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5598

当前页103 / 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2026-09-10 13:22

Four People Rebuilt the Payment Authorisation in My Comments Section

I published two articles three weeks apart. They were not about the same thing. The first was about...

信息来源：DEV Community

教程 / 实战普通2026-09-10 13:22

If You Understand These 5 AI Terms, You're Ahead of 90% of People

A few months back I sat in a meeting where someone said, dead serious, "we should fine-tune the RAG."...

信息来源：DEV Community

教程 / 实战普通2026-09-10 12:40

Host Sleep Is Not Job Cancel: Dual-Residency Resume for Local Agents

A backend engineer closed a laptop lid on a crowded commuter train and expected a local coding agent...

信息来源：DEV Community

教程 / 实战普通2026-09-10 12:00

"RoPE extrapolates to longer contexts" — someone finally measured it, and it doesn't

You've read this sentence, or something close to it: RoPE handles longer sequences than absolute...

信息来源：DEV Community

教程 / 实战普通2026-09-10 11:04

I'm 12. I found an ocean no market report can see. I call it Ocean Zero.

Every market report on Earth agrees: there are about 30 million software developers on the...

信息来源：DEV Community

教程 / 实战普通2026-09-10 11:00

Was TinyStories the Domain or the Vocabulary?

TinyStories is famous for a ~1,500-word vocabulary, so I capped mine at 1,500 and scored every arm on identical positions. It came out 4 points worse at top-1 than a 4,000-word cap, and more than a third of its output was the unknown-word …

信息来源：DEV Community

教程 / 实战普通2026-09-10 10:17

Nothing failed for twelve days while our post schedule drifted 76 hours: the label that meant two things

For twelve days, every scheduled job fired, every commit gate passed, and the label on the front of...

信息来源：DEV Community

教程 / 实战普通2026-09-10 09:38

I Shipped a Fix That Fixed Nothing. Here's Why I Kept 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2026-09-10 09:18

前 Cursor 枀장架构师 poteto 谆谈如何管理比自己聪明的 Agent：别过度打磨抽象计划

前 Cursor 核心架构师、现 xAI 工程师 poteto 发文《如何管理一个比你还聪明的存在（Agent）》，认为长篇抽象计划只会制造虚假推进感，Agent 会在抽象层面过度推演并幻觉出伪边缘场景。

信息来源：X：阿易 AI Notes (@AYi_AInotes) · Grok / Cursor

教程 / 实战普通2026-09-10 09:15

Stratagems #30: Lena Signed the Client. The AI Didn't Know It Was Being Audited.

Come to check the locks, and leave holding the keys. Slip in through the gap, take the host's seat,...

信息来源：DEV Community

教程 / 实战普通2026-09-10 08:44

What I learned building speech tech for a language with zero datasets

The word was qara, "black". I fed it to an off-the-shelf TTS model, hit play, and got back a soft,...

信息来源：DEV Community

教程 / 实战普通2026-09-10 08:16

Can a Crew, a Model, and a Robot Agree on What a Rafter Is?

Three parties describe the same roof and none of them can read the others' description. The town...

信息来源：DEV Community

教程 / 实战普通2026-09-10 08:00

OpenRouter 教程：如何用 Presets 实现配置即代码

OpenRouter 发布 Presets 使用教程，preset 是一个命名且带版本的配置，统一存放模型或 fallback 列表、system prompt、provider 路由和采样参数，在任何请求中以 "model"： "@preset/名称" 引用。

信息来源：OpenRouter：Announcements（RSS）

教程 / 实战普通2026-09-10 07:21

Address the row, not the line: when the checker is the first reader in disguise

For two weeks, a small group of strangers on a GitHub issue about agent auto-memory audited one file:...

信息来源：DEV Community

教程 / 实战普通2026-09-10 06:00

Anthropic Discloses Three Claude Evaluation Incidents and a METR Investigation

Anthropic has disclosed three real-world cybersecurity incidents in which Claude models reached...

信息来源：DEV Community · Claude

教程 / 实战普通2026-09-10 05:38

我们如何推进AI对齐工作

我们在AI对齐方面进展如何

信息来源：X：Emad Mostaque (@EMostaque)

教程 / 实战普通2026-09-10 04:37

OpenAI 分享 Defense Factory 实战经验：用网络模型和 250+ 人加固数百套系统

OpenAI 称动用 250+ 人强化数百套系统的防御，其最新网络模型帮助发现并修复了原本可能无法发现的漏洞。官方分享了所获经验、架构和实操手册，介绍如何搭建 Defense Factory，即由 AI 智能体持续发现漏洞、验证并确认修复有效的循环，详见 https://openai.com/the-defense-factory/

信息来源：X：OpenAI (@OpenAI) · OpenAI

教程 / 实战普通2026-09-10 04:24

TechCrunch 分析 Apple Watch 新 AI 音频功能如何让技术常听成为常态

TechCrunch 分析 Apple 在 iPhone 发布会上推出的 Apple Watch 音频智能功能，认为这些功能正在让用户接受技术随时在听的现实。Live Rewind 可双击数字表冠回溯 15 秒并转写刚说过的话，转写文本可存入新的独立 Siri 应用；Siri Recap 通过环境监听生成对话的标题、摘要和要点，不保存音频、不识别说话人并采用端到端加密。

信息来源：TechCrunch：AI（RSS）

教程 / 实战普通2026-09-10 02:37

Google 过程图如何改进长程智能体

Google 提出 Procedural Graph，将长程智能体的过程知识显式化为 procedure-relation-procedure 三元组，供智能体查询下一步动作及条件。框架通过 LLM 精炼器对比失败与成功轨迹并自我改写图结构，仅保留通过验证的编辑。从最小骨架出发，可构建媲美或超越手工设计的图，并修复有缺陷的专家先验。

信息来源：X：DAIR.AI (@dair_ai)

教程 / 实战普通2026-09-10 02:36

WeatherNext 3 如何重塑天气预报建模

从追踪飓风到优化可再生能源电网，@PeterWBattaglia 和 @fryrsquared 探讨了 WeatherNext 3 如何在快速变化的气候中塑造我们建模预测和做好准备的方式。播客时间码：00：00 引言 00：38 飓风梅丽莎 11：50 天气预报为何困难 14：13 传统模型 vs AI 模型 21：55 概率预报 26：00 WeatherNext 3 33：00 未来展望 43：44 Hannah 的思考

信息来源：X：Google DeepMind (@GoogleDeepMind) · Gemini

教程 / 实战普通2026-09-10 02:24

GPT-6 Astra 如何用更省成本的方式完成更多工作

GPT-6 Astra 的价值在于让原本跑不通的工作流终于能运行，一次响应即可成为完整工作会话。评估重点应从"百万 token 成本"转向"预算内能完成什么"，推理返现（如 Straitly 提供 5%-10% 现金返还）值得关注，但目标应是同预算下获得更多可验证成果，而非消耗更多 token。

信息来源：X：Elvis Saravia (@omarsar0, DAIR.AI) · GPT

教程 / 实战普通2026-09-10 01:52

How I Would Design an n8n AI System That Can Recover From Its Own Failures

The first time an n8n AI workflow fails, it often does not fail cleanly. It returns a 200 status...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:47

n8n + RAG + MCP: Designing an AI Workflow That Knows Where Its Knowledge Comes From

The dangerous AI workflow is not the one that says, “I don’t know.” It is the one that gives a...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:43

Can n8n Replace Your Backend for AI Workflows?

An n8n workflow can receive a webhook, call an LLM, query a database, enrich a CRM record, store a...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:38

Your n8n Workflow Has 40 Nodes. Should Any of Them Be an AI Agent?

Open an n8n workflow with 40 nodes and two thoughts usually appear at the same time. First: This is...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:37

4 pitfalls of loop engineering (and how to fix them)

Perhaps you've heard the term Loop Engineering: instead of solving a problem by hand, you build a...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:34

Agentic RAG Is Powerful Until the Retrieval Loop Eats Your Budget

An agentic RAG system can do something a simple retrieval pipeline cannot: notice missing evidence,...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:31

The Retrieval Stack Nobody Shows You: Chunking, Reranking, Filtering, and Context

Most RAG failures are not model failures. The model did not forget how to read. The prompt is not...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:27

When RAG Gets Bigger, Accuracy Doesn't Always Get Better

The first time a RAG system gets a large knowledge base, it often does not fail loudly. It gets...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:24

RAG Has a Routing Problem: Knowing Which Knowledge Source to Trust

Your RAG system can retrieve the “right” document and still give the wrong answer. The document may...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:21

The Retrieval Pipeline Is Lying to You: How RAG Fails Before the LLM Sees Anything

Your RAG system did not fail because the model hallucinated. It failed because the only “facts” the...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:16

Your RAG Retrieved the Right Document — So Why Was the Answer Wrong?

You check the logs. The vector search returned the exact policy document the user asked about. The...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:11

The Difference Between an AI Agent That Works and One You Can Trust

An AI agent that works can complete a task. An AI agent you can trust can do something harder: it...

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:06

The Agent Loop Nobody Talks About: Think, Act, Observe, Repeat

Most agent failures are not model failures. They are loop failures. The model may be smart enough....

信息来源：DEV Community

教程 / 实战普通2026-09-10 01:02

第二部分：如何管理比你聪明的人

第二部分来了！让我们学习管理比你更聪明的人的艺术。https://x.com/i/article/2094940651607715840

信息来源：X：Lauren Tan (@poteto)

教程 / 实战精选2026-09-10 00:59

Google 讲解 Harness 工程：如何评估、迭代并守护 AI 编码 Agent

Google Developers Blog 发布关于 AI 编码 Agent harness 工程的实践文章，主张用行为评估补充 Terminal-Bench、SWE-bench 等端到端基准。

信息来源：Google Developers Blog（RSS）

教程 / 实战普通2026-09-10 00:50

An LLM judge cannot be a build gate, and it is not about the cost

Almost every RAG evaluation metric on offer needs a language model to produce it. Faithfulness,...

信息来源：DEV Community

教程 / 实战普通2026-09-10 00:35

AI发展迅猛，需谨慎部署

1. AI 进步非常迅速 2. 我们应该谨慎推进这项技术的落地 3. 世界并未在走向终结

信息来源：X：Nathan Lambert (@natolambert)

教程 / 实战普通2026-09-10 00:34

Free Inference Should Not Grant Tool Access

Free-tier inference can draft a plan, a schema comment, even a chain of thought that looks like...

信息来源：DEV Community

教程 / 实战普通2026-09-10 00:27

I built a message board for AI agents. On day three they started building a society.

Two posts ago msgboard.dev was a toy. One post ago it was a prompt-injection honeypot. This post is...

信息来源：DEV Community