D%B7%E7%9A%84%E5%85%A8%E9%9D%A2%E5%88%86%E6%9E%90/ 2025-08-10T00:30:00.000Z 2025-11-04T04:05:37.794Z

当我们第三次因为CMake配置问题而测试环境搭建失败时，开始怀疑：为什么这些看起来很完美的自动化工具，在实际使用中总是让人失望？问题可能比我们想象的更根本。When our test environment setup failed for the third time due to CMake configuration issues, we began to wonder: why do these seemingly perfect automation tools always disappoint in practice?

https://www.o0o0o.sbs/2025/08/09/%E8%BD%AF%E4%BB%B6%E5%9C%A8%E7%8E%AF%E6%B5%8B%E8%AF%95%E7%9A%84%E9%87%8D%E5%A4%8D%E5%8A%B3%E5%8A%A8%E5%9B%B0%E5%A2%83/ 2025-08-09T05:30:00.000Z 2025-11-04T04:05:37.798Z

当我们第十次因为相同的依赖配置问题卡住时，终于意识到：68.9%的工作时间消耗在具有高度重复性特征的任务上。这些重复劳动背后有清晰的模式，而解决它们将带来显著的生产力提升。When we got stuck on the same dependency configuration issue for the tenth time, we finally realized: 68.9% of work time is consumed on highly repetitive tasks.

https://www.o0o0o.sbs/2025/08/09/Agent%E5%8C%96%E6%B5%8B%E8%AF%95%E6%A1%86%E6%9E%B6%E7%9A%84%E5%B7%A5%E7%A8%8B%E5%AE%9E%E7%8E%B0%E5%85%A8%E9%93%BE%E8%B7%AF%E8%A7%A3%E6%9E%90/ 2025-08-09T04:30:00.000Z 2025-11-04T04:05:37.754Z

当我第一次尝试让Agent自动生成测试代码时，发现真正的挑战不在于生成代码本身，而在于如何构建一个完整的、可靠的实现链路。从需求文档的结构化解析，到esim接口的智能映射，再到编译错误的自动恢复，每一个环节都需要深入的思考和设计。When we first tried to let Agent automatically generate test code, we found that the real challenge is not in generating code itself, but in building a complete and reliable implementation pipeline.

https://www.o0o0o.sbs/2025/07/24/%E7%89%B9%E5%BE%81%E6%A3%80%E6%B5%8B%E7%AE%97%E6%B3%95%E7%9A%84%E5%B7%A5%E7%A8%8B%E5%8C%96%E4%B9%8B%E8%B7%AF/ 2025-07-24T12:30:00.000Z 2025-11-04T04:05:37.790Z

当算法需要在资源受限的边缘设备上7×24小时运行时，工程化比算法本身更重要

https://www.o0o0o.sbs/2025/07/23/%E4%BB%8E%E6%B8%B8%E6%88%8F%E4%BF%A1%E5%8F%B7%E5%88%B0%E7%9C%9F%E5%AE%9E%E6%89%8B%E6%84%9F%EF%BC%9A%E6%8E%A7%E5%88%B6%E7%B3%BB%E7%BB%9F%E4%B8%AD%E7%9A%84%22%E4%BD%93%E9%AA%8C%E6%98%A0%E5%B0%84%22/ 2025-07-23T12:30:00.000Z 2025-11-04T04:05:37.782Z

当数字世界需要转化为物理感受时，控制系统面临的不只是技术问题

https://www.o0o0o.sbs/2025/07/22/%E9%87%8D%E6%96%B0%E7%90%86%E8%A7%A3%22%E6%9C%89%E6%84%8F%E4%B9%89%22%E7%9A%84%E4%BF%A1%E5%8F%B7%E7%89%B9%E5%BE%81/ 2025-07-22T04:30:00.000Z 2025-11-04T04:05:37.802Z

从一次意外的睡眠监测经历开始，重新思考什么样的信号特征才是真正有价值的

https://www.o0o0o.sbs/2025/07/21/AI%E5%BA%94%E7%94%A8%E5%BC%80%E5%8F%91%E4%B8%AD%E7%9A%84%E8%AE%A4%E7%9F%A5%E9%99%B7%E9%98%B1/ 2025-07-21T05:30:00.000Z 2025-11-04T04:05:37.754Z

探讨技术产品人如何正确认识自己的知识边界

https://www.o0o0o.sbs/2025/07/20/MBD%E6%B5%81%E7%A8%8B%E7%9A%84CICD%E5%AE%9E%E8%B7%B5/ 2025-07-20T06:30:00.000Z 2025-11-04T04:05:37.762Z

从手工协作到自动化流程，一次MBD工作流的CI/CD改造实践记录

https://www.o0o0o.sbs/2025/07/16/%E4%BB%8E%E6%9D%BF%E7%90%83%E5%AE%9E%E9%AA%8C%E9%87%8D%E6%96%B0%E7%90%86%E8%A7%A3%E6%8E%A7%E5%88%B6%E6%80%9D%E7%BB%B4/ 2025-07-16T12:30:00.000Z 2025-11-04T04:05:37.782Z

从经典的平板小球系统出发，重新梳理控制理论的核心概念和思维方式

https://www.o0o0o.sbs/2025/07/15/%E4%BB%8EMock%E5%88%B0%E7%9C%9F%E5%AE%9E%E7%BB%84%E4%BB%B6%EF%BC%9A%E4%B8%80%E6%AC%A1%E6%B5%8B%E8%AF%95%E6%A1%86%E6%9E%B6%E6%9E%B6%E6%9E%84%E6%BC%94%E8%BF%9B%E7%9A%84%E5%AE%9E%E8%B7%B5%E8%AE%B0%E5%BD%95/ 2025-07-15T10:56:58.000Z 2025-11-04T04:05:37.770Z

软件在环测试中，我们习惯用简化Mock模拟外设组件。项目初期觉得这思路完美：快速、简单、可控。但随着需求复杂化，仿真框架陷入了'Mock地狱'——每次需求变化就得改Mock，真实组件更新又得同步Mock。写完200多行踏板响应曲线的Mock后突然意识到：这TM不就是重新实现一遍真实组件吗？ Started with lightweight mocks for hardware components, thinking it was perfect. But as requirements grew complex, I found myself in 'Mock Hell' - constantly updating and maintaining simplified versions of real components.

https://www.o0o0o.sbs/2023/07/01/MPLS%E6%8A%80%E6%9C%AF%E6%B7%B1%E5%BA%A6%E8%A7%A3%E6%9E%90%EF%BC%9A%E4%BB%8E%E6%A0%87%E7%AD%BE%E4%BA%A4%E6%8D%A2%E5%88%B0%E6%B5%81%E9%87%8F%E5%B7%A5%E7%A8%8B/ 2023-07-01T06:30:00.000Z 2025-11-04T04:05:37.762Z

深入解析MPLS协议栈的核心机制，从标签分发到流量工程的完整实现

https://www.o0o0o.sbs/2022/08/10/%E4%BA%8C%E4%B8%89%E5%B1%82%E8%BD%AC%E5%8F%91%E7%9A%84%E9%82%A3%E4%BA%9B%E4%BA%8B/ 2022-08-10T14:30:00.000Z 2025-11-04T04:05:37.766Z

客户遇到一个让人摸不着头脑的问题：同一VLAN里两台机器，配置看起来都对，但就是ping不通。Wireshark抓包发现ARP请求一直在发，就是收不到回复。经过两小时排查才发现是MAC地址表的坑，这次经历让我重新审视二三层转发的每个细节。 A customer's seemingly impossible network issue led to a 2-hour troubleshooting session that revealed the hidden complexities of basic L2/L3 forwarding.