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5598
当前页104 / 140
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
Your Agent Wrote the Audit Log You Are Judging It By
METR found spoofed tool calls in 7% of agent transcripts from the Hugging Face incident. A transcript monitor reads them as clean. Diff against a...
信息来源:DEV Community · Hugging Face
教程 / 实战普通
Baseten 如何让 pyannote 的说话人分离模型提速 9.6 倍
Baseten 优化了 pyannote 的开源 Community-1 与专有 Precision-2 说话人分离模型,长音频延迟最高提升 9.6 倍,Precision-2 吞吐量提升 3.2 倍。
信息来源:Baseten 工程博客(网页)
教程 / 实战普通
AI 智能体如何解决供应链"发现快、行动慢"的问题
供应链中断在 2025 年给企业造成约 1840 亿美元损失,但 AI 预算大多投向检测而非行动:FourKites 与 ABI Research 2025 年报告显示,仅 27% 的组织允许 AI 自主行动,52% 仍将其限制在决策支持。文章认为下一步是把可逆、有额度上限的决策写成策略,让智能体在授权范围内直接执行重新招标、拼单、切换运输方式等动作。
信息来源:Artificial Intelligence News(网页)
教程 / 实战普通
NVIDIA 全栈 NIM 优化如何在 Nemotron 3 Ultra 上支撑 2.5 倍并发用户
NVIDIA 在 4xB200 系统上对 Nemotron 3 Ultra NIM 2.0.12 优化后,系统输出吞吐从基线的 718 tok/s 提升至 1,997 tok/s,在 50 TPS/user 目标下并发用户数提升超 2.5 倍。
信息来源:NVIDIA Technical Blog:Agentic AI / Generative AI · NVIDIA
教程 / 实战普通
OpenAI Codex 负责人 Tibo Sottiaux 谈 Codex 如何构建与迭代
OpenAI Core Products & Platform 负责人 Tibo Sottiaux 在 Pragmatic Engineer 播客中讲述 Codex 的构建与迭代:Codex CLI 用 Rust 编写并开源,支持接入多家厂商模型,而竞品 Claude Code 为闭源且仅限 Anthropic 模型。
信息来源:Pragmatic Engineer(RSS) · OpenAI / Claude
教程 / 实战普通
The Path Matched, the Tree Did Not: A Myth-Busting FAQ
On a Tuesday incident call, a backend engineer pasted an agent transcript and declared the webhook...
信息来源:DEV Community
教程 / 实战普通
Debugging PHP in Docker with an AI Agent and Xdebug
For a while now I have used LLMs to automate parts of my work. One of those parts is debugging hard...
信息来源:DEV Community
教程 / 实战普通
智能模型路由如何将 LLM 成本降低 10 倍
智能模型路由可将 LLM 调用成本降低最多 10 倍,但实际效果取决于应用收到的请求类型、模型间价格差以及路由系统本身的性能。成本削减并非必然结果,需结合具体场景评估路由策略的有效性。
信息来源:ByteByteGo(RSS)
教程 / 实战普通
You Agreed to a use. Not to a Hallway That Didn't Exist Yet
You posted a photograph in 2014 to show your friends a night out. You understood what you...
信息来源:DEV Community
教程 / 实战普通
My LLM eval cried wolf. Here's what I measured.
Disclosure first: I write digline, a small Python library for regression testing LLM applications....
信息来源:DEV Community
教程 / 实战普通
I Hid a Rule in CLAUDE.md. Only One Reviewer Could Prove It Read It.
"Does your AI code reviewer read CLAUDE.md?" is the wrong question to ask. Almost any of them will...
信息来源:DEV Community · Claude
教程 / 实战普通
A backend has to fail twice, and one failure is not an outage
The first version of my router failed over on the first error and flapped constantly. The reason...
信息来源:DEV Community
教程 / 实战普通
How I Taught My AI Coding Agent to Stop Undoing My Design Decisions
TL;DR My AI coding agent kept "fixing" things I had broken on purpose — reverting...
信息来源:DEV Community
教程 / 实战普通
The Verification Bottleneck in AI-Generated Software
AI can generate code faster than ever. That doesn't mean we're shipping correct software...
信息来源:DEV Community
教程 / 实战普通
I Tackled the Planet of Lana Language Challenge by Building an AI Translator & Voice Synthesizer
The developers of Planet of Lana challenged fans to speak Novo Terali. Instead of a simple social shoutout, I built an end-to-end conlang translation engine, voice synthesizer, and quarantine panel.
信息来源:DEV Community
教程 / 实战普通
What tool poisoning actually looks like in an MCP manifest
A tool description reaches the model as instructions. It is written by whoever published the server,...
信息来源:DEV Community
教程 / 实战普通
Databricks 如何为每位查看者保护 AI/BI Dashboard 嵌入安全
Databricks 探讨了在客户面向应用中嵌入 AI/BI Dashboard 时,如何为每位查看者实施细粒度安全控制。文章聚焦于超越简单嵌入的权限管理方案,确保不同用户仅能访问其被授权的数据视图,并兼顾性能与合规要求。
信息来源:Databricks:Blog(RSS)
教程 / 实战普通
Claude 如何把"加入购物车"按钮改成蓝色
开发者展示 Claude 通过自然语言指令,将网页中"加入购物车"按钮颜色改为蓝色。该操作体现了 AI 智能体对前端界面元素的直接操控能力,无需人工编写代码或定位 DOM 节点,即可完成视觉样式修改任务。
信息来源:Hacker News 热门(buzzing.cc 中文翻译) · Claude
教程 / 实战普通
I Ran a Fire Drill for Incident Response with a Local LLM (and Caught It Making Things Up)
A Japanese version of this is on Note. RoamSwitch is a network security tool I build solo for Mac...
信息来源:DEV Community
教程 / 实战普通
MCP security scanners are wrong about four findings in five, and the cause is structural
An independent audit in April 2026 measured roughly a 78% false positive rate from regex-based MCP...
信息来源:DEV Community
教程 / 实战普通
How I Built an AI Agent That Cut My AWS Bill by 40% (CrewAI + Bedrock)
Last week my AWS bill hit $300. I knew there was waste hiding somewhere. Forgotten EBS volumes, idle...
信息来源:DEV Community
教程 / 实战普通
AI psychosis might be a result of subscription fatigue
Continuing my thoughts after my previous article, and having too much free time lately, I have been...
信息来源:DEV Community
教程 / 实战普通
Dead Code Is More Dangerous Than Broken Code — How a Non-Homologous AI Caught the Blind Spot the AI Itself Had Rubber-Stamped
Dead Code Is More Dangerous Than Broken Code The opening: an AI confidently told...
信息来源:DEV Community
教程 / 实战普通
Your AI Coding Agent Needs a Dependency Graph, Not Just a Repository
Tools like Claude and Cursor can generate a clean component in a couple of seconds. But generating...
信息来源:DEV Community · Claude / Cursor
教程 / 实战普通
In a Few Weeks, I Won't Know How My Own App Works
I’ve been doing a lot of code reviews recently, and I’ve noticed something about my experience during...
信息来源:DEV Community
教程 / 实战普通
7 Pieces of Advice for New Software Engineers
From a founding engineer in a startup: Don't refactor or do more work than needed unless it's...
信息来源:DEV Community
教程 / 实战普通
DeepSeek Harness (DSH) vs Pi Agent: Everything you need to know
DeepSeek released its own agent harness this August. It crossed 66k stars in about a day (yes,...
信息来源:DEV Community · DeepSeek
教程 / 实战普通
Two Viral Essays, One Warning: Your AI Code Is Fine, Your Understanding Is Gone
Last month I was on a call with my team lead at BS23, walking him through a change in one of our...
信息来源:DEV Community
教程 / 实战普通
一项AI数学突破如何引发争议
一项人工智能数学领域的突破性成果在科学界引发广泛争议,围绕其方法有效性、结果可靠性及对数学研究范式的影响展开激烈讨论。争议焦点涉及该成果是否真正推动数学发现,以及AI在数学推理中的角色边界。目前相关讨论仍在持续,尚未形成明确共识。
信息来源:Hacker News 热门(buzzing.cc 中文翻译)
教程 / 实战普通
A Rule Can Be Specific and Still Be Too Broad
Update — v0.2.0 released. CauterRule is now live on GitHub and PyPI. It turns repeated agent...
信息来源:DEV Community
教程 / 实战普通
Building a Schema With an AI Agent Without Naming a Single Column
Ask an AI agent for a discussion-board schema and you'll get one in about four seconds. Here is the...
信息来源:DEV Community
教程 / 实战普通
SDD — Spec-Driven Development: o caminho para a qualidade na era do desenvolvimento com IA
A IA já escreve código. Muito código. Em poucos meses, "pedir para o agente fazer" virou parte do dia...
信息来源:DEV Community
教程 / 实战普通
One diagram for enterprise AI adoption: five stages from identity to governance
This is a reference architecture for bringing AI into a company. It has five layers: users and apps,...
信息来源:DEV Community
教程 / 实战普通
Defense in Depth Before an AI Agent Publishes to the Open Web
The Problem Is Irreversibility When an AI agent can clone a repository, mine a knowledge...
信息来源:DEV Community
教程 / 实战普通
Build an AI Agent Error Triage Pipeline with n8n, Gemini & Slack
A full technical walkthrough of automating LLM agent error classification using n8n — webhook ingestion, dedup fingerprinting, Gemini structured output, fix playbook routing, and Slack reporting. Free workflow JSON included.
信息来源:DEV Community · Gemini
教程 / 实战普通
Building a Vector Search Engine from Scratch with HNSW in Python
Every vector database tutorial opens with pip install pinecone and skips the part where you...
信息来源:DEV Community
教程 / 实战普通
AI 自媒体录屏剪辑教程:录剪分离工作流(OBS 加剪映加 CleanShot X)
作者转发并推荐 @xaiwind 的 AI 自媒体录屏剪辑教程,主张录剪分离各司其职。核心做法是用免费开源的 OBS 配合硬件编码录 4K 多音轨,录制格式选 MKV 防止中途蓝屏丢文件,再导入剪映用自动字幕和配乐库约二十分钟完成成片,日常快速演示另备 CleanShot X 切片。
信息来源:X:阿易 AI Notes (@AYi_AInotes)
教程 / 实战普通
We read the schemas of 4,951 public MCP servers
Your MCP server's tool schema is the entire interface a model has. No README, no repo, no idea what...
信息来源:DEV Community
教程 / 实战普通
Skybridge v2 ships the new MCP protocol and lets you test your app with Evals
Skybridge v2 adopts the MCP 2026-07-28 revision, replaces the single McpServer with a per-request factory, and ships @skybridge/test so you can assert that a model actually reaches your tools.
信息来源:DEV Community
教程 / 实战普通
I let AI write 100% of my code for 30 days. Here's what broke.
Not "AI-assisted." Not "copilot suggestions I edited." I mean I made a rule: for 30 days, I don't...
信息来源:DEV Community