Development on Youqing Han - DevOps Engineer

Vibe Coding 上线之后：一次 150+ commit 的复盘

ihanyouqing#gmail.com (#->@) (Youqing Han) — Mon, 14 Sep 2026 11:00:00 +0800

演示里 vibe coding 很爽：自然语言描述需求，几秒就出一版能跑的代码。真正麻烦的是上线以后。

我参与的一个多服务集成功能：用 AI 辅助开发，功能第一周就能演示，但修到生产可用断断续续花了一个多月，累计 150+ commits，其中大约六成是 fix——回归、边界、字段名对不上、并发坑。能力不是问题，缺的是工作方式。

一句话概括我后来信的原则：

AI 写出来的代码，质量上限 ≈ 你给它的规格有多完整。

下面是按那次复盘整理出来的经验和教训。

六类常见翻车

那次 post-mortem 里，问题大致落在这六类，后面复杂功能里也反复撞见：

1. 接口契约漂移

同一概念拆到多个模块、多个 AI session 里实现时，命名各自为政。数据模型里是 user_identifier，API 变成 user_id，前端又写成 userId。AI 没有跨会话记忆，没人盯就会散。

表现： 联调时才发现字段对不上；前端空白或显示错数据。

2. 外部系统的边角料

训练数据里有通用文档，没有你们线上那版 SDK 的怪脾气。空字符串被拒、锁名最长 64、非法 ID 抛的是 validation 而不是 not-found——文档往往写得轻描淡写。

表现： 本地 happy path 全绿，生产里冒出没处理过的异常。

3. 内部基础设施的隐性约定

审批流在非生产环境会跳步、异步任务要等事务提交后再发、分布式锁 key 有长度限制……这些东西通常不在 AI 能读到的文档里。它会按「通用正确」来写。

表现： 单测/隔离测都过，进 staging 端到端就怪。

4. 并发语义想当然

看起来像有保护的 ORM 调用，常常没有。update() 按状态过滤再改计数，可能返回「匹配 1 行」——匹配了，但未必改了。行锁挡不了两个并发 insert 同时通过 uniqueness 检查。

DevHelper: The Ultimate Native macOS Developer Toolkit

ihanyouqing#gmail.com (#->@) (Youqing Han) — Wed, 03 Sep 2025 00:00:00 +0000

Introduction

DevHelper appears to be a comprehensive native macOS application designed specifically for software developers, DevOps engineers, and technical professionals who need efficient access to essential development tools in a unified, high-performance environment.

Target Audience

Software Developers: Frontend, backend, and full-stack developers
DevOps Engineers: Infrastructure and automation specialists
Data Engineers: Working with various data formats and APIs
Security Developers: Cryptographic and security testing needs
QA Engineers: API testing and data validation requirements

Use Cases

Daily Development: Quick access to common utilities during coding sessions
API Development: Testing and debugging HTTP endpoints
Data Processing: Formatting and validating JSON, SQL, and other data formats
Security Testing: JWT validation, cryptographic operations, and hash generation
Code Review: Formatting and comparing code snippets
Troubleshooting: IP queries, timestamp conversions, and regex testing

Core Functionality

DevHelper consolidates 16 essential development tools into a single, native macOS application, eliminating the need to switch between multiple web-based tools or browser tabs.

Complete macOS Development Environment Setup for SRE/DevOps Engineers

ihanyouqing#gmail.com (#->@) (Youqing Han) — Mon, 01 Sep 2025 00:00:00 +0000

Introduction

This guide is a synthesis of my personal configuration journey and best practices for SRE/DevOps engineers. It details how to set up a robust, reproducible macOS development environment, covering essential tools, configuration management, and automation strategies. The goal is to make environment setup painless, portable, and scalable—whether you’re starting over, joining a new team, or enabling seamless collaboration.

Prerequisites

Before we begin, ensure you have:

macOS 12.0 (Monterey) or later
Administrator access to your machine
Basic familiarity with command line operations
A GitHub account for configuration storage

Core Development Tools

Install required develop tools & Set update/upgrade

Python Version Management Tools: A Comprehensive Guide

ihanyouqing#gmail.com (#->@) (Youqing Han) — Fri, 22 Aug 2025 00:00:00 +0000

Introduction

Python’s ecosystem offers a diverse range of tools for managing versions, dependencies, and virtual environments. Each tool serves specific use cases and comes with its own advantages and trade-offs. In this comprehensive guide, we’ll explore the most popular Python version management tools and their pros and cons.

Core Package Management Tools

pip

pip is Python’s default package installer and the foundation of Python package management. It’s the standard tool for installing packages from the Python Package Index (PyPI) and other package indexes.