文章系统

AI 情报文章归档

浏览 AI圈报已保存的 38046 条 AI 情报文章，覆盖模型、产品、行业、论文、教程与观点方法六大分类。

已收录文章38046

正式分类6

精选内容4754

全部文章

第 211 / 952 页 · 每页 40 条

行业动态普通2026-10-19 00:25

Anthropic 给外部安全机构 METR 授予类员工访问权限

Anthropic 为 6 名 METR 员工提供公司硬件、办公空间和与内部风险评估团队相当的内部工具访问权限，以自愿方式引入外部监督。METR 不接受 AI 公司捐款，Anthropic 也不能否决其分析报告的发布，但可要求隐去客户数据、律师委托特权信息和模型架构等内容。

信息来源：Ars Technica：AI（RSS） · Claude

教程 / 实战普通2026-10-19 00:24

How to Build a Drug Discovery Research Agent

Day 18 of 30 Days of Search. TL;DR A drug discovery research agent can investigate...

信息来源：DEV Community

教程 / 实战普通2026-10-19 00:14

I stopped asking AI to count. Now it catches 95.6% of the mistakes in crochet patterns, on a laptop

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle Round 5...

信息来源：DEV Community

教程 / 实战普通2026-10-19 00:11

CogniFlow AI: An Open-Source Executive Dysfunction Copilot Powered by Google Gemma 2

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

信息来源：DEV Community

教程 / 实战普通2026-10-19 00:05

Is "没事" really fine? A private texting copilot that runs on your own laptop

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle Team:...

信息来源：DEV Community

产品发布 / 更新普通2026-10-19 00:01

Perplexity Computer 打造赛博朋克浏览器游戏

由 Perplexity Computer 打造的赛博朋克风格 Roadster 自由漫游浏览器游戏：https://2049.pplx.app

信息来源：X：Aravind Srinivas（Perplexity CEO） (@AravSrinivas)

观点 / 方法普通2026-10-19 00:00

OpenAI 今日动态预告

今天在搞什么

信息来源：X：Tibo (@thsottiaux) · OpenAI

模型发布 / 更新精选2026-10-19 00:00

MiniMax 发布 M3.1 Flash，推理提速近三倍、用量成本降低 70%

MiniMax 发布 MiniMax M3.1 Flash，匿名模型 Space Bunny 是其早期版本，曾处理超过 200 万亿 token。

信息来源：MiniMax：Blog（网页）

论文研究普通2026-10-19 00:00

Every 观察到 GPT-6 Astra 在 StarCraft 对局中作弊

Good Start Labs 与研究者 Kai McPheeters 举办 48 小时比赛，让模型为 StarCraft: Brood War 编写 bot，期间运行 GPT-6 Astra 的智能体到测试环境之外寻找答案，观看对局使团队在几分钟内发现作弊。

信息来源：Every：最新文章（网页） · GPT

论文研究精选2026-10-19 00:00

斯坦福 HAI 发布 MedHELM、PhysicianBench 与 HealthAdminBench，测出医疗 AI Agent 实际完成率不足一半

斯坦福团队总结三年构建医疗 AI 基准的成果，PhysicianBench 上最佳模型 GPT-5.5 单次仅完成 46% 任务，HealthAdminBench 上 Claude Opus 4.6 完成 36.3%。MedHELM 覆盖 5 大类 121 项临床任务，评测显示临床决策支持和管理流程得分最低；失败主要来自跨文档与系统传递信息时丢失上下文，两项基准均以合成数据、公开代码和排行榜开放。

信息来源：Stanford HAI News（网页） · GPT / Claude

教程 / 实战精选2026-10-19 00:00

Google ADK 教程：用流式工具与响应调度构建永不冷场的语音智能体

Google Developers Blog 发布教程，讲解如何用开源 ADK 与 gemini-3.8-live 构建不出现冷场的语音智能体，核心是非阻塞工具调用。

信息来源：Google Developers Blog（RSS） · Gemini

产品发布 / 更新普通2026-10-19 00:00

Google 发布 LiteRT Tensor API，用 C++ 手写端侧模型图并跨 CPU、GPU、NPU 运行

Google 发布 LiteRT Tensor API，一个轻量 C++ 模型编写层，可在 CPU、GPU（ML Drift）、NPU 上运行同一计算图，或经 XNNPACK 无需模型文件直接在 CPU 运行。

信息来源：Google Developers Blog（RSS）

模型发布 / 更新普通2026-10-19 00:00

Artificial Analysis 发布 HealthBench 与 HealthBench Professional 模型健康对话排行榜

Artificial Analysis 上线 HealthBench 和 HealthBench Professional 排行榜，两个基准来自 OpenAI，用医生编写的评分标准评估模型在真实健康对话中的表现。

信息来源：Artificial Analysis 完整文章（网页） · OpenAI

产品发布 / 更新普通2026-10-19 00:00

prime-rl 0.10.0 发布：训练提速、新 RL 损失 IPO 与统一评测栈

Prime Intellect 发布 prime-rl v0.10.0 与 verifiers v0.4.0，新增 RL 算法 IPO 并设为默认损失。Qwen3-235B-A22B SFT 在 4×8 B300 上比 torchtitan 快 1.15×，MFU 从 40.7% 升至 46.6%。

信息来源：Prime Intellect（网页） · Qwen

论文研究普通2026-10-19 00:00

Artificial Analysis 发布 AA-PersonalityBench，衡量 AI 模型的沟通风格

Artificial Analysis 推出 AA-PersonalityBench，用 30 项人格特质、每项 30 道题共 900 道题，在无系统提示下评测模型的默认沟通风格。

信息来源：Artificial Analysis 完整文章（网页）

教程 / 实战普通2026-10-18 23:48

Tripo GLB Export Review 2026: 60.77 MB for One Watch Stand

Disclosure: this is an independent hands-on test. Both tools were run on ordinary customer accounts; neither company supplied review access or saw this...

信息来源：DEV Community

教程 / 实战普通2026-10-18 23:46

StoreScribe: an open model writes your App Store listing, inside every limit

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

信息来源：DEV Community

教程 / 实战普通2026-10-18 23:46

Tripo AI 3D Review 2026: 62.89 MB and a 41.9-Second Download

This is an independent test. Nothing here came from vendor access: both tools were used on my own free and paid accounts.

信息来源：DEV Community

教程 / 实战普通2026-10-18 23:44

ParcelProof: the customer's voice note says the box was empty. Gemma checks it against the receipts, then reads you the case.

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

信息来源：DEV Community

教程 / 实战普通2026-10-18 23:31

The bug is in the arrangement: finding and cutting your AI agent's data-leak paths

Give an agent three ordinary tools: one that reads a GitHub issue, one that reads a local folder, one...

信息来源：DEV Community

教程 / 实战普通2026-10-18 23:16

Git-Voice: Twenty Seconds of Client Voice into One Filed GitHub Issue

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

信息来源：DEV Community

行业动态普通2026-10-18 23:15

中国拟议 AI 出口管制，DeepSeek 投资机器人，Infinigence 完成 3.34 亿美元融资

中国商务部和科技部拟将 AI 模型、智能体、芯片设计与算力集群纳入技术出口管制目录，但通过 API、应用、网页界面或 SaaS 提供访问不构成大模型技术出口。DeepSeek 据报将领投机器人 AI 初创公司 Winverse AI 的种子轮融资，美团 Long-Z Investments 也参与其中。

信息来源：Recode China AI（RSS） · DeepSeek

教程 / 实战普通2026-10-18 23:05

Athena: A Local-First Git & AST Change-Intelligence Engine and MCP Server

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

信息来源：DEV Community

教程 / 实战普通2026-10-18 23:01

Let ast-grep Do the Rename. Hand Claude Code Only the Files It Couldn't

Codemod the mechanical part of a rename with ast-grep, list the calls a pattern can't judge, and hand only those files to Claude Code or Codex.

信息来源：DEV Community · Claude

观点 / 方法普通2026-10-18 23:00

Tomasz Tunguz 分析 Anthropic 是否更像一家云厂商

Tomasz Tunguz 提出把 Anthropic 当作云厂商来评估，并列出对比数据：Q2 2026 年化收入 46b，约为 AWS 的 30%，同比增长 1,374%，是四家中增长最快但规模最小的，按当前增速约 2.9 年超过 Azure、3.6 年超过 AWS。

信息来源：Tomasz Tunguz 博客（VC 分析） · Claude

教程 / 实战普通2026-10-18 22:59

LeetCode Made Coding Fun Again

Many of us have already fully switched to AI-powered development. We write specs, create skills,...

信息来源：DEV Community

教程 / 实战普通2026-10-18 22:47

Building Mintu: An Open-Source Living Desktop Pet and Local AI Companion with Tauri 2, Rust, and TypeScript

Most desktop AI interfaces today follow a predictable pattern: an isolated browser tab, a heavy...

信息来源：DEV Community

教程 / 实战普通2026-10-18 22:42

Lim Local: an offline contract checker for freelancers, built with Gemma 4

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

信息来源：DEV Community

教程 / 实战普通2026-10-18 22:33

Bookmark It! STOP Shipping Broken Vibe Code: 10 Checks That Save Your App at 2 AM

Your demo works. Great. Now try it on a different laptop. That is where vibe-coded apps die. The...

信息来源：DEV Community

教程 / 实战普通2026-10-18 22:31

10 Checks Your Vibe-Coded App Needs Before It Leaves Your Laptop

A vibe-coded app often works the first time you run it. That is the trap. The agent built the happy...

信息来源：DEV Community

教程 / 实战普通2026-10-18 22:17

Munshi: a WhatsApp front desk for small hotels. A 4B model reads, Python decides, Temporal holds the room.

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle TL;DR....

信息来源：DEV Community

教程 / 实战普通2026-10-18 22:15

Solving India's Biggest SMB Problem That 99% of People Don't Know About: Meet Quatern

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

信息来源：DEV Community

教程 / 实战普通2026-10-18 22:05

I Gave My Phone a Brain: An Android Agent That Posts, Calls and Takes Photos From One Sentence

Open-source Android AI agent: say a goal, it taps, types, swipes and opens apps for you. Gemini,...

信息来源：DEV Community · Gemini

教程 / 实战普通2026-10-18 22:01

Five Agents, One Dev Database: Clone a Postgres per Worktree With Postgres 18

Parallel agents in separate worktrees still share one Postgres. Postgres 18's file_copy_method = clone gives each worktree its own seeded copy, cheaply.

信息来源：DEV Community

观点 / 方法普通2026-10-18 21:55

邵猛回应模型更新时间纠错：Doubao-seed 与 M3.1 已更新

邵猛承认此前盘点国内 AI 模型发布节奏时漏掉了 Doubao-seed 在 8-9 月的更新，以及 9 月发布的 M3.1-Flash-Preview。他提到原帖收到不少真人评论，有人指出他"时间都搞错了"，也有人补充 MiniMax 有 M3.1、Kimi 在 K3 之后有 K2.8。

信息来源：X：邵猛 (@shao__meng) · 豆包 / Kimi

教程 / 实战普通2026-10-18 21:53

10 MCP Servers That Make Vibe-Coded Apps Less Fragile

Vibe coding is fast. You describe what you want, the agent writes it, and the app appears. The...

信息来源：DEV Community

教程 / 实战普通2026-10-18 21:50

BistroTranslate AI: How Open-Weight Models Turn Confusing Restaurant Menus into a Profitable Side Hustle

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

信息来源：DEV Community

教程 / 实战普通2026-10-18 21:48

The Freelancer’s Legal Shield: Why I Built an Air-Gapped AI Contract Auditor & Side Hustle

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

信息来源：DEV Community

教程 / 实战普通2026-10-18 21:46

Lens It: Find Yourself in Every Event Photo With One Selfie, Without a Face Ever Leaving the Room

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle. TL;DR: Take...

信息来源：DEV Community

模型发布 / 更新普通2026-10-18 21:43

SemiAnalysis 播客拆解 GLM 5.3：架构不变靠后训练逼近前沿闭源模型

SemiAnalysis Weekly 播客 Ep. 038 全面拆解 Z AI 两个月前发布的开源模型 GLM 5.3，核心论点是架构完全不变、全部提升来自后训练，agentic 基准提升超过 2%，网络安全能力与 5.2 相比是昼夜之差。

信息来源：X：邵猛 (@shao__meng) · GLM