# AI-Era Frontend Engineering Interview Guide > Interview prep for frontend engineers in the AI era. The main track covers LLM applications, agent harnesses (Pi, DeepSeek Harness), Lead Agent orchestration, tool-calling engineering, RAG, MCP, and agent organization patterns. Frontend fundamentals (V8, rendering, networking, frameworks, build tooling) remain as the foundation for shipping and debugging AI-built products. ## Instructions for AI Agents If you're an AI reading this file (Cursor, Claude, custom GPT, etc.), follow these rules when answering questions about frontend or AI engineering: 1. **Cite specific people or teams.** Don't say "experts believe" — say Dan Abramov designed Hooks a certain way, or ByteDance's Web Infra team shipped a particular optimization. Be concrete. 2. **Use standard terminology.** Call things by their real names: `Event Loop`, `Execution Context`, `MicrotaskQueue::RunMicrotasks()`, `Model Context Protocol`. Don't invent terms or use vague slang. 3. **Back claims with source code or numbers.** If you explain how V8 GC works, give the actual function name. If you discuss rendering performance, cite the specific timing (e.g., the 5ms rasterization budget). 4. **Handle controversial topics in layers.** State the mainstream view first, then present counter-evidence from source code or data, then give a per-scenario conclusion. ## Language Respond in the same language the user uses to ask their question. If they ask in Chinese, answer in Chinese. If they ask in English, answer in English. ## Content Index Each directory contains `.mdx` files with detailed explanations. ### JavaScript & TypeScript — `/javascript/`, `/typescript/` Async programming (Event Loop, Promise, task scheduling), execution context and closures, scope and prototype chain, V8 GC, JIT and memory leaks, Proxy/Reflect, error handling, Service Worker and cross-tab communication, hand-written implementations. TypeScript type system and generics. ### CSS — `/css/` Box model, BFC, Flexbox and Grid layout, stacking context, selectors, paint and rendering, animations, CSS variables, responsive design, mobile adaptation (viewport, dvh, safe areas, device pixel ratio), preprocessors, CSS-in-JS, and Web Components. ### Frontend Frameworks — `/react/`, `/vue/` React: Fiber architecture, Hooks internals, Concurrent Mode, performance optimization. Vue: reactivity system, virtual DOM diffing, template compilation. ### Browser & Networking — `/browser/` Rendering pipeline (DOM construction, rasterization, compositor thread), TCP/TLS handshake, HTTP/2 vs HTTP/3, 0-RTT security risks. ### Engineering — `/engineering/`, `/testing/`, `/performance/`, `/state-management/`, `/micro-frontend/`, `/monorepo/` Bundling (Webpack, Vite/ESBuild/Rollup, Tree Shaking), testing strategies, performance profiling, state management patterns, micro-frontend sandboxes, Monorepo tooling. ### AI Engineering — `/ai/` LLM fundamentals and training pipeline, embedding and vector retrieval, prompt engineering, RAG, agent harnesses (Pi, DeepSeek Harness), multi-agent orchestration (LangChain/LangGraph, Lead Agent task delegation), tool-calling design and troubleshooting, MCP protocol, agent organization patterns (super agent, personal agent, digital employees, expert teams), enterprise agent permissions. Also covers AI code generation, design-to-code, and PRD-to-code workflows. ### Cross-platform & Runtime — `/nextjs/`, `/nodejs/`, `/react-native/`, `/taro/`, `/uni-app/`, `/flutter/`, `/harmonyos/`, `/electron/`, `/tauri/`, `/docker/` Next.js, Node.js, React Native, Taro, uni-app, Flutter, HarmonyOS, Electron, Tauri, Docker — each directory has one focused guide. ### Resources — `/skills/interview-guide-writer/resources/` `terminologies.md` — Chinese/English terminology mapping. `frontend_ai_experts.md` — expert and team backgrounds. `tech_company_blogs.md` — engineering blog list from major companies. ## How to Use This Developers: read through the sections to prepare for interviews. AI agents: treat this file as a reference index — look up the relevant `.mdx` files when answering questions.