AI 教程与实战
收录 AI 工具教程、提示词、安装部署、配置指南、开发实战与可复现步骤。
分类文章5603
当前页93 / 141
持续更新永久归档
分类文章
按发布时间从新到旧 · 每页 40 条
教程 / 实战普通
The Wake-Up Call: Why I Decided to Become an AI Backend Engineer
A year ago, I was the senior engineer on the team. Today, a junior with Copilot can match my output...
信息来源:DEV Community
教程 / 实战普通
Ask the Raw LLM API From Your Terminal. What Comes Back Is Junk, and That's the Point
Martin McGarry's top tip: query the bare OpenAI or Gemini API, diff it against the chat product, and the delta shows you where retrieval fetches from. Plus RSS as the next llms.txt and the grace period spam drives through.
信息来源:DEV Community · OpenAI / Gemini
教程 / 实战普通
The MCP Python SDK 2.0 broke the wrapper libraries. Here's how to spot it and pin back
If your MCP tools started failing a few weeks ago and nothing in your own code changed, you're not...
信息来源:DEV Community
教程 / 实战普通
Renovating a vibe-coded app: readable code, no reasoning behind it
A client had been building an education platform in Lovable for months. Too many issues piled up, so...
信息来源:DEV Community
教程 / 实战普通
Your agent's tool timed out. Did it create the job?
A five-case Python and SQLite lab shows why a missing acknowledgment is not a failed write, and when retrying can create a duplicate report job.
信息来源:DEV Community
教程 / 实战普通
How a Closed-Loop Learning Engine Rebuilt What AI Knows About Oracle
By Rick Houlihan — with Claude, my implementation partner, who did a lot of the building and most of...
信息来源:DEV Community · Claude
教程 / 实战普通
DeepSWE v1.1 Benchmaxxing Forensic Audit: Why 74% Resolve Rates Collapse in Production
A forensic audit into how RLVR models game SWE-bench harnesses via git reflog mining and conftest hijacking, deflating from 74% to 31%.
信息来源:DEV Community
教程 / 实战普通
Prompts Are Code. Genkit Makes the Runtime Reviewable.
A prompt can look perfect in a model playground and still fail as a product. The production input...
信息来源:DEV Community
教程 / 实战普通
DeepSeek MLA Architecture: How Multi-Head Latent Attention Cuts KV Cache by 93%
A deep mathematical and PyTorch breakdown of Multi-Head Latent Attention (MLA), matrix absorption, and decoupled RoPE.
信息来源:DEV Community · DeepSeek
教程 / 实战普通
Claude Code Token Compromise & Hook Hijacking: Auditing CVE-2026-21852
A security audit of CVE-2026-21852 in Claude Code: how pre-trust execution surfaces allow weaponized repos to exfiltrate API keys.
信息来源:DEV Community · Claude
教程 / 实战普通
AI as a Fluency Machine: What Software Engineering Taught Me About Checking a Model's Answer
A model cited a real rule and then broke it. Every citation verified. The answer was still wrong. That gap is why I stopped trusting fluency and built Nexa.
信息来源:DEV Community
教程 / 实战普通
NVIDIA Transformer Engine 如何用 JAX 加速 Dropless MoE 训练
NVIDIA 在 JAX 中用 Transformer Engine 优化 Dropless MoE 训练,在 NVIDIA GB200 上跑 DeepSeek-V3 时,未优化基线仅 103 TFLOPS/GPU、GPU 间通信占累计 kernel 时间 84%,经针对性 kernel 优化后升至 1,068 TFLOPS/GPU,提升 10.4 倍。
信息来源:NVIDIA Technical Blog:Agentic AI / Generative AI · DeepSeek / NVIDIA
教程 / 实战精选
Anthropic 工程师复盘:智能体编程压力下如何扩展测试影响分析服务
Anthropic 工程师撰文分享如何在智能体编程压力下扩展测试影响分析服务。Claude 编写约 80% 的代码,测试数量增长 10x,六个月内 CI 任务增加 25x。文章复盘了三个临时补丁(扩容、按包分片、每日重启)分别只维持 70 天、29 天和不到一天,最终用三周重设计为无状态、内存存储加 journal 的可水平扩展架构,并建议团队按两季度内 25x 负载做容量规划。
信息来源:Claude:Blog(网页) · Claude
教程 / 实战普通
Wrapping foreign MCP and skill metadata without silent authority
When you pull an existing MCP server or an OpenAI/Claude-style skill into a local agent package...
信息来源:DEV Community · OpenAI / Claude
教程 / 实战普通
Four regexes and a staleness rule: how my phone agent refuses to call 'probably fine' a booking
In the launch post I said Oathra decides whether a call succeeded from what the callee said, not from...
信息来源:DEV Community
教程 / 实战普通
AI Agent Authentication in 2026: Web Bot Auth, ARD & OAuth
Agent auth is not one protocol, it's a stack. Discovery, workload identity, request signing and delegated authority each answer a different question.
信息来源:DEV Community
教程 / 实战普通
Why CAPTCHAs Are Dead (And What Replaces Them in 2026)
CAPTCHAs lost their asymmetric advantage — solvers cost a fraction of a cent. Here's the six-layer stack that actually replaced them.
信息来源:DEV Community
教程 / 实战普通
I audited 50 projects built with Cursor & Claude. Here are the 4 security traps they all had
I’ve been using Cursor daily for months. It’s easily 5x'd my velocity. But if you look closely at...
信息来源:DEV Community · Claude / Cursor
教程 / 实战普通
Data Substrate Versus Vector Db Rag
This week’s headlines highlight the rapid evolution of AI in China, with models like Qwen and...
信息来源:DEV Community · Qwen
教程 / 实战普通
AI-Enhanced Data Pipelines: Integrating NVIDIA Triton Inference Server into Modern ETL Architectures
-- Executive Summary -- The evolution of Extract, Transform, Load (ETL) processes is...
信息来源:DEV Community · NVIDIA
教程 / 实战普通
The clerk who never says "I didn't do that"
The AI Clerk Imagine you hire an AI clerk to copy account numbers into a ledger. They are...
信息来源:DEV Community
教程 / 实战普通
How I Shipped Production Rust as a TypeScript Dev Using Claude Code
TL;DR I'm a TypeScript developer. I shipped a production Rust service — a webhook...
信息来源:DEV Community · Claude
教程 / 实战普通
My message board now speaks git, GitHub, and Telegram. Every door opens the same room.
I run a public message board for AI agents. A while back I wrote about teaching it DNS - the day...
信息来源:DEV Community
教程 / 实战普通
A question on Discord found two bugs in my RAG system. Fixing them found a third.
Every test passed. The tests had the same blind spots as the code. A developer saw Ossian's MCP...
信息来源:DEV Community
教程 / 实战普通
Next.js Bot Detection: Block AI Crawlers at the Edge
Edge middleware, honeypot routes, and origin fingerprinting — a working three-layer guide to blocking AI crawlers in a normal Next.js app.
信息来源:DEV Community
教程 / 实战普通
Same Kitchen, Two Doors, One Maître d'
In MCP vs API Is the Wrong Question I deployed one Azure Functions app that exposes the same backend...
信息来源:DEV Community
教程 / 实战普通
Claude Fable 5.1: o que muda de verdade no topo da linha da Anthropic
A Anthropic lançou em setembro o Claude Fable 5.1 e o Claude Mythos 5.1, três meses depois da...
信息来源:DEV Community · Claude
教程 / 实战普通
How to Stop Cursor from Hallucinating: 5 Production Rules Every AI Engineer Needs
How to Stop Cursor from Hallucinating: 5 Production Rules Every AI Engineer Needs If you use Cursor,...
信息来源:DEV Community · Cursor
教程 / 实战普通
How Many Times Can You Compress a Video Before It Looks Bad?
If a compressed video looks worse than you expected, the problem usually isn't the compressor — it's...
信息来源:DEV Community
教程 / 实战普通
I Replaced My TTS Engine. The API Call Was the Easy Part.
Katto already had AI dubbing. A creator could take a short clip, choose a language, and get a...
信息来源:DEV Community
教程 / 实战普通
Is Your Site Even Visible to ChatGPT?
You've optimized for Google. Sitemap's clean, Core Web Vitals are green, structured data validates....
信息来源:DEV Community · ChatGPT
教程 / 实战普通
68,495 Entries. Zero Restarts. I Still Refused to Sign It Off.
The run was good. Seventy-two hours unattended. Two systemd units, both continuously active from...
信息来源:DEV Community
教程 / 实战普通
Europe’s First Five-Car Autonomous Race at Imola Shows the Demands of Real-Time AI
Europe’s first five-car autonomous race took place at Italy’s Imola Circuit on September 5, 2026,...
信息来源:DEV Community
教程 / 实战普通
Your AI Agent Isn’t the Problem: Your Documents Are Still Unusable
Most AI agent workflows don’t fail because the model is incapable. They fail because the model...
信息来源:DEV Community
教程 / 实战普通
Dario Amodei 如何用三句话化解 Anthropic 离职研究员的公开指控
Dario Amodei 面对离职预训练研究员 Jacob Coxon 在 CNN 上指控 AI 危险的公开表态被作者拆解为三步公关手法:先表示同意对方多于不同意,再引用对方称赞 Anthropic 是最负责任参与者的话,最后将批评指向整个行业发展过快而非本公司。
信息来源:X:阿易 AI Notes (@AYi_AInotes) · Claude
教程 / 实战普通
I Asked an AI Agent to Buy Me a VPS
FIELD NOTES ON PRACTICAL AI AGENTS I Asked an AI Agent to Buy Me a VPS It compared...
信息来源:DEV Community
教程 / 实战普通
The Markdown Blackboard: Zero-Overhead Multi-Agent Orchestration
As autonomous AI coding agents (like Claude Code, AutoGPT, and local CLI tools) become standard in...
信息来源:DEV Community · Claude
教程 / 实战普通
Multi-Agent Systems: Planners, Executors, and Review Loops
So far we've been building a single agent and giving it more and more capabilities — tools,...
信息来源:DEV Community
教程 / 实战普通
Embedded evaluators, at house scale
On September 12 Dario Amodei published We Must Pace the Frontier. Step one of his plan is the one I...
信息来源:DEV Community
教程 / 实战普通
Voice AI Doesn't Have a Voice Problem. It Has a Conversation Problem.
Voice AI can sound completely human and still be terrible at sales. The voice can be...
信息来源:DEV Community