#AI编程
9349 related articles
AI Agent in Practice: From Prompts to Few-Shot for Building Enterprise-Grade Agents
From prompts to Few-shot, break down the practical path for deploying enterprise AI Agents. Through a real Sony store sales scenario, explain how examples change large model outputs and why the core skill of an AI Agent builder is business organization, not programming.
LangChain + LangGraph + MCP in Practice: Building Enterprise-Grade AI Agents from Scratch
A deep dive into LangChain, LangGraph, MCP, and enterprise-grade Agent practice—from low-code platforms to precise token control, covering AI engineering thinking and three Agent-building approaches.
Harness Architecture Explained: Context Engineering, Task Orchestration, and the Skills Lifecycle
The Harness architecture is becoming a frequent interview topic for AI large model positions. This article explains context engineering, task orchestration, and the Skills lifecycle, outlining the capability requirements and transition paths for engineering implementation.
DeepSeek Harness in the Enterprise: From Agent Fundamentals to Project Development
Bilibili's Teacher Dayu explains DeepSeek Harness enterprise deployment—from Agent fundamentals and native model limitations to hands-on projects covering automated development, enterprise customer service, and multimodal RAG, ideal for beginners exploring AI engineering paths.
Claude Code in Practice: Installation, Configuration, and Efficient Development Tips
A detailed guide to Claude Code installation, configuration, and practical tips: NPM setup, bypassing login with CC Switch to avoid bans, permission modes and built-in commands, MCP, Skills, Hooks, SubAgent usage, context management, and prompt optimization.
A Google Engineer of 12 Years: CS Curricula Should Be Burned Down—In the AI Era, Move Beyond Code
Former Google engineer Titus Winters on the AI era: CS curricula should be rebuilt, deterministic tools rather than AI should refactor code, why productivity metrics miss AI's real value, and the shift in software engineering education.
LangChain + RAG Interview Prep: Understanding LangChain's Core Value in One Article
Based on the Bilibili LangChain + RAG interview prep tutorial, this article explains LangChain's role as a large model application development framework, detailing the three limitations it addresses—limited knowledge, lack of external interaction, and insufficient state retention—and offers official documentation learning paths and interview preparation advice.
Deep Agent Sub-Agent Explained: An Enterprise AI Agent Framework in the LangChain Ecosystem
An in-depth look at Deep Agent: the LangChain team's framework for long-horizon tasks, covering the sub-agent mechanism, its place in the LangChain/LangGraph ecosystem, and enterprise use cases.
ChatGPT Codex in Practice: A Zero-Basics Guide from Configuration to AI Programming
A hands-on ChatGPT Codex guide for beginners: the core differences between Codex and Claude Code, what AI agent tools really are, and how to tackle the biggest hurdle—setting up a GPT account.
Dify + RAG in Practice: Building a Game Assistant with Self-Hosted Deployment
A complete Dify self-hosted deployment guide based on a Bilibili tutorial: comparing with Coze's cloud hosting, covering WSL and Docker setup, image pulling, Dify startup, and building a RAG game assistant with knowledge base and workflow.
Claude Code in Practice: Harness Engineering for the Full AI Development Workflow
Bilibili instructor Zhuge Laoshi's Claude Code course explained: use Harness engineering-based programming to complete the full AI development workflow, from requirement breakdown to code verification, and apply it to enterprise e-commerce projects with domestic models such as GLM to boost interview competitiveness.
Claude Code Step-by-Step Guide: Installation, Configuration, and Permission Modes Explained
A step-by-step Claude Code guide covering NPM installation, using ccswitch to avoid login bans, strategies for the four permission modes, and prompt writing tips for stable use in China.
Open-Source DeepSeek Skill Pack: Approaching Claude Opus with 29 Skills
DeepSeek Stack is an MIT open-source skill pack with 29 skills. Through plan loops, verify loops, and multi-model review, it helps the far cheaper DeepSeek V4.1 Flash approach Opus 5.5 at 80% on coding tasks. Here's how to install it and the full skill list.
Vercel Security Review Public Beta: Automatic Vulnerability Scanning on Every Push
Vercel Security Review is in public beta: once enabled for Pro and Enterprise teams, every GitHub push automatically reviews for vulnerabilities, with results in the Vulnerabilities dashboard, one-click fixes via Vercel Agent, and natural-language security policies.
Vercel CLI Adds doctor Command: One-Click Configuration Diagnostics
Vercel CLI launches the vercel doctor command to check CLI version, login status, project links, and configuration, with fix suggestions. Supports --json structured output, ideal for non-interactive coding agents.
Revolut Ditches Claude for Self-Hosted AI Coding, Slashing Costs 10x
UK fintech Revolut cut AI coding costs 10x by dropping Claude for a self-hosted solution. This analysis examines the cost logic, feasibility boundaries, and market signals.
LangChain Tool Returning 200 Means It's Safe? Detecting API Schema Drift
Explore how to detect HTTP 200 schema drift when LangChain tools call third-party APIs, including bounded sampling, known-good window comparison, optional field handling, and where to place checks.
Recreating Meta Muse Charm with a $40 ESP32: A DIY AI Pendant Hands-On
Recreating Meta Muse Charm with a $40 Waveshare ESP32S3 and Claude: voice chat, smart home control, virtual pet — all hand-written and isolated on a guest network.
Vercel CI Decouples from Projects: Direct GitHub Repository Connection
Vercel CI now runs on private GitHub repositories without linking a Vercel project. Learn the prerequisites, setup steps, and impact on development workflows.
Weekend Hack: Filtering 43 Million HN Comments with Claude to Build a Clean YouTube Recommendation Site
A developer used Claude Code to analyze 43 million Hacker News comments from two decades, curating 6,454 community-recommended YouTube videos into HackerTube, a login-free recommendation site — all in about two hours.