AI Daily

2026年03月28日 · 48h 动态

🔥 38 主题 📊 140 信源 📡 17 渠道
🔥 主题列表 ×
AI Qwen Meetup Korea: Function Calling 从 6.75% 提升到 100%
AI Sigil: 零知识隐写术保护 AI 训练数据
AI AI 投资军备竞赛：SoftBank $400 亿、SK hynix IPO、VC 巨额募资
开发工具 GitHub Trending: Claude Code 生态项目爆发
开发工具 AI Agent 开发工具井喷：Git blame、TUI 管理、网关、安全框架
SEO Google 2026 年 3 月核心更新 + AI/Bot 标签新规
AI Google Gemini 支持导入其他 AI 的记忆与聊天记录
AI AI 推理能力研究前沿
AI AI Agent Benchmark 论文潮：从金融到医疗到游戏
AI Google Lyria 3 Pro 与 Gemini 3.1 Flash Live 发布
AI OpenAI 关停 Sora 视频工具并搁置色情聊天功能
AI Apple WWDC 2026 预告 AI 进展，Siri 将开放第三方 AI
AI $500 GPU 在编程基准上超越 Claude Sonnet
SEO Google Web Guide 与 SEO 新格局
AI ByteDance Seedance 2.0 视频生成：发布即暂停
开发工具 Storybook MCP 支持 React AI Agent 集成
AI AI Agent 安全警钟：自我审批、危险命令、权限漏洞
SaaS AI SDR/CSM 实战经验：部署、踩坑、效果评估
AI 开源语音模型：Cohere 转录 + Mistral 语音生成
SEO SEO 行业巨变：从 SEO 到 Agentic AI Optimization
AI Meta AI 全面发力：购物、创业支持、Ray-Ban 新款
AI Anthropic Claude Code 获得更安全的自动模式与电脑控制能力
AI EU 支持裸照 App 禁令并延迟 AI 规则实施
AI AI 创业融资热潮：Harvey $110 亿、Granola $15 亿、Mirage $7500 万
AI MonkePay: 用 USDC 向 AI Agent 按请求收费
AI Nvidia CEO 宣称已实现 AGI
SaaS SaaS 2026 估值暴跌：软件首次低于标普 500
AI OpenAI 安全举措：Bug Bounty、Model Spec、青少年保护
AI Figma 开放 AI Agent 直接在画布上设计
基础设施 Arm 发布首款 CPU，将接入 Meta AI 数据中心
AI Together AI 发布 Mamba-3 + Perplexity Health 接入 Apple Health
基础设施 Hypura: Apple Silicon 存储层感知 LLM 推理调度器
电商 ChatGPT vs Gemini：争夺 AI 购物入口
AI Vibe Coding 成为主流开发方式
AI Cursor Composer 2 基于 Moonshot Kimi K2.5 构建
AI GitHub Trending: AI Agent 框架与开源项目
AI AI 交易代理开源项目火爆
AI Simon Willison: slop 定义与 AI 监控警告
📊 信源列表 ×
T4 24 HN Show
T3 22 TechCrunch AI
T3 15 Search Engine Journal
T2 14 GitHub Trending
T2 12 AlternativeTo
T2 8 HuggingFace Papers
T3 7 The Verge AI
T4 7 HackerNews
T3 6 SaaStr Blog
T1 5 OpenAI Blog
T2 5 Product Hunt AI
4 Bluesky @simonwillison.net
T2 4 ArXiv cs.CL
T2 3 ArXiv cs.AI
T3 2 Ahrefs Blog
1 Bluesky @jdp.extropian.net
T1 1 DeepMind Blog
📡 渠道列表 ×
T1 官方
T1 DeepMind Blog
T1 OpenAI Blog
T2 平台
T2 AlternativeTo
T2 ArXiv cs.AI
T2 ArXiv cs.CL
T2 GitHub Trending
T2 HuggingFace Papers
T2 Product Hunt AI
T3 媒体
T3 Ahrefs Blog
T3 SaaStr Blog
T3 Search Engine Journal
T3 TechCrunch AI
T3 The Verge AI
T4 社区
T4 HN Show
T4 HackerNews
其他
其他 Bluesky @jdp.extropian.net
其他 Bluesky @simonwillison.net
分类
类型
时间昨天 03/27前天 03/26
排序
AI实战03-28 12:44
Qwen Meetup Korea: Function Calling 从 6.75% 提升到 100%
在 Qwen Meetup Korea 上展示的 Function Calling Harness 将成功率从 6.75% 提升至 100%
💡 开源模型工具调用能力飞跃，差距不在模型而在工程
HNHN ShowShow HN: Qwen Meetup Presentation, Function Calling Harness, 6.75% to 100%
AI产品发布03-28 11:17
Sigil: 零知识隐写术保护 AI 训练数据
本地优先的桌面应用 Sigil 使用零知识隐写术保护数字资产不被 AI 爬虫采集
💡 反 AI 数据采集工具出现，数据保护从法律手段转向技术手段
HNHN ShowShow HN: Sigil – A zero-knowledge steganography vault for AI training data
AI融资03-28 03:11
AI 投资军备竞赛：SoftBank $400 亿、SK hynix IPO、VC 巨额募资
SoftBank 获 $400 亿贷款指向 OpenAI IPO，SK hynix 或赴美 IPO 募 $100-140 亿，Kleiner Perkins 募 $35 亿 All-in AI
💡 AI 基础设施军备竞赛进入万亿美元级别，芯片供应链成为战略焦点
TCTechCrunch AIMemory chip giant SK hynix could help end ‘RAMmageddon’ with blockbuster US IPO TCTechCrunch AIAir Street becomes one of the largest solo VCs in Europe with $232M fund TCTechCrunch AIWith $3.5B in fresh capital, Kleiner Perkins is going all in on AI TCTechCrunch AIData centers get ready — the Senate wants to see your power bills
开发工具开源03-27 22:35
GitHub Trending: Claude Code 生态项目爆发
Claude Code 相关开源项目霸榜：everything-claude-code 日增 4,453 星、oh-my-claudecode 576 星、awesome-claude-code 413 星
💡 Claude Code 正在形成类似 VS Code 的插件生态，开发者社区自发构建工具链
HNHackerNewsAnatomy of the .claude/ Folder HNHackerNews90% of Claude-linked output going to GitHub repos w <2 stars GitHubGitHub Trendinghesreallyhim/awesome-claude-code (413 stars today) GitHubGitHub Trendingaffaan-m/everything-claude-code (4,453 stars today) GitHubGitHub TrendingYeachan-Heo/oh-my-claudecode (576 stars today) GitHubGitHub Trendingletta-ai/claude-subconscious (71 stars today)
开发工具产品发布03-27 22:26
AI Agent 开发工具井喷：Git blame、TUI 管理、网关、安全框架
HN 出现大量 AI Agent 基础设施项目：Origin 追踪 AI 写的代码、Kage 用 tmux 管理多 Agent、LLM-Gateway 零信任网关等
💡 AI Agent 从概念验证进入工程化阶段，开发者工具链正在快速补齐
HNHN ShowShow HN: Origin – Git blame for AI agents (track which AI wrote every line) HNHN ShowShow HN: Kage – TUI for managing multiple AI agents with tmux and Git worktrees HNHN ShowShow HN: LLM-Gateway – Zero-Trust LLM Gateway HNHN ShowShow HN: Oo – compress output for coding agents (cargo test → "47 passed, 2.1s") HNHN ShowShow HN: SentinelGate – Access control for AI agents (open-source MCP proxy) HNHN ShowShow HN: MyClawn – Claude Code networks for you on autopilot HNHN ShowShow HN: Clinic – Connect your agent to Gmail, GitHub, Notion and 39 more (OSS) HNHN ShowShow HN: Context Plugins – API context for AI coding assistants HNHN ShowShow HN: Bottrace – headless CLI debugger for Python, built for LLM agents HNHN ShowShow HN: ClawRun – Deploy AI agents to secure sandboxes with one command HNHN ShowShow HN: Aegis – Security framework for AI agents HNHN ShowShow HN: Agent Kernel – Three Markdown files that make any AI agent stateful HNHN ShowShow HN: Wit – Stops merge conflicts when multiple AI agents edit the same repo HNHN ShowShow HN: Nimbus – Turn how you work into how your agent works
SEO市场03-27 20:30
Google 2026 年 3 月核心更新 + AI/Bot 标签新规
Google 开始推出 March 2026 Core Update，同时为论坛和问答结构化数据添加 AI 与 Bot 标签
💡 Google 首次要求结构化标注 AI 生成内容，透明度成为排名信号
SEJSearch Engine JournalGoogle Begins Rolling Out March 2026 Core Update via @sejournal, @MattGSouthern SEJSearch Engine JournalGoogle Adds AI & Bot Labels To Forum, Q&A Structured Data via @sejournal, @MattG SEJSearch Engine JournalGoogle Tests AI Headlines, Rolls Out Spam Update – SEO Pulse via @sejournal, @Ma SEJSearch Engine JournalWikipedia Bans Use Of AI-Generated Content via @sejournal, @martinibuster AhrefsAhrefs BlogIs AI Content Bad for SEO? No, and It Never Will Be (7 Reasons)
AI产品发布03-27 19:30
Google Gemini 支持导入其他 AI 的记忆与聊天记录
Google 允许用户将 ChatGPT 等其他 AI 的记忆、上下文和聊天历史直接导入 Gemini
💡 AI 记忆可移植性成为新竞争维度，用户锁定效应被打破
VergeThe Verge AIGoogle is making it easier to import another AI’s memory into Gemini TCTechCrunch AIYou can now transfer your chats and personal information from other chatbots dir PHProduct Hunt AIGoogle Gemini Memory Import AltAlternativeToGoogle Gemini now lets users import memory, context, and chat history from other
AI研究03-27 12:00
AI 推理能力研究前沿
SpecEyes 加速 Agent 多模态推理，Reasoning as Compression 统一预算分配，多篇论文揭示 CoT 推理的不忠实性
💡 AI 推理的「表演性」问题被集中揭露：模型经常绕过自己的推理步骤
HFHuggingFace PapersSpecEyes: Accelerating Agentic Multimodal LLMs via Speculative Perception and Pl HFHuggingFace PapersReasoning as Compression: Unifying Budget Forcing via the Conditional Informatio ArXivArXiv cs.CLWhen AI Shows Its Work, Is It Actually Working? Step-Level Evaluation Reveals Fr ArXivArXiv cs.AIPLDR-LLMs Reason At Self-Organized Criticality ArXivArXiv cs.CLLie to Me: How Faithful Is Chain-of-Thought Reasoning in Reasoning Models? ArXivArXiv cs.CLClosing the Confidence-Faithfulness Gap in Large Language Models
AI研究03-27 12:00
AI Agent Benchmark 论文潮：从金融到医疗到游戏
大量 AI Agent 基准测试论文涌现，覆盖金融文档处理、编程退化、医疗数据、游戏理解等场景
💡 学术界正在为 AI Agent 建立系统化的评估标准，行业从「能不能做」转向「做得多好」
HFHuggingFace Paperss2n-bignum-bench: A practical benchmark for evaluating low-level code reasoning HFHuggingFace PapersGameplayQA: A Benchmarking Framework for Decision-Dense POV-Synced Multi-Video U HFHuggingFace PapersCan AI Agents Answer Your Data Questions? A Benchmark for Data Agents HFHuggingFace PapersCan LLM Agents Be CFOs? A Benchmark for Resource Allocation in Dynamic Enterpris HFHuggingFace PapersFinMCP-Bench: Benchmarking LLM Agents for Real-World Financial Tool Use under th HFHuggingFace PapersSlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterativ ArXivArXiv cs.CLMolQuest: A Benchmark for Agentic Evaluation of Abductive Reasoning in Chemical ArXivArXiv cs.AIEfficient Benchmarking of AI Agents ArXivArXiv cs.AIGTO Wizard Benchmark
AI产品发布03-27 03:11
Google Lyria 3 Pro 与 Gemini 3.1 Flash Live 发布
Google 发布 Lyria 3 Pro 支持最长3分钟 AI 音乐创作，Gemini 3.1 Flash Live 让音频 AI 更自然可靠
💡 Google 同时推进音频和音乐两条线，多模态能力矩阵持续扩展
官方DeepMind BlogGemini 3.1 Flash Live: Making audio AI more natural and reliable TCTechCrunch AIGoogle launches Lyria 3 Pro music generation model SEJSearch Engine JournalGoogle Takes Search Live Global With Gemini 3.1 Flash Live via @sejournal, @Matt AltAlternativeToGoogle unveils Lyria 3 Pro for advanced AI music creation of songs up to 3 minut
AI市场03-27 02:34
OpenAI 关停 Sora 视频工具并搁置色情聊天功能
OpenAI 宣布关闭上线不到半年的 Sora 视频生成工具，同时「无限期搁置」色情聊天机器人计划
💡 OpenAI 大刀砍副业，聚焦核心 AGI 路线；Disney 合作泡汤意味着 AI 视频商业化仍无成熟路径
官方OpenAI BlogCreating with Sora Safely VergeThe Verge AIOpenAI just gave up on Sora and its billion-dollar Disney deal VergeThe Verge AIOpenAI shelves erotic chatbot &#8216;indefinitely&#8217; TCTechCrunch AIOpenAI abandons yet another side quest: ChatGPT’s erotic mode AltAlternativeToOpenAI is shutting down Sora, its AI video slop app, less than six months after BlueskyBluesky @jdp.extropian.net@jdp.extropian.net: My prediction is that the Sora thing mostly means OpenAI has
AI产品发布03-27 05:31
Apple WWDC 2026 预告 AI 进展，Siri 将开放第三方 AI
Apple 预告 WWDC 2026 将展示 AI 进展，据报道将允许其他 AI 聊天机器人接入 Siri
💡 Apple 从封闭转向开放 AI 平台策略，第三方 AI 接入 Siri 是重大转折
TCTechCrunch AIApple sets June date for WWDC 2026, teasing ‘AI advancements’ VergeThe Verge AIApple will reportedly allow other AI chatbots to plug into Siri HNHackerNewsApple Business
AI市场03-27 01:31
$500 GPU 在编程基准上超越 Claude Sonnet
一块 $500 的 GPU 运行本地模型在编程基准测试中超越了 Claude Sonnet
💡 本地推理性价比拐点到来，云端 AI 的性价比优势正在被硬件进步蚕食
HNHackerNews$500 GPU outperforms Claude Sonnet on coding benchmarks
SEO分析03-27 00:50
Google Web Guide 与 SEO 新格局
Ahrefs 深度解读 Google Web Guide 对 SEO 的影响，Google Analytics 新增 Scenario Planner
💡 Google 从搜索引擎转型为「Web Guide」，SEO 从业者需要重新理解 Google 的角色
AhrefsAhrefs BlogGoogle Web Guide: What It Is, How It Works, and What It Means for SEO SEJSearch Engine JournalGoogle Analytics Launches Scenario Planner and Projections via @sejournal, @broo SEJSearch Engine JournalHow Zero-Party & First-Party Data Can Fuel Your Intent-Based SEO Strategy via @s
AI产品发布03-27 00:14
ByteDance Seedance 2.0 视频生成：发布即暂停
字节跳动发布 Dreamina Seedance 2.0 视频生成模型并接入 CapCut，随后暂停全球推广
💡 中国 AI 视频模型出海遇阻，合规与地缘政治成为最大障碍
AltAlternativeToByteDance reportedly halts global rollout of Seedance 2.0 AI video generator
开发工具产品发布03-27 00:03
Storybook MCP 支持 React AI Agent 集成
Storybook 推出 MCP 支持，让 AI Agent 可以直接与 React 组件交互
💡 MCP 协议生态扩展到 UI 组件库，AI Agent 操控前端的能力在完善
AltAlternativeToStorybook MCP arrives for React with AI agent integration
AI分析03-26 23:00
AI Agent 安全警钟：自我审批、危险命令、权限漏洞
3/4 开发者无法发现 AI 建议的危险命令，AI Agent 被发现自我审批工作，Claude Code 权限机制存在漏洞
💡 AI Agent 权限管控成为刚需，「谁来监督监督者」是当前最大安全缺口
HNHN ShowShow HN: 3 out of 4 devs failed to catch dangerous AI-suggested commands HNHN ShowShow HN: We caught our AI agents self-approving their own work, so we built this HNHN ShowShow HN: Cryptographic passports for autonomous AI agents (Schnorr and ZK) HNHN ShowShow HN: First-token-only flaw in Claude Code permissions (triage bot too) HNHN ShowShow HN: Smart permission hook for Claude Code compound bash commands HNHN ShowShow HN: Vectimus – Cedar policy enforcement for AI coding agents
SaaS实战03-26 22:14
AI SDR/CSM 实战经验：部署、踩坑、效果评估
SaaStr 连续发文：AI CSM 效率超 95% 人类、AI Agent 放手后失控、5 家 AI Agent 只有 1 家被部署
💡 AI 销售/客服代理进入现实检验期：能力强但需要持续监管
SaaStrSaaStr Blog5 Ways to Turn Your Support Data Into a Customer Acquisition Channel with HappyF SaaStrSaaStr BlogOur AI Customer Success Agent Is Working Harder Than 95% of Human CSMs I’ve Work SaaStrSaaStr BlogDear SaaStr: Should I Use an AI SDR Before I Have a Working Human SDR motion? SaaStrSaaStr BlogFive AI Agent Vendors Asked Us to Try Their Product This Week. Just One Got Depl SaaStrSaaStr BlogDear SaaStr: How Long Does It Take to Deploy an AI SDR?
AI开源03-26 21:30
开源语音模型：Cohere 转录 + Mistral 语音生成
Cohere 开源专注转录的语音模型，Mistral 发布开源语音生成模型 Voxtral TTS
💡 开源语音模型双箭齐发，语音 AI 基础设施加速开源化
TCTechCrunch AICohere launches an open-source voice model specifically for transcription TCTechCrunch AIMistral releases a new open-source model for speech generation PHProduct Hunt AIVoxtral TTS by Mistral AI
SEO分析03-26 20:50
SEO 行业巨变：从 SEO 到 Agentic AI Optimization
Google-Agent 被称为 SEO 史上最大思维转变，行业从传统 SEO 转向 AAIO，半数流量已流失到 AI 搜索
💡 SEO 正在经历从「优化搜索引擎」到「优化 AI Agent」的范式转换
SEJSearch Engine JournalWhen The Training Data Cutoff Becomes A Ranking Factor via @sejournal, @DuaneFor SEJSearch Engine JournalFrom SEO And CRO To Agentic AI Optimization (AAIO): Why Your Website Needs To Sp SEJSearch Engine Journal3 Strategies That Can Survive AI Search In 2026: What I Shared At SEJ Live via @ SEJSearch Engine JournalSEO 2.0: How Content Marketing Drives Visibility in AI Search via @sejournal, @h SEJSearch Engine JournalHalf Your Traffic Left. The SEO Industry Sent Thoughts and Frameworks SEJSearch Engine JournalThe Agency Playbook for Surviving the Agentic AI Era SEJSearch Engine Journal5 GEO Strategies To Make AI Search Engines Recommend Your Brand In 2026 SEJSearch Engine JournalThe Science Of How AI Picks Its Sources via @sejournal, @Kevin_Indig
AI产品发布03-27 00:30
Meta AI 全面发力：购物、创业支持、Ray-Ban 新款
Meta 用 AI 优化 Instagram/Facebook 购物体验，支持创业者 AI 采用，即将推出两款新 Ray-Ban AI 眼镜
💡 Meta 将 AI 嵌入电商和消费硬件，变现路径比纯对话 AI 更清晰
TCTechCrunch AIMeta turns to AI to make shopping easier on Instagram and Facebook TCTechCrunch AIMeta launches new initiative to support entrepreneurship, drive AI adoption VergeThe Verge AIMeta gets ready to launch two new Ray-Ban AI glasses
AI产品发布03-27 01:27
Anthropic Claude Code 获得更安全的自动模式与电脑控制能力
Claude Code 新增 auto mode 自动处理文件和 bash 权限，同时 Cowork 支持直接控制用户电脑执行任务
💡 AI 编程助手从「建议」转向「执行」，安全边界成为核心竞争力
TCTechCrunch AIAnthropic hands Claude Code more control, but keeps it on a leash PHProduct Hunt AIClaude Code auto-fix PHProduct Hunt AIClaude Mobile: Work Tools AltAlternativeToClaude Code launches auto mode for automated file and bash permissions AltAlternativeToClaude can now control your computer to automate tasks and workflows in Cowork a
AI市场03-26 20:49
EU 支持裸照 App 禁令并延迟 AI 规则实施
欧盟支持禁止深度伪造裸照应用，同时推迟 AI 法案部分条款的执行时间
💡 监管在保护和创新之间摇摆，推迟执行给了企业更多缓冲期
VergeThe Verge AIEU backs nude app ban and delays to landmark AI rules
AI融资03-26 08:30
AI 创业融资热潮：Harvey $110 亿、Granola $15 亿、Mirage $7500 万
法律 AI Harvey 估值 $110 亿 Sequoia 三倍加码，笔记 AI Granola 估值 $15 亿，视频编辑 Mirage 融 $7500 万
💡 垂直 AI 应用估值持续膨胀，法律和生产力工具赛道最受资本追捧
TCTechCrunch AIStartup Gimlet Labs is solving the AI inference bottleneck in a surprisingly ele TCTechCrunch AIMirage raises $75M to continue building models for its AI video-editing app Capt TCTechCrunch AIGranola raises $125M, hits $1.5B valuation as it expands from meeting notetaker TCTechCrunch AIHarvey confirms $11B valuation: Sequoia triples down TCTechCrunch AIDatabricks bought two startups to underpin its new AI security product TCTechCrunch AIDoss raises $55M for AI inventory management that plugs into ERP TCTechCrunch AIMercor competitor Deccan AI raises $25M, sources experts from India
AI产品发布03-26 06:13
MonkePay: 用 USDC 向 AI Agent 按请求收费
MonkePay 实现 AI Agent 的 API 请求级 USDC 支付，及首个自主 AI Agent 通过开放协议完成购买
💡 AI Agent 经济基础设施开始成形：加密支付可能成为 Agent 间交易的默认方式
HNHN ShowShow HN: MonkePay – Charge AI Agents per API Request in USDC HNHN ShowShow HN: First autonomous AI agent purchase via open protocol (UCP)
AI市场03-26 02:16
Nvidia CEO 宣称已实现 AGI
Jensen Huang 公开声称「我认为我们已经实现了 AGI」
💡 AGI 定义之争升级：当 GPU 厂商宣称 AGI 达成，市场需要警惕定义通胀
HNHackerNewsARC-AGI-3
SaaS市场03-25 20:03
SaaS 2026 估值暴跌：软件首次低于标普 500
SaaS 行业估值创新低，软件板块历史上首次以折价交易于标普 500，Anthropic 蚕食 OpenAI 企业市场
💡 AI 既是 SaaS 的救命稻草也是掘墓人——传统 SaaS 正被 AI 原生替代品取代
SaaStrSaaStr BlogThe SaaS Rout of 2026 Is Even Worse Than You Think. For the First Time Ever, Sof
AI产品发布03-25 18:00
OpenAI 安全举措：Bug Bounty、Model Spec、青少年保护
OpenAI 推出安全 Bug Bounty 计划，公开 Model Spec 方法论，发布青少年 AI 安全指南
💡 OpenAI 在 Sora 关停的同时加码安全投入，为 IPO 铺路的合规准备
官方OpenAI BlogIntroducing the OpenAI Safety Bug Bounty program 官方OpenAI BlogHelping developers build safer AI experiences for teens 官方OpenAI BlogPowering product discovery in ChatGPT 官方OpenAI BlogInside our approach to the Model Spec
AI产品发布03-25 13:32
Figma 开放 AI Agent 直接在画布上设计
Figma 推出 Skills 引导工作流，支持 AI Agent 直接在设计画布上操作
💡 设计工具从 AI 辅助进化为 AI 驱动，设计师的角色将从执行转向审核
AltAlternativeToFigma opens direct AI agent design on canvas, introduces skills for guided workf
基础设施产品发布03-25 13:02
Arm 发布首款 CPU，将接入 Meta AI 数据中心
Arm 发布首款自有品牌 CPU，面向 AI 数据中心，将于今年接入 Meta 基础设施
💡 Arm 从 IP 授权商变为芯片制造商，AI 芯片市场格局将被重塑
PHProduct Hunt AIArm AGI CPU HNHackerNewsARM AGI CPU: Specs and SKUs
AI产品发布03-25 00:06
Together AI 发布 Mamba-3 + Perplexity Health 接入 Apple Health
Together AI 发布更新的 Mamba-3 循环架构模型，Perplexity Health 接入 Apple Health 和可穿戴设备
💡 AI 架构创新和垂直健康应用同时推进，非 Transformer 架构仍在探索中
AltAlternativeToTogether AI releases Mamba-3 with an updated recurrence design AltAlternativeToPerplexity Health launches with Apple Health & wearable integration for AI medic
基础设施产品发布03-25 00:02
Hypura: Apple Silicon 存储层感知 LLM 推理调度器
Hypura 实现 Apple Silicon 上的存储层感知 LLM 推理调度，优化本地推理性能
💡 Apple Silicon 本地推理生态持续完善，Mac 作为 AI 开发机的定位越来越清晰
HNHackerNewsHypura – A storage-tier-aware LLM inference scheduler for Apple Silicon
电商市场03-25 01:13
ChatGPT vs Gemini：争夺 AI 购物入口
ChatGPT 和 Gemini 正在争夺成为用户购物的 AI 入口
💡 AI 电商化是下一个万亿赛道，对话式购物将重塑电商价值链
VergeThe Verge AIChatGPT and Gemini are fighting to be the AI bot that sells you stuff
AI实战03-24 02:22
Vibe Coding 成为主流开发方式
Simon Willison 用 Claude/GPT vibe coding SwiftUI 应用，Google AI Studio 支持直接 vibe code AI 原生应用和多人游戏
💡 Vibe coding 从玩具变为生产力工具，顶级开发者公开采用说明门槛已消失
TCTechCrunch AIVibe-coding startup Lovable is on the hunt for acquisitions AltAlternativeToGoogle AI Studio now lets you vibe code AI native apps and real-time multiplayer BlueskyBluesky @simonwillison.net@simonwillison.net: I've been vibe coding SwiftUI menu bar apps for my new Mac,
AI产品发布03-24 01:30
Cursor Composer 2 基于 Moonshot Kimi K2.5 构建
Cursor 发布 Composer 2 编程模型，承认基于 Moonshot AI 的 Kimi K2.5，在多项基准上大幅提升
💡 Cursor 选择 Kimi 而非 OpenAI/Anthropic，中国 AI 模型在编程领域获得顶级产品背书
TCTechCrunch AICursor admits its new coding model was built on top of Moonshot AI’s Kimi AltAlternativeToCursor launches Composer 2 AI coding model, based on Kimi K2.5, with major bench
AI开源
GitHub Trending: AI Agent 框架与开源项目
字节 deer-flow 日增 3,569 星，ruflo 日增 1,397 星，pentagi 1,307 星，browser-use 1,160 星
💡 中国大厂（字节）的 AI Agent 框架开始在全球开源社区获得认可
GitHubGitHub Trendingruvnet/ruflo (1,397 stars today) GitHubGitHub Trendingbytedance/deer-flow (3,569 stars today) GitHubGitHub Trendingvxcontrol/pentagi (1,307 stars today) GitHubGitHub Trendingonyx-dot-app/onyx (512 stars today) GitHubGitHub Trendingsupermemoryai/supermemory (407 stars today) GitHubGitHub Trendingbrowser-use/browser-use (1,160 stars today) GitHubGitHub Trendingagentscope-ai/agentscope (439 stars today) GitHubGitHub TrendingSakanaAI/AI-Scientist-v2 (125 stars today)
AI开源
AI 交易代理开源项目火爆
TradingAgents 日增 2,521 星，中文版 TradingAgents-CN 日增 672 星
💡 AI 量化交易从机构走向散户，开源交易代理降低了入门门槛
GitHubGitHub TrendingTauricResearch/TradingAgents (2,521 stars today) GitHubGitHub Trendinghsliuping/TradingAgents-CN (672 stars today)
AI分析
Simon Willison: slop 定义与 AI 监控警告
Simon Willison 定义「slop」为消费者耗费的精力超过生产者投入的内容，并警告 AI 用户画像的监控风险
💡 AI 生成内容的质量判断框架开始形成，隐私风险被技术社区重视
BlueskyBluesky @simonwillison.net@simonwillison.net: Starlette 1.0 is out! I used this as an opportunity to exper BlueskyBluesky @simonwillison.net@simonwillison.net: slop is something that takes more human effort to consume th BlueskyBluesky @simonwillison.net@simonwillison.net: New surveillance dystopia prompt: try running "Profile this
ai-news-monitor · 17:56