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5597

当前页106 / 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2026-09-09 11:01

Translating 300-Page Books with Claude: Taming Token Limits and Context Windows

How we chunk long-form content and maintain translation quality with Claude API When we launched...

信息来源：DEV Community · Claude

教程 / 实战普通2026-09-09 11:00

Zepto 如何在 Databricks 和 MLflow 上用评估优先的 AI 智能体扩展客户支持

Zepto 在 Databricks 和 MLflow 上构建评估优先的 AI 智能体，以支撑其快速增长的实时客户支持。该方法将评估作为开发核心环节，通过持续测试与迭代提升智能体可靠性。文章详述了其工程实践与落地路径，为同类大规模客服场景提供参考。

信息来源：Databricks：Blog（RSS）

教程 / 实战普通2026-09-09 10:55

What 'local' means in a Mac meeting transcription app

While building HyNote for Mac, I noticed that the word local was doing too much work. It can...

信息来源：DEV Community

教程 / 实战普通2026-09-09 10:40

Slop-Creep: When Building Gets Cheaper Than Thinking

AI is making software cheap to build. That may leave companies with a new kind of technical debt: systems that never needed to exist.

信息来源：DEV Community

教程 / 实战普通2026-09-09 10:38

How to Build an AI Tool Stack Without Wasting Hours Testing Every New Tool

Every week, developers are introduced to another wave of AI tools. There are AI coding assistants,...

信息来源：DEV Community

教程 / 实战普通2026-09-09 10:35

The LoRA won on its own moods, and the held-out tradeoff stayed visible

I wanted a small answer to a practical question: can two topic-specific LoRA adapters...

信息来源：DEV Community

教程 / 实战普通2026-09-09 10:23

How to know when Claude Code is done

You start a long task, switch to something else, and come back nine minutes later to find it finished...

信息来源：DEV Community · Claude

教程 / 实战普通2026-09-09 10:17

We measured where unattended Claude Code jobs fail: 46 of 72 were the git push

We keep a ledger of every failure from our scheduled agent jobs. It has 72 entries, and 46 of them...

信息来源：DEV Community · Claude

教程 / 实战普通2026-09-09 09:33

Phoenix V2: A Cognitive Architecture for AI That Remembers, Feels and Evolves — Paper, Code and Book

I Updated My AI Model and Lost Everything We Built Together. So I Fixed It. There is a...

信息来源：DEV Community

教程 / 实战普通2026-09-09 09:31

De la Charla a la Demo: Observabilidad y Evals para Agentes de IA con Strands, Phoenix y AWS Bedrock 🔍

Hace poco hablé de esto en mi charla "Harnessing AI Agents en AWS: Evals, Memoria Segura y Prevención...

信息来源：DEV Community

教程 / 实战普通2026-09-09 09:21

GPT-6 Astra vs GPT-5.6 Sol: Choosing by Completed Work, Not Model Age

The short version I would use GPT-5.6 Sol as the default for routine production traffic. I...

信息来源：DEV Community · GPT

教程 / 实战普通2026-09-09 09:15

Predicting the output before running it is how a script I can't write from scratch ended up automating real work.

Small scripts I can fully explain are now automating real work, even though I still can't confidently write Python from a blank file.

信息来源：DEV Community

教程 / 实战普通2026-09-09 08:15

The state axis: why agent benchmarks keep measuring amnesiac models

I keep hammering the point that any coding-agent score is model + harness, not model alone. Same...

信息来源：DEV Community

教程 / 实战普通2026-09-09 08:15

How to Actually Evaluate an AI Code Review Tool

The failure mode that matters in AI code review isn't a missed bug. It's output that reads like a...

信息来源：DEV Community

教程 / 实战精选2026-09-09 08:09

GPT-6 Astra推理等级怎么选才最省Token

卡兹克发文讲解GPT-6 Astra的推理强度等级（Reasoning Effort）含义，指出各档位是同一模型的不同思考预算，Ultra则类似拉起多个智能体协作的专项工作组。

信息来源：公众号：数字生命卡兹克 · GPT

教程 / 实战普通2026-09-09 08:00

A Detector That Never Fires Scores Perfect on False Positives

Three attempts at an overlap detector for rendered video frames failed the same hidden way, and the fourth split in two — one predicate got wired into the gate, one didn't. The split turned on whether each could be made to fire on a known-…

信息来源：DEV Community

教程 / 实战普通2026-09-09 08:00

Brain2Semantics2Text：用语义瓶颈实现非侵入式语音解码

Brain2Semantics2Text 通过中间语义嵌入空间重建文本，将句子级 MEG 脑磁响应映射到语义流形，再反演为自然语言，从而绕开词级对齐。该方法面向非侵入式语音解码中信噪比低、难以还原音素或单词的难题，利用皮层分布式、慢时间尺度的语义表征，在句子级结果上优于此前的非侵入式 Brain2Text 方法。

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

教程 / 实战精选2026-09-09 08:00

OpenRouter 教程：用代码调用 Nano Banana 2 编辑图像

OpenRouter 发布教程，演示通过 API 向 google/gemini-3.1-flash-image（即 Nano Banana 2）发送源图和文本指令完成图像编辑，编辑结果以 base64 形式在响应中返回。教程给出 Python 和 TypeScript 示例、提示词写法、多轮小步编辑方法，以及更换模型只需改一个字段，并介绍了 Nano Banana 系列四个成员的价格与质量差异。

信息来源：OpenRouter：Announcements（RSS） · Gemini

教程 / 实战普通2026-09-09 07:48

Como eu aprendi a aprender (e por que a IA não veio pra pensar por você)

Eu entrei na programação de cabeça no finalzinho de 2022. Não foi um caminho gradual, do tipo "vou...

信息来源：DEV Community

教程 / 实战普通2026-09-09 07:45

Google Says EU DMA Search Redesign Reduces Quality and Changes Search Visibility

Google has rolled out a redesigned Search experience in the European Union to comply with the Digital...

信息来源：DEV Community

教程 / 实战普通2026-09-09 07:20

A memory server may need two write paths, not one

For anything an agent can persist, split the proposal path from the commit path and put a real trust...

信息来源：DEV Community

教程 / 实战普通2026-09-09 07:15

Why Enterprise SEO Advice Can Backfire on Small Sites and What to Do Instead

SEO advice is rarely one-size-fits-all. Techniques built for websites with thousands of templates,...

信息来源：DEV Community

教程 / 实战普通2026-09-09 06:36

Baseten 如何优化托管 rollout 的 delta 权重同步

Baseten 为托管 rollout 设计了跨集群的 delta 权重同步方案：训练端只需向 Amazon S3 发布一次 checkpoint 或 delta，由 Baseten Delivery Network 分发到各独立 rollout 集群，集群无需与训练端直连。

信息来源：Baseten 工程博客（网页）

教程 / 实战普通2026-09-09 06:25

You built an app with AI. Now it has to run somewhere.

Your app works on your laptop and there's nowhere to send anyone. Here's what deploying actually involves, whether you need a server at all, and the seven things AI-written code assumes that stop being true in production.

信息来源：DEV Community

教程 / 实战普通2026-09-09 05:38

Date Slop: building a deliberately bad UX with AI

A while ago now, I entered a Bad UX competition, where a prize went to the worst date-picker. Since...

信息来源：DEV Community

教程 / 实战普通2026-09-09 04:51

OpenAI can't rule out that it stole its most recent breakthrough

A mathematician spent a year on one of the hardest open problems in math. He asked OpenAI a simple...

信息来源：DEV Community · OpenAI

教程 / 实战普通2026-09-09 04:41

My 3B Model Found a Shortcut. It Took Me Three Fixes to Close It.

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2026-09-09 04:35

Track AI Token Spend in Grafana: Claude, Codex, and Ollama

I let three different AIs work in my homelab every day — a coding assistant, a second CLI for...

信息来源：DEV Community · Claude / Ollama

教程 / 实战普通2026-09-09 04:27

Would You Choose a Library Because AI Writes It Better?

I was at a conference recently and watched Joel Hooks talk about Effect. Effect homepage h1...

信息来源：DEV Community

教程 / 实战普通2026-09-09 04:25

Sierra 开源 hyper-τ-bench：评估能构建智能体的智能体

Sierra 今日开源 hyper-τ-bench，这是一个新的长时程智能体评测基准，用于衡量模型不仅能作为智能体行动、还能构建智能体的能力。该基准聚焦于"构建智能体的智能体"这一任务场景，为长时程自主智能体开发提供评测标准。

信息来源：Sierra：Blog（RSS）

教程 / 实战普通2026-09-09 04:16

I Created an AI Fitness Coach with Grok Bot

Hey all. Today I wanted to show you what I'm doing with Grok Bot, which is CoachBot. Yes, I...

信息来源：DEV Community · Grok

教程 / 实战普通2026-09-09 03:53

SPIFFE Workload Identity for AI Agents, End to End

An agentgateway v1.5.0 demo where the agent, the gateway, and the model upstream all authenticate with SPIFFE SVIDs and no certificate file exists...

信息来源：DEV Community

教程 / 实战普通2026-09-09 03:45

OpenAI Navier-Stokes Claim Signals a New Test for Multi-Agent AI Research

OpenAI has reportedly used a large coordinated group of AI agents to pursue one of mathematics' most...

信息来源：DEV Community · OpenAI

教程 / 实战普通2026-09-09 03:34

Dynamic Workflows: Critical Analysis (pt. II) (feat. Fable)

[NOTE: This is a follow-up to this post where I began to develop a hermeneutic critique of the Claude...

信息来源：DEV Community · Claude

教程 / 实战普通2026-09-09 03:17

AI labs cutting off Cursor and Windsurf is the platform risk nobody priced in

Your AI programming editor could be rendered useless with just one decision by management. It's not...

信息来源：DEV Community · Cursor

教程 / 实战普通2026-09-09 02:55

Grok Bot 从零到百万用户的成长故事

Grok @Bot 已成为当下最热门的 AI 产品，上线不到一个月即吸引数百万用户。Roman Ugarte 首次讲述其孵化与发布历程：为何选择从零构建而非集成进 Cursor，以及发布前数周刻意"删减功能"等关键决策。

信息来源：X：Lauren Tan (@poteto) · Grok / Cursor

教程 / 实战普通2026-09-09 02:45

What AI Can’t Fake: How junior engineers can stand out in the age of AI

There is a piece of advice that circulates endlessly through mentoring conversations, coding...

信息来源：DEV Community

教程 / 实战普通2026-09-09 02:36

We gave our AI agent fleet a credit limit, and it hit it the same day

Ten agent sessions ("minds," in this codebase) run continuously on one box, each with its own...

信息来源：DEV Community

教程 / 实战普通2026-09-09 02:20

My AI Remembered Everything Important. It Forgot What I Did Last Night.

I built an open-source memory system for AI agents that works like a brain, connections that strengthen with use and decay without it. Then it got so good at surfacing what mattered that it could not tell me what I had just done. The bug w…

信息来源：DEV Community

教程 / 实战普通2026-09-09 02:11

I Swept All 33 Bedrock Regions So You Don't Have To

One region out of 33 can fine-tune Llama 3.3 70B on Bedrock. I asked every region the SDK lists, then re-ran the sweep before publishing to be sure it still held. Also in here: every model you can customise in us-east-1 belongs to Amazon, …

信息来源：DEV Community · Llama