📅 2026-04-26

Sunday, April 26, 2026

1 sections ~15 min read 🌏 Bilingual 中/EN
📅 Week in Review
← 2026-04-25 📂 All Days 2026-04-27 →

📅 Week in Review — Week 7 (10 min read)

📊 NeetCode: 27/150 · SysDesign: 26/40 · Behavioral: 26/40 · Frontend: 26/50 · AI: 12/30

🔥 31-day streak!

🗓️ This Week's Journey / 本周回顾

  • Monday (Day 27): 设计新闻流 (News Feed)、括号生成 (Generate Parentheses)、闭包 (Closures)、建立信任 (Building Trust)、AI 动态。
  • Tuesday (Day 28): 设计聊天系统 (Chat System)、每日温度 (Daily Temperatures)、异步控制流 (Promises)、自研与外购 (Build vs Buy)、MCP 协议。
  • Wednesday (Day 29): 设计搜索引擎 (Search Engine)、车队 (Car Fleet)、事件循环 (Event Loop)、应对组织变革 (Navigating Org Change)、AI 动态。
  • Thursday (Day 30): 综合复习日 (Review Day)。
  • Friday (Day 31): 设计视频流媒体 (YouTube/Netflix)、直方图最大矩形 (Largest Rectangle in Histogram)、原型与 this (Prototypes & this)、衡量项目成功 (Measuring Success)、AI 动态。
  • Saturday (Deep Dive): 深度解析 Google Maps 架构 (Design Google Maps)。

🧠 System Design: Key Takeaways / 系统设计要点

  1. Fan-out 策略 (Fan-out Strategies): 在新闻流和聊天系统中，写时扇出 (Fan-out on write) 适合普通用户和中小群组，而读时扇出 (Fan-out on read) 适合大V。
  2. 异步与解耦 (Async & Decoupling): 视频转码 (YouTube) 和消息推送 (Chat) 都极度依赖消息队列 (Kafka) 来解耦 CPU 密集型或高并发任务。
  3. 空间与时间索引 (Spatial & Time Indexing): 搜索引擎依赖倒排索引 (Inverted Index)，而 Google Maps 依赖四叉树/Geohash 进行空间索引。
  1. Fan-out Strategies: In News Feeds and Chat Systems, fan-out on write is ideal for normal users/small groups, while fan-out on read is better for celebrities.
  2. Async & Decoupling: Video transcoding (YouTube) and message pushing (Chat) heavily rely on message queues (Kafka) to decouple CPU-heavy or high-concurrency tasks.
  3. Spatial & Time Indexing: Search engines rely on Inverted Indexes, while Google Maps relies on Quadtrees/Geohashes for spatial indexing.

💻 Algorithms: Patterns Mastered / 算法模式总结

单调栈模式 (Monotonic Stack Pattern):

  • #739 Daily Temperatures: 使用单调递减栈寻找下一个更大元素。
  • #853 Car Fleet: 结合排序与单调栈，通过到达时间判断追及关系。
  • #84 Largest Rectangle in Histogram: 使用单调递增栈，在弹出元素时利用左右边界计算最大面积（该模式的最难点）。

Monotonic Stack Pattern:

  • #739 Daily Temperatures: Monotonic decreasing stack to find the next greater element.
  • #853 Car Fleet: Sorting + stack to simulate catch-up scenarios based on arrival time.
  • #84 Largest Rectangle in Histogram: Monotonic increasing stack to calculate max area using left/right boundaries upon popping (the hardest variation).

🗣️ Soft Skills: What to Practice / 软技能练习重点

  • 建立信任 (Building Trust): 在新团队中，信任来自可预测性、透明度和稳定交付，而非单纯的人际关系。
  • Build vs Buy: 评估总拥有成本 (TCO) 和核心竞争力。不要在非核心的通用基础设施上浪费工程资源。
  • 衡量成功 (Measuring Success): 成功不仅是技术上线，还包括用户影响和业务价值。需要建立分层的指标框架。
  • Building Trust: In a new team, trust comes from predictability, transparency, and delivery, not just being nice.
  • Build vs Buy: Evaluate TCO and core differentiators. Don't waste engineering cycles on commodity infrastructure.
  • Measuring Success: Success is not just shipping code; it's user impact and business value. Build a multi-layered metrics framework.

🎨 Frontend: Concepts to Lock In / 前端知识巩固

  • 闭包 (Closures): 函数及其捕获的词法作用域。注意在循环中捕获变量时的陷阱。
  • 异步控制流 (Async Control Flow): Promise.all 用于并行请求，避免不必要的串行 await 阻塞。
  • 事件循环 (Event Loop): 区分微任务 (Microtasks, 如 Promise) 和宏任务 (Macrotasks, 如 setTimeout)，以及它们在渲染前后的执行顺序。
  • 原型与 this (Prototypes & this): 掌握 this 的四种绑定规则，理解箭头函数如何继承外层词法作用域的 this。
  • Closures: Functions + captured lexical scope. Watch out for variable capture traps in loops.
  • Async Control Flow: Use Promise.all for parallel requests to avoid sequential await blocking.
  • Event Loop: Differentiate between Microtasks (Promises) and Macrotasks (setTimeout), and their execution order relative to rendering.
  • Prototypes & this: Master the 4 rules of this binding and how arrow functions inherit lexical this.

🤖 AI: What Stuck / AI 知识点

  • MCP (Model Context Protocol): AI 工具的“USB 标准”，允许大模型以统一、标准化的方式连接外部工具和数据源。
  • AI 动态 (News): 据报道，Google 发布了分离训练和推理的第八代 TPU (8t/8i)，并强制工程师使用内部 AI 编程代理。同时据报道，Anthropic 的 Claude Mythos 帮助 Mozilla 发现了 271 个 Firefox 漏洞。
  • MCP (Model Context Protocol): The "USB standard" for AI tools, allowing LLMs to connect to external tools and data sources in a unified, standardized way.
  • AI News: Reportedly, Google released 8th-gen TPUs (8t/8i) separating training and inference, and mandated internal AI coding agents. Also reportedly, Anthropic's Claude Mythos helped Mozilla find 271 Firefox bugs.

⚠️ What to Review / 需要复习的内容

  • 单调栈的边界处理 (Monotonic Stack Boundaries): 特别是 Largest Rectangle in Histogram 中的 extend-left 技巧，需要反复手写加深肌肉记忆。
  • 事件循环的执行顺序 (Event Loop Execution Order): 容易在复杂的异步嵌套中判断失误，需多做代码输出推演。
  • Monotonic Stack Boundaries: Especially the extend-left trick in Largest Rectangle in Histogram. Needs repeated practice for muscle memory.
  • Event Loop Execution Order: Easy to misjudge in complex async nesting. Needs more code output tracing practice.

🏆 Win of the Week / 本周亮点

成功攻克了单调栈模式中最难的 #84 Largest Rectangle in Histogram，并完成了对 Google Maps 架构的深度解析！

Successfully conquered the hardest monotonic stack problem (#84 Largest Rectangle in Histogram) and completed a deep dive into Google Maps architecture!

🎯 Next Week Preview / 下周预告

  • Algorithms: 二分查找 (Binary Search) 模式。
  • System Design: 分布式事务与共识算法 (Distributed Transactions & Consensus)。
  • Frontend: React 性能优化与底层原理。