AI 情报文章归档
浏览 AI圈报 已保存的 37920 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37920
正式分类6
精选内容4747
全部文章
第 30 / 948 页 · 每页 40 条
教程 / 实战普通
Claude Code reads AGENTS.md now, but only when CLAUDE.md is missing
Claude Code shipped AGENTS.md support on September 18, 2026. The changelog line for 2.1.277 is one...
信息来源:DEV Community · Claude
论文研究普通
Identifiable alignment of unpaired representations
Aligning two models without paired data requires a distance that both share, yet existing methods use Euclidean or spherical geometry. This is either an arbitrary choice or imposed by their loss function. We show that, under a precise form…
信息来源:arXiv
教程 / 实战普通
Building a free AI UI builder on Cloudflare: what small models taught me
A 2 days ago I launched openconstructor, a free, open-source browser editor: you describe an...
信息来源:DEV Community
教程 / 实战普通
An email told my AI support agent to refund $900. Here's the guardrail that said no.
I'm building PurseString, an open-source MCP server that lets AI agents spend money safely. I'm in...
信息来源:DEV Community
教程 / 实战普通
Auto-Research for Physical Products
The idea started with a conversation about math. A lot of recent progress in AI-assisted math comes...
信息来源:DEV Community
论文研究普通
MOMENT-DP: Momentum Aware Evaluation of Correlated Noise for Differentially Private Training
Temporally correlated noise can substantially improve the utility of differentially private optimization, yet candidate mechanisms are commonly designed or evaluated using only partial information about the optimization process, such as wo…
信息来源:arXiv
论文研究普通
Semantically Aligned KV Cache Compression for Efficient Agentic Autoregressive Video Generation
Agentic autoregressive video generation combines semantic planning, sequential generation, and visual verification for controllable synthesis. Growing visual histories enlarge the key-value (KV) cache, increasing memory use and attention c…
信息来源:arXiv
论文研究普通
How Much Training-Inference Mismatch Can On-Policy Distillation Tolerate?
On-policy distillation (OPD) trains a student language model on its own responses, guided by a fixed teacher. Generating these responses is expensive, so systems use quantized inference and asynchronous rollouts to reduce the cost. Both, h…
信息来源:arXiv
论文研究普通
Learning to Rank Propagators in Constraint Programming
The order in which propagators are executed can substantially affect the performance of constraint programming solvers, yet propagator priorities are typically determined by fixed, manually designed heuristics. In this work, we explore lea…
信息来源:arXiv
教程 / 实战普通
The 11 CMS SEO Features Every Platform Should Include in 2026
A CMS can make routine SEO either a manageable editorial task or a continuing queue of developer...
信息来源:DEV Community
论文研究普通
SWHForest: A Structural Curation and Search Layer for Large-Scale Code Corpora
Pre-training code language models relies on massive code corpora usually streamed over the internet. However, researchers rarely need a multi-terabyte archive in its entirety; instead, their workflow centers on curated subsets: selecting w…
信息来源:arXiv
论文研究普通
How to "train" a surgeon? Benchmarking generalist agents in surgical scene understanding
Evaluating general-purpose multimodal large language models (MLLMs) for surgical scene understanding remains limited due to scarce surgical data and fragmented evaluation protocols. We address these challenges through a two-stage evaluatio…
信息来源:arXiv
行业动态普通
Chamath:算力涨价或迫使 SpaceX 重定价 Anthropic 合约
Chamath 在 All-In Podcast 上称,若数据中心稀缺把算力价格推高 2 到 3 倍,SpaceX 与 Anthropic 的算力合约定价将由表格决定,而非两家公司关系。
信息来源:X:Rohan Paul (@rohanpaul_ai) · Claude
教程 / 实战普通
Why We Built a Financial Firewall for Autonomous AI Agents on Solana: ACID Reservations, Zero Key Leakage, and the Subscriptions Program
Why We Built a Financial Firewall for Autonomous AI Agents on Solana: ACID Reservations,...
信息来源:DEV Community
行业动态普通
蚂蚁集团灵光用户突破 2000 万,灵光 2.0 上线一周管理超 116 万项事务
蚂蚁集团旗下 AI 助手灵光宣布用户突破 2000 万,灵光 2.0 自 10 月 27 日上线以来一周内帮助用户管理超 116 万项事务,执行超 360 万次任务。
信息来源:IT之家·人工智能
教程 / 实战普通
I audited my own tool: what limitcast got wrong and what I fixed
I read limitcast the way I would read a stranger's tool: corrupt its state, feed it bad input, time...
信息来源:DEV Community
教程 / 实战普通
eBay API Setup for an AI Listing Agent
Published November 07, 2026 on matbanik.info The AI agent that lists my eBay items can group my...
信息来源:DEV Community
教程 / 实战普通
I published 30 runnable API examples — here is what the real outputs taught me
A green run status is not enough to trust an API. I want to inspect the dataset, understand the...
信息来源:DEV Community
教程 / 实战普通
PhantomCommit AI: Bringing Forgotten Code Back to Life with Open-Source AI
What I Built Code never truly dies. It just waits to be understood. 👻 Every developer has...
信息来源:DEV Community
教程 / 实战普通
Andrej Karpathy 用 Claude 周末搭建视频精选平台 HackerTube
Andrej Karpathy 用 Claude Opus 5.5 在一个周末搭建了 HackerTube,从 20 年 Hacker News 的 4300 万条评论中筛出 6454 个视频,并称观看体验明显优于 YouTube 自身的推荐。
信息来源:The Decoder:AI News(RSS) · Claude
观点 / 方法精选
GPT-Image-2.6 实测:10 个 GIF 玩法提示词与三项新能力评测
作者实测前两天上线的 GPT-Image-2.6,整理出 10 个 GIF 玩法提示词(黏土定格动画、表情包、剪辑贴纸、loading 图、PPT 动态图表、爆炸拆解图等)并测试三项新能力。
信息来源:X:卡尔的 AI 沃茨 (@aiwarts) · GPT
教程 / 实战普通
OpenAI Says ChatGPT’s Search-Like Queries Are About One-Fifth of Google’s
OpenAI has offered a striking new comparison for the scale of AI-assisted discovery: ChatGPT’s...
信息来源:DEV Community · OpenAI / ChatGPT
教程 / 实战普通
The Answer Looked Fine. The AI Bill Didn’t.
A Friday release adds a friendly touch to a support chatbot. Greet customers by loyalty tier. Give a...
信息来源:DEV Community
观点 / 方法普通
用 n8n 和 TurboDocx 从 Slack 审批到 SharePoint 归档发送合同签署
n8n 发布 Verified Node Spotlight,演示用 TurboDocx 节点补齐合同电子签名环节:Excel 中标记 Ready to send 的行自动生成合同 PDF,经 Slack 审批后发送签署,客户签完后已签 PDF 归档到 SharePoint 并回写表格状态。发送流程每 15 分钟运行一次,TurboDocx 免费版每月含 5 次文档生成和 5 次签名。
信息来源:n8n Blog
行业动态普通
FT 报道 Anthropic 以 34 亿美元购入 1.7GW Caterpillar 燃气发电机
据 FT 报道,Anthropic 以 34 亿美元购入 1.7GW 的 Caterpillar 燃气发电机,是已知 AI 实验室最大规模的电力设备采购。这些订单原属圣安东尼奥供应商 OnePWR Solutions,两年前下单但因融资失败被 Anthropic 接手;由于下单早于涨价,设备将在未来两年内交付。
信息来源:X:Rohan Paul (@rohanpaul_ai) · Claude
教程 / 实战普通
Why Multi-Pet AI Art Is an Engineering Nightmare (And How We Solved It)
Anyone who has owned more than one pet knows the family-photo problem: someone is always mid-blink,...
信息来源:DEV Community
论文研究普通
Meta 论文介绍大规模 AI 辅助 A/B 实验的 Agent 设计与经验教训
Meta 发布论文《Learnings and Pitfalls from AI-Assisted A/B Experimentation at Scale》,介绍其 ABTest Agent:LLM 只负责选择和排序实验步骤,所有数字都由经过验证的确定性工具计算,报告模板要求每张表格可追溯到代码,知识库仅收录专家审核过的文档。
信息来源:X:Rohan Paul (@rohanpaul_ai)
教程 / 实战普通
Worker Previews Give Every Change a Place to Prove Itself
TL;DR Cloudflare launched Worker Previews on September 22, 2026: every Git branch gets a...
信息来源:DEV Community
教程 / 实战普通
I run 16 CLI commands, 47 content checks, and 27 AI crawler tests. Here's which ones actually matter
GEO Optimizer runs 16 CLI commands, 47 content checks, and 27 AI crawler tests per audit. After...
信息来源:DEV Community
教程 / 实战普通
My free open-source AI UI builder got 562 visitors in its first 2 days — here's what worked
Two days ago I launched openconstructor (https://openconstructor.pw), a free, open-source browser...
信息来源:DEV Community
教程 / 实战普通
Text-to-SQL em português com RAG e LoRA: o que aprendi medindo um assistente sobre dados sintéticos
Montei um assistente que transforma perguntas em português sobre uma frota de ônibus em SQL, e passei...
信息来源:DEV Community
观点 / 方法普通
Anil Seth、Ned Block 与 Tononi 在 Nature 撰文称当今 AI 不具备意识且硅基硬件难以产生意识
Anil Seth、Ned Block 和 Giulio Tononi 在 Nature 发表评论,认为今天的 AI 没有意识,在硅基硬件上大概率也不可能有意识,并反对把正确的计算等同于意识。
信息来源:The Decoder:AI News(RSS)
教程 / 实战普通
Reproducible LLM Evals: Record the Run, Not Just the Score
Your eval suite scored 87% on Tuesday and 84% on Thursday. Nobody changed the prompt. Was it the...
信息来源:DEV Community
教程 / 实战普通
Your AI Agent Is a Confused Deputy: What Developers Need to Know About AI Security
Prompt injection, hallucinated packages and AI-run intrusions, plus the design rules that hold up when the model gets fooled.
信息来源:DEV Community
教程 / 实战普通
[AI Hands-on] Refactoring LINE Business Card Assistant: Python to Go with Antigravity Remote Agent, Google ADK, and Firestore -
Background Recently, I've been rewriting several of my daily core services from Python...
信息来源:DEV Community
论文研究普通
上海交大发布 AI 原生科学实验室 AutoDNA,成果登《自然》
上海交大 IPADS 实验室联合化学化工学院团队在《自然》发表成果,构建 AI 原生科学实验室 AutoDNA(思源一号),实现自主分子生物学实验操作。系统采用多智能体架构分解实验流程,把仪器能力抽象为可调用的原子服务,硬件支持微米级定位和微升级液体处理,集成基于 OpenHarmony 的控制终端,已在核酸检测、酶促 DNA 合成、癌症检测等场景完成应用验证。
信息来源:IT之家·人工智能
教程 / 实战普通
How I made "jump to message" work in ChatGPT, Claude and Gemini
I'm a third-year student, and a lot of my studying and project work happens in AI chats. The chats...
信息来源:DEV Community · ChatGPT / Claude / Gemini
教程 / 实战普通
I built a "what am I signing?" checker for Solana in a weekend
Most people approve crypto transactions blind. A site says "Claim your airdrop!", the wallet pops up,...
信息来源:DEV Community
教程 / 实战普通
'Almost Right' Agent Code Is Worse Than Wrong Code. It's Devs' Top AI Complaint
Stack Overflow 2025: 66% of devs flagged AI code that's 'almost right, but not quite', the #1 frustration. Wrong code fails loudly. Almost-right ships.
信息来源:DEV Community
行业动态普通
Asahi Linux 因 AI 洗代码放弃 MIT,全部仓库转向 GPL v2 或更新版本
Asahi Linux 宣布将所有仓库迁移到 GNU GPL v2 或更新版本,未来仅使用强 copyleft 许可证。原因是有开发者从其个人 fork 提取未公开代码,或用 AI 重写以掩盖来源,用于替代该项目并商业化 Apple Silicon Linux。
信息来源:Hacker News:AI 热帖