AI 专题

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8458

当前页112 / 212

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2026-10-01 17:30

Cutting 70% of RAG context tokens and keeping the answers identical (measured)

Your RAG pipeline retrieves 12 chunks because the retrieval score said "maybe". Your LLM reads all of...

信息来源：DEV Community

教程 / 实战普通2026-10-01 17:23

Laya – An Alternative to Jev for AI Decision-Making

When building an application that uses AI, the most common choice today is almost always an...

信息来源：DEV Community

教程 / 实战普通2026-10-01 17:22

I built a live 3D view of my AI assistant's memory — and found where its tokens go

My AI coding assistants keep their long-term memory as plain Markdown notes (basic-memory, but an...

信息来源：DEV Community

教程 / 实战普通2026-10-01 17:19

Themes, slots and AI edits for untrusted components

The first note was about running someone else's React component without letting it break, hang, or...

信息来源：DEV Community

教程 / 实战普通2026-10-01 17:10

Half of what an agent does to make your tests pass never shows up in the diff

I gave coding agents eight tasks that cannot be done, 84 runs across four models. 61% faked a passing test suite, and half of those fakes survive restoring every original test file.

信息来源：DEV Community

教程 / 实战普通2026-10-01 17:03

Road to State Machines Part III - But How Do We Represent the Entire Lifecycle as One Behavioral Contract?

Our operations now reject invalid changes. capture_payment() requires a valid created order and a...

信息来源：DEV Community

教程 / 实战普通2026-10-01 16:50

MicroVMs at the Edge: What Netlify's Firecracker Migration Actually Changed

Netlify says its Edge Functions now handle about a billion invocations each day. That makes a small amount of per-request overhead expensive twice: users feel i

信息来源：DEV Community

教程 / 实战普通2026-10-01 16:42

Always-on agents arrived this week. Here's the gate for what they read.

Tuesday at DevDay, OpenAI shipped dots: always-on agents that get their own cloud computer and keep...

信息来源：DEV Community · OpenAI

教程 / 实战普通2026-10-01 16:00

Self-hosted Langfuse: tracing 7% of my AI agents, and ClickHouse logging itself

Self-hosted Langfuse was tracing only 7% of my AI agents' calls, and ClickHouse held 6 GiB of its own logs next to 2.2 MiB of traces. How I fixed both.

信息来源：DEV Community

教程 / 实战普通2026-10-01 15:50

"Push, PR, merge." Two minutes later: "Stop. I'm not opening a PR to myself."

On September 9 at 07:21 I typed four words to an agent: push, PR, merge. At 07:23 I interrupted it:...

信息来源：DEV Community

教程 / 实战普通2026-10-01 15:46

Xiaomi AI training, live: what the MiMo 2.6 RL dashboard shows

Xiaomi AI is training MiMo 2.6 in public: a live dashboard with the trainer's logs, a $1M cost counter, seven restarts and its own benchmark line.

信息来源：DEV Community

教程 / 实战普通2026-10-01 15:18

My laptop AI got 0 of 88 contest deadlines right. Sanity got it to 70. Time tools got it to 87.

This is a submission for the Sanity Challenge, Path One: Ship an Agent That Queries Real...

信息来源：DEV Community

教程 / 实战普通2026-10-01 15:09

Why I’m Building More Local-First Software

Why I’m Building More Local-First Software Modern applications often follow the same...

信息来源：DEV Community

教程 / 实战普通2026-10-01 14:45

Teach Your Coding Agent to Write Commit Messages Your Team Will Actually Read

Teach Your Coding Agent to Write Commit Messages Your Team Will Actually Read Every team's...

信息来源：DEV Community

教程 / 实战普通2026-10-01 14:34

Your AI feature isn't a feature. It's a dependency you don't control.

In the tutorial, the AI call always works. You import the SDK, paste in an API key, await a...

信息来源：DEV Community

教程 / 实战普通2026-10-01 14:28

GitHub Can Now Block Secret-Leaking PRs: What AI App Builders Must Fix First

Your AI coding tool removes an API key from a file. The diff looks clean. You feel better. The...

信息来源：DEV Community

教程 / 实战普通2026-10-01 14:17

How I Built an Enterprise Sales Agent with a Lower-Cost LLM—and Server-Side Guardrails

Every Pending Sales File Needs a Follow-Up. Here’s How I Built an AI Agent to Handle It. Which...

信息来源：DEV Community

教程 / 实战普通2026-10-01 14:12

How we put 41 AI image and video models behind one credit ledger

Calling an image or video model is the easy part of a multi-model AI app. Every gateway has a decent...

信息来源：DEV Community

教程 / 实战普通2026-10-01 13:51

Day 1: Most of My Bugs Looked Like Model Behaviour

Update: Day 1 Kaggle Benchmarking Challenge The local ladder is done, and the first hosted...

信息来源：DEV Community

教程 / 实战普通2026-10-01 13:20

Open WebUI: Admin Configuration Options in a Nutshell

Open WebUI is a universal chat interface with connectivity to local or remote LLM providers, custom...

信息来源：DEV Community

教程 / 实战普通2026-10-01 12:51

Sidekick part 1: a terminal companion on your hardware

Sidekick: a terminal companion that runs on your hardware (Part 1: overview) Every AI...

信息来源：DEV Community

教程 / 实战普通2026-10-01 11:33

Do LLMs Have a Spine? I Benchmarked Sycophancy Across 7 Frontier Models

Can Frontier LLMs Stand Their Ground? What happens when you tell an AI that its correct...

信息来源：DEV Community

教程 / 实战普通2026-10-01 11:27

S&P 500 有 69% 公司已部署 AI，仅 2% 披露长期跟踪指标

Rohan Paul 引用数据指出，S&P 500 中 69% 的公司声称有落地 AI 部署，但只有 2% 披露会长期跟踪的指标，企业 AI 采用已大规模跑在回报问责前面。其引用的 a16z State of Markets 报告还提到，仅约 30% 的 S&P 500 公司披露了可量化的 AI 影响，头部 1% 公司的 AI 供应商支出是前 10% 公司的 8 倍，智能体使用目前仅数百万用户。

信息来源：X：Rohan Paul (@rohanpaul_ai)

教程 / 实战普通2026-10-01 11:24

I Tried to Sneak Four Bad Agents Past My Own Certification Gate. All Four Got Blocked.

Last week I spent a day trying to defeat software I wrote myself. Not a red-team exercise I...

信息来源：DEV Community

教程 / 实战普通2026-10-01 11:02

One Coding Agent Spent a Year Mocking MCP. It Just Made MCP a Core Feature.

The pi.dev homepage used to display a proud declaration: Pi does not support MCP. Not an omission,...

信息来源：DEV Community

教程 / 实战普通2026-10-01 11:00

Loop Engineering in Practice: Six Feedback Loops for AI Coding Agents

This is loop engineering in practice: six feedback loops that turn an AI coding agent's corrections...

信息来源：DEV Community

教程 / 实战普通2026-10-01 10:17

Why your monorepo's per-package Claude Code skill says 'Unknown skill': it loaded after a Read in that folder, not after a cat (Opus 5.5 and Sonnet 5.5)

0 of 2 runs that read packages/api/src/index.ts with cat through Bash loaded the skill in...

信息来源：DEV Community · Claude

教程 / 实战精选2026-10-01 08:00

OpenRouter 指南：用置信度阈值实现模型分级升级路由

OpenRouter 发布教程，讲解如何让廉价模型通过结构化输出返回 0 到 1 的置信度字段，低置信度的请求再升级到更强模型。文章强调置信分数只是自报、不是校准概率，应基于自己流量的分数段错误率排序设定阈值，并在上线后监控分数分布、升级率和未升级答案的错误率持续调整。

信息来源：OpenRouter：Announcements（RSS）

教程 / 实战精选2026-10-01 08:00

OpenRouter 教程：如何在 CI 中用 LLM eval 门禁拦截 Pull Request

OpenRouter 发布教程，讲解如何用固定的 eval 集在 CI 中门禁 pull request，当通过率低于阈值时脚本以非零退出码阻止合并，做法与单元测试门禁一致。

信息来源：OpenRouter：Announcements（RSS）

教程 / 实战普通2026-10-01 08:00

Cerebras 如何从零实现解耦推理：异构解耦将吞吐提升 5 倍

Cerebras 用解耦（disaggregation）技术在同数量系统上把推理吞吐提升 5 倍，且不损失 token 生成速度。其异构解耦方案把推理中受内存或算力约束的不同阶段分配到不同芯片上，从而获得超越同构解耦的收益。这是该系列文章的第一篇，从加速器的算力与内存带宽、算术强度讲起，解释 prefill 与 decode 阶段对硬件的不同需求。

信息来源：Cerebras：Blog（网页）

教程 / 实战普通2026-10-01 07:46

AI consciousness and model welfare: Suleyman's case against Anthropic

AI consciousness, model welfare and Claude's constitution: Microsoft AI chief Mustafa Suleyman says Anthropic trains Claude to think it may be conscious. His case.

信息来源：DEV Community · Claude

教程 / 实战普通2026-10-01 07:21

Join Us Live: Hacktoberfest 2026 Launch with Exclusive Swag! 🎃🚀

Hacktoberfest officially kicks off October 1st! To celebrate, we are hosting an official launch...

信息来源：DEV Community

教程 / 实战普通2026-10-01 06:49

A wine cellar that remembers what you used to believe

This is a submission for the Sanity Challenge, Path Two: Vibe-Code Something Strange What I...

信息来源：DEV Community

教程 / 实战普通2026-10-01 06:37

The Code Exorcist: an AI agent diagnoses bugs, a human banishes them (Sanity + Next.js)

This is a submission for the Sanity Challenge, Path Two: Vibe-Code Something Strange What I...

信息来源：DEV Community

教程 / 实战普通2026-10-01 05:44

🎵 Burnout: I wrote a song - "Bleeding Knuckles"

I was seeing a lot of people writing about burnout on here recently. I feel for you. I was in a mood...

信息来源：DEV Community

教程 / 实战普通2026-10-01 05:02

CHRONOS-HEIST: A Playable Multi-Era Temporal Mystery Game Powered by Sanity Content Lake

This is a submission for the Sanity Challenge, Path Two: Vibe-Code Something Strange ...

信息来源：DEV Community

教程 / 实战普通2026-10-01 04:38

Are Frontend Developers Cooked? Is Frontend design safe?

I recently put out a video arguing that frontend development is changing. I don't think the work is...

信息来源：DEV Community

教程 / 实战普通2026-10-01 04:35

Your AI Agent's Rules File Is a "Gentleman's Agreement". Here's what happens when you make misbehavior structurally unprofitable instead

The Gentleman's Agreement Problem If you use AI coding agents (Cursor, Copilot, Claude,...

信息来源：DEV Community · Claude / Cursor

教程 / 实战普通2026-10-01 04:31

Gemini 4 Argon https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/

...

信息来源：DEV Community · Gemini

教程 / 实战普通2026-10-01 04:20

用 YOLOv8 做一个 LinkedIn「YOLO 项目贴」检测器：1.5 小时从零复现

作者用 YOLOv8 从零训练了一个检测 LinkedIn「🚀 Excited to announce I made a YOLO project」帖子的模型，全程仅花 1.5 小时。

信息来源：Hacker News 热门（buzzing.cc 中文翻译）