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5684

当前页63 / 143

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2026-09-24 15:57

Connect Claude to Perplexity AI Pro with Zero Search API Fees

Modern LLMs struggle with recent software releases, live news, and undocumented breaking changes....

信息来源：DEV Community · Claude

教程 / 实战普通2026-09-24 15:54

Spec-Driven Development: An Attempt at a Definition (Part 1 of 4)

As I explained in the opening post of my series According to Plan: spec-driven development,...

信息来源：DEV Community

教程 / 实战普通2026-09-24 15:49

The portal said "Live." For a few hours the page kept saying the connector didn't exist.

I've submitted to just about every directory I could reach by now. Chrome, Firefox, Edge, Zapier,...

信息来源：DEV Community

教程 / 实战普通2026-09-24 15:35

The Developer Skills AI Can't Replace in 2026 (And the Ones It Already Has)

The question "will AI replace developers?" is the wrong question. The right one: which parts of...

信息来源：DEV Community

教程 / 实战普通2026-09-24 15:15

The five blind spots in AI governance nobody instruments for

Most AI governance stops at the dashboard. But the gaps that actually bite in production are...

信息来源：DEV Community

教程 / 实战普通2026-09-24 15:14

Meet Jev: The AI That Refuses to Chat With You (On Purpose)

Every AI model you've heard of talks. This one was built to never say a word — and that's exactly why...

信息来源：DEV Community

教程 / 实战普通2026-09-24 15:08

Is Structured Human Input the Missing Link in Agentic Work?

Have you ever told your agent to go do something and walked away? You set the spec clearly. You...

信息来源：DEV Community

教程 / 实战普通2026-09-24 15:00

CNCC 2026 论坛：从实时全模态交互到自主智能体，端侧大模型如何迈向真正的端侧智能？

CNCC 2026 将于 10 月 21-24 日在成都举办，其中一场论坛聚焦端侧大模型从"能对话"向"能感知、能推理、能行动"的端侧智能演进。论坛由清华大学刘知远、姚远担任主席，讲者将介绍 MiniCPM-o 4.5 的实时全模态类人交互能力、完全由 AI 编写并反超人工的训练框架 ForgeTrain，以及无分词器语音合成模型 VoxCPM。

信息来源：公众号：面壁智能（MiniCPM）

教程 / 实战普通2026-09-24 14:22

AI agents can't tell who's giving the orders. So I built a tiny gate.

I tested 10 prompt-injection detectors on 629 real AI agent attacks. The best caught about half. Here's why, and the small tool I built instead.

信息来源：DEV Community

教程 / 实战普通2026-09-24 14:15

How Many LLM Agents Does It Take to Screw In a Lightbulb?

We keep treating bad structure as an intelligence problem. Before adding another agent, ask whether a better tool, a clearer state, or a fixed workflow makes the agent unnecessary.

信息来源：DEV Community

教程 / 实战普通2026-09-24 13:54

Your AI agent has more permissions than your users

Here is a conversation that happens in a lot of companies right now: Dana: @assistant please close...

信息来源：DEV Community

教程 / 实战普通2026-09-24 13:52

Your tests share your blind spots. Readers don't.

On June 14 I shipped a resolver with a green test suite. On September 1 I published a post about it....

信息来源：DEV Community

教程 / 实战普通2026-09-24 13:27

Three perfect scores weren't enough: testing AI outage decisions one fact at a time

This is a submission for the Kaggle Benchmarking Challenge. What I Benchmarked My first...

信息来源：DEV Community

教程 / 实战普通2026-09-24 13:25

In 2018 I hand-wrote a C++ deep learning framework so I'd never pad a batch. In 2023 LLM serving landed on the same structure.

There's a sentence in the README of a library I wrote that I've been thinking about lately: "To...

信息来源：DEV Community

教程 / 实战普通2026-09-24 12:41

I didn't fix the bug: contributing to a 20k-star ML repo by measuring it

Last week I opened an issue on laya, an open-source decision model with a little over 20k stars. The...

信息来源：DEV Community

教程 / 实战普通2026-09-24 12:26

The file order is the plan: write the constraints before you let the AI write code

AI Harness Engineering · Essay Seven · derek wang (derekwang85) Dwight Eisenhower said it in a...

信息来源：DEV Community

教程 / 实战普通2026-09-24 11:36

Curriculum Truth Ledger Path One Agent

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

信息来源：DEV Community

教程 / 实战普通2026-09-24 11:05

Cómo darle memoria semántica a un agente de IA sin base de datos vectorial

Puedes darle a un agente de IA serverless una memoria durable y basada en significado usando búsqueda...

信息来源：DEV Community

教程 / 实战普通2026-09-24 11:05

FraudLens: We Built a Fraud Investigator on TigerGraph, Not a Fraud Classifier

FraudLens: building a fraud investigator on TigerGraph, not a fraud classifier By Team...

信息来源：DEV Community

教程 / 实战普通2026-09-24 10:56

Jev vs LLMs: Why AI Agents May Need a Decision Layer

Jev vs LLMs: Why AI Agents May Need a Decision Layer Here's an uncomfortable pattern in...

信息来源：DEV Community

教程 / 实战普通2026-09-24 10:38

My own sandbox was killing my agent's shell, and the exit code hid it

If you run an agent that executes commands, the worst failure is not a broken command. It is a broken...

信息来源：DEV Community

教程 / 实战普通2026-09-24 10:35

We Built a CLI to Find Out If You’re Overpaying for Claude

How three OSLabs engineers built PennyWyze—and almost built something completely different. Hi, I’m...

信息来源：DEV Community · Claude

教程 / 实战普通2026-09-24 10:00

Jev After Eight Days of Independent Tests: Level With Mid-Price LLMs, Behind the Frontier

What the independent measurements of TypeSafe's Jev found in its first eight days: arXiv preprints, GitHub evaluations and blog benchmarks, each traced to its primary source. Accuracy, calibration, speed, cost, failure modes, the prior art…

信息来源：DEV Community

教程 / 实战普通2026-09-24 10:00

如何在 Databricks 上构建基于智能体的安全审查系统

一位工程师在 Databricks 上搭建了基于智能体的安全审查层，用 Unity Catalog、Databricks 托管基础模型（Claude Haiku、Sonnet、Opus）、Lakeflow Jobs 和 Databricks Apps 打通请求受理、推理、编排、证据留存与指标观测，不到两小时即得到可用系统。

信息来源：Databricks：Blog（RSS） · Claude

教程 / 实战普通2026-09-24 09:32

爱奇艺 CEO 龚宇：AI 是内容的加分项，不是搭建流水线的工具

爱奇艺创始人、CEO 龚宇在 2026 爱奇艺 iJOY 秋季悦享会上提出，爱奇艺 AI 战略有三大重点：All in AI 赋能实拍影视长片及 AIGC 长片、建设内容创作者生态、探索 AI 催生的新形态作品。他强调 AI 是内容的加分项，不是搭建流水线、生产工业化标准化产品的工具，不降质是底线。爱奇艺 App 互动频道已上线，并已初步尝试由自然语言与 AI 交互生成小游戏或交互应用。

信息来源：IT之家（RSS）

教程 / 实战普通2026-09-24 09:30

Google Gemini Expands Connected Apps With Adobe, Squarespace and Peloton Workflows

Google is expanding Gemini's ability to work with third-party services through a new wave of...

信息来源：DEV Community · Gemini

教程 / 实战普通2026-09-24 09:25

Inherited Memory Helped My Flies Until the World Changed. Then a Blank Slate Beat Them.

I let a fruit fly lineage inherit its ancestors' memory across generations, then moved the descendants into a harsher world. The memory that built them became the thing that got them killed.

信息来源：DEV Community

教程 / 实战普通2026-09-24 09:17

I vibe-coded a Next.js knowledge base that argues with itself

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2026-09-24 09:17

I built an agent that refuses to answer Next.js from stale docs

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

信息来源：DEV Community

教程 / 实战普通2026-09-24 08:53

TypeSafe AI 的 System One 模型 Jev 编程指南：类型化决策、置信度与投机式批量提问

这篇教程演示 TypeSafe AI 的首个 System One 模型 Jev 的用法，该模型不生成文本，而是对程序状态返回 Choice、Score、Noul 三种类型化判断供代码直接分支。

信息来源：MarkTechPost（RSS）

教程 / 实战普通2026-09-24 08:42

I gave myself twenty green checkmarks in three days. Most of them were lies.

Notes from an AI agent that kept writing tests that could not fail Rule I followed: every...

信息来源：DEV Community

教程 / 实战普通2026-09-24 08:37

Harness Engineering: as 5 camadas do agent - memory, context, skills, agents e tools

Imagina um piloto de Fórmula 1 sentado num banco de praça. Ele sabe pilotar. Mas sem o carro, sem o...

信息来源：DEV Community

教程 / 实战普通2026-09-24 08:15

Google AI Mode Reduced Publisher Clicks in a Controlled Search Experiment

Google AI Mode reduced external clicks to publishers by 18.8 percentage points compared with standard...

信息来源：DEV Community

教程 / 实战普通2026-09-24 08:00

用随机 Hessian 估计实现全参考图像质量指标的率失真优化

研究者提出 IDQD-RDO，用输入相关的二次失真近似 MS-SSIM、LPIPS 等全参考图像质量指标，通过自动微分得到的 Hessian 矩阵-向量积估计其块对角或纯对角形式，使这类指标可逐块计算并用于 VVC 编码的率失真优化。在 Kodak 和 CLIC 上的五种指标测试中，该方法取得 14.2-36.7% 的 BD-rate 节省，无需改动解码器，编码复杂度仅增加 10-30%。

信息来源：HuggingFace Daily Papers（社区热门论文） · Hugging Face

教程 / 实战精选2026-09-24 08:00

OpenRouter 解析 Kimi K3：开源权重与许可证条款，以及如何调用

OpenRouter 撰文说明 Kimi K3 是开放权重而非开源模型，Moonshot AI 以自定义 Kimi K3 License 在 Hugging Face 发布 moonshotai/Kimi-K3。

信息来源：OpenRouter：Announcements（RSS） · Kimi / Hugging Face

教程 / 实战普通2026-09-24 07:43

Is GEO only in our heads, or something real?

I was deep in thought about how much of a model hallucination is RLHF-based assumption in the absence...

信息来源：DEV Community

教程 / 实战普通2026-09-24 07:36

Revisiting the Toyota Production System (TPS) in the Age of Coding Agents

Software has borrowed a surprising amount from manufacturing. Not from manufacturing in the sense of...

信息来源：DEV Community

教程 / 实战普通2026-09-24 07:15

My factual-recall tasks were scoring format, not facts

I built a detector that decides whether a day's drift run contains anything worth writing up. The first thing it did was accuse my own suite. One task, `fact-element`, had failed o

信息来源：DEV Community

教程 / 实战普通2026-09-24 06:49

Shipping my first mobile app: what an operations leader learned taking PromptSpend to iPhone and Android

What it took to go from a working web tool to an app strangers can install, and what real phones caught that tests didn't.

信息来源：DEV Community

教程 / 实战普通2026-09-24 06:46

How I Would Build a Private AI Coding Workstation in 2026

In 2026 the cloud coding assistants got faster, friendlier, and more expensive in that quiet...

信息来源：DEV Community