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5616

当前页68 / 141

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2026-09-22 14:57

Your AI eval is green because it never called the model

We run an eval suite on every PR. Seven fixtures, four metrics, deterministic, no API key, no...

信息来源：DEV Community

教程 / 实战普通2026-09-22 14:18

Mistakes I Made Building My First Full- Stack AI App

A brutally honest reflection of my experience with rate limits, hallucinating prompts, and deployment...

信息来源：DEV Community

教程 / 实战普通2026-09-22 14:14

Apex Bestiary: a living cryptid archive powered by Sanity

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

信息来源：DEV Community

教程 / 实战普通2026-09-22 13:23

Stochastic Parrot or Alien Mind? What Really Is An LLM?

Note: This article is a developer-focused adaptation of my original blog post in Italian. You can...

信息来源：DEV Community

教程 / 实战普通2026-09-22 13:21

Resisting Mode Gravity: Why Bigger LLMs Produce Mediocre Output

Why swapping in a frontier LLM still yields polite boilerplate. Deconstructing Mode Gravity, the triple lock-in flywheel, Differential Diffing, and Model vs. Harness decoupling.

信息来源：DEV Community

教程 / 实战普通2026-09-22 13:08

Designing an eval harness for prompt-injection detection: what measuring my defenses actually taught me

Designing an eval harness for prompt-injection detection: what measuring my defenses...

信息来源：DEV Community

教程 / 实战普通2026-09-22 12:51

I interviewed the people who rejected me. Here's what their hiring process is actually optimizing for.

Last week I wrote about getting rejected from a job for "relying on AI" in the coding round — by an...

信息来源：DEV Community

教程 / 实战普通2026-09-22 12:48

How to Keep AI-Generated Code Maintainable After 6 Months

AI can help you write code incredibly fast. That part is no longer surprising. The harder question...

信息来源：DEV Community

教程 / 实战普通2026-09-22 12:43

Managing the Hidden Overhead of AI Software Engineering

There is a scene that is becoming more and more common in software development right now. You have...

信息来源：DEV Community

教程 / 实战普通2026-09-22 12:25

I built a Magic rules agent, then tried to prove it wasn't guessing

Submission for the DEV x Sanity Challenge, Path One. What I Built JudgeStack answers...

信息来源：DEV Community

教程 / 实战普通2026-09-22 11:54

My agents now have a canvas. They use it to draw diagrams, open browser previews or handoff work each other.

I use Excalidraw a lot at work to draw diagrams, plan projects, and visualize how new features will...

信息来源：DEV Community

教程 / 实战普通2026-09-22 11:32

The 15-minute Claude Code setup, and the four things people get wrong

Four files, fifteen minutes, and five of those minutes are one step. Plus the four claims about Claude Code setup that get repeated everywhere and are wrong.

信息来源：DEV Community · Claude

教程 / 实战普通2026-09-22 11:24

Five ways a voice agent tells you it booked a table when it didn't

The failure is never that the model lies about the transcript. It is that a clerk said something that is not a yes, and the model read it as one.

信息来源：DEV Community

教程 / 实战普通2026-09-22 11:23

Your agent said the tests passed. Check whether they ran.

A step that could not run is easy to report as a step that did not fail. Here is how to tell the difference, and a ten line hook that stops the most common version of it.

信息来源：DEV Community

教程 / 实战普通2026-09-22 10:54

外科医生用 Codex 构建工具并查文献

看看外科医生 @pridgenmd 如何使用 Codex 构建自己的工具，并在 PubMed 上查阅科学文献：

信息来源：X：OpenAI Developers (@OpenAIDevs) · OpenAI

教程 / 实战普通2026-09-22 10:47

Your agent's cost problem isn't the model. It's the steps you never measured.

A few weeks back I watched an agentic pipeline burn a month of budget in three days. The instinct...

信息来源：DEV Community

教程 / 实战普通2026-09-22 10:37

腾讯混元如何把 770B 参数模型压进 214 GiB

腾讯混元量化团队将 Hy4 preview 的权重从约 1.5TB 压缩到 214 GiB，参数量仍保持 770B，仅改变权重表示方式。其 Sherry 算法与 STQ1_0 格式让每组四个权重仅需 5 bit，混合精度模型平均约 2.38 bit/权重。

信息来源：X：腾讯混元 (@TencentHunyuan)

教程 / 实战普通2026-09-22 10:35

The code was never the point: Hacktoberfest, AI, and the value of the climb

I have a small drawer of Hacktoberfest T-shirts. Four of them, I think. Every October for years I...

信息来源：DEV Community

教程 / 实战普通2026-09-22 10:31

Your agent's memory needs a forgetting curve, not a bigger database

Everything your AI agent remembers becomes stale eventually. Here's the engineering behind forgetting: decay curves, core memories, and why we never hard-delete.

信息来源：DEV Community

教程 / 实战普通2026-09-22 10:29

Your Claude Code hook exits 1. It is not blocking anything.

Your Claude Code hook exits 1 when it wants to stop something. It is not stopping anything. That is...

信息来源：DEV Community · Claude

教程 / 实战普通2026-09-22 10:22

Is Kasparov still right?

Kasparov's conclusion: "Weak human + machine + better process was superior to a strong computer alone...

信息来源：DEV Community

教程 / 实战普通2026-09-22 10:17

Six CLAUDE.md files, six codewords: three at launch, two after a Read, one behind an env var

We put a codeword in six CLAUDE.md files and found that the parent-directory, working-directory, and...

信息来源：DEV Community · Claude

教程 / 实战普通2026-09-22 10:14

Jev Decides, OpenAI Reasons: A Practical Hybrid Agent Architecture

Code Controls, Jev Decides, OpenAI Reasons: A Hybrid Agent Architecture That Doesn't Waste LLM...

信息来源：DEV Community · OpenAI

教程 / 实战普通2026-09-22 10:12

Building an Interactive Fashion Hero with HTML, CSS & GSAP

I wanted to experiment with something different from the usual landing pages. So I...

信息来源：DEV Community

教程 / 实战普通2026-09-22 09:35

Beyond Localhost MCP

MCP Is Easy Until You Take It to Enterprise Scale Why local MCP workflows eventually need a gateway,...

信息来源：DEV Community

教程 / 实战普通2026-09-22 09:24

Why 'monitoring' isn't enough for AI agents — and how I made delegation cryptographically verifiable

The problem nobody talks about with AI agents We're rushing to give AI agents autonomy. An...

信息来源：DEV Community

教程 / 实战普通2026-09-22 09:22

How to Search Your Claude Code Conversation History

You talked through an approach with your agent a few days ago. The agent explained the tradeoffs, you...

信息来源：DEV Community · Claude

教程 / 实战普通2026-09-22 09:21

Beyond Bigger Models: Toward a Modular Cognitive Architecture

Beyond Bigger Models: Toward a Modular Cognitive Architecture From Exception-Driven Reasoning to AI...

信息来源：DEV Community

教程 / 实战普通2026-09-22 09:21

Stop Making AI Agents Grind Through Huge Codebases — I Built a Deterministic Wiki Build System for Them

Joining an unfamiliar codebase, what you want isn't more code — it's a wiki that explains how the...

信息来源：DEV Community

教程 / 实战普通2026-09-22 09:10

AIFeed - signed content permissions for AI web crawlers

I built this because robots.txt kept coming up in conversations about AI crawlers and the answer was...

信息来源：DEV Community

教程 / 实战普通2026-09-22 08:57

The Real Fruit Fly Brain Told Me Where I Was Cheating

I replaced my toy neural network with the real mushroom body wiring from a fruit fly. It worked, and immediately showed me what I had been faking.

信息来源：DEV Community

教程 / 实战普通2026-09-22 08:14

小米 MiMo-V2.6-Pro 如何提升新材料研发效率

小米 MiMo-V2.6-Pro 面向新材料研发场景，覆盖从文献综述到分子设计，再到 PFAS 捕获的自动化干实验室实验。该模型旨在提升新材料研发流程中的生产力。

信息来源：小米 MiMo：官网发布与博客

教程 / 实战普通2026-09-22 08:03

MCP Debate: Token Tax, Context Bloat, and What Devs Can Do

A viral post argues MCP causes context bloat and a token tax. Here is how to measure your own tool overhead and cut it.

信息来源：DEV Community

教程 / 实战普通2026-09-22 08:00

英国 AISI 如何用 EvalEval 基础设施让评测结果可复现

英国 AI Security Institute（AISI）正采用 EvalEval 的 Every Eval Ever schema 和 Evaluation Cards 平台公开分享评测结果，以提升评测科学的可复现性。

信息来源：Hugging Face：Blog（RSS） · Hugging Face

教程 / 实战普通2026-09-22 08:00

SpaceXAI 如何用 Grok Bot 扩展客户支持团队

SpaceXAI 发布长文，介绍 Cursor 并入后如何让 Grok Bot 参与客户支持运营。合并团队工单量增长 175% 但未新增人手，否则可能需多招 200 人。

信息来源：xAI：News（网页） · Grok / Cursor

教程 / 实战普通2026-09-22 08:00

StableVQ：稳定向量量化分词器训练的实用指南

StableVQ 通过 Dynamic STE、Region VQ Loss 和 Decoupled Schedule 三项改进，解决向量量化分词器中编码器-解码器与码本训练相互纠缠导致的稳定性问题。该方法基于 shared-projection 码本构建，不引入可学习参数，在 ImageNet 上于不同码本规模和初始化设置下均提升了训练稳定性、码本利用率和重建质量。

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

教程 / 实战精选2026-09-22 08:00

OpenRouter 解读 NVIDIA Nemotron 3.5 Lightning 如何承担 Agent 高频执行调用

OpenRouter 发文解读 NVIDIA 的 Nemotron 3.5 Lightning，这是一款 30B 总参数、约 3B 激活参数的混合专家开源权重模型，定位于工具调用、编码等高频、边界清晰的 Agent 执行步骤，与负责复杂推理的 Nemotron 3 Ultra（550B 总参数、55B 激活）形成分工。

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

教程 / 实战普通2026-09-22 07:32

Claude 生成《荒原》注释指南

AI 可以成为探索远离编程话题的绝佳工具。我让 Claude Fable 5.1 整理了一份艾略特诗作《荒原》的注释指南，包含多条穿行诗歌的路径、录音、学术资料等。我相当 impressed。https://the-waste-land.netlify.app/

信息来源：X：Ethan Mollick (@emollick) · Claude

教程 / 实战普通2026-09-22 07:15

Gemini Notebook Expands Its Workspace, but Artifact Organization Is Still Undocumented

Google has expanded Gemini Notebook, formerly NotebookLM, with capabilities including notebook code...

信息来源：DEV Community · Gemini

教程 / 实战普通2026-09-22 06:47

Museum of Almost Right: change an assumption, change the answer

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

信息来源：DEV Community