// paginated posts

const page = 18

const article = {
  title: "DeepSeek急招Agent方向！一口气放17个岗位，重度Vibe Codin",
  publishedAt: "2026.03.26",
  tags: ["科技","AI","人工智能"],
  href: "/blogs/ai-daily-2026-03-26"
}

export default article

const article = {
  title: "DeepSeek急招Agent方向！一口气放17个岗位，重度Vibe Codin",
  publishedAt: "2026.03.26",
  tags: ["科技","AI","人工智能"],
  href: "/blogs/ai-daily-2026-03-26"
}

export default article

article.tsexport default

publishedAt2026.03.26

const article = definePost()

DeepSeek急招Agent方向！一口气放17个岗位，重度Vibe Codin

# DeepSeek急招Agent方向！一口气放17个岗位，重度Vibe Coding优先

科技AI人工智能

readMore()

const article = {
  title: "Claude电脑控制发布/对齐评估漏洞/白宫AI政策框架",
  publishedAt: "2026.03.25",
  tags: ["AI资讯","大模型","Claude"],
  href: "/blogs/ai-daily-2026-03-25"
}

export default article

const article = {
  title: "Claude电脑控制发布/对齐评估漏洞/白宫AI政策框架",
  publishedAt: "2026.03.25",
  tags: ["AI资讯","大模型","Claude"],
  href: "/blogs/ai-daily-2026-03-25"
}

export default article

article.tsexport default

publishedAt2026.03.25

const article = definePost()

Claude电脑控制发布/对齐评估漏洞/白宫AI政策框架

2026年3月24日AI资讯日报：Anthropic发布Claude电脑自主操控功能，Mistral推出企业私有建模平台Forge，谷歌发布270M超轻量Gemma微调方案。大模型对齐评估漏洞曝光，简单微调可绕过审查。沃顿研究揭示人类认知投降现象，SK海力士利润首超三星。

AI资讯大模型Claude

readMore()

const article = {
  title: "AI资讯日报：Anthropic发布架构师认证 微信接入OpenClaw",
  publishedAt: "2026.03.24",
  tags: ["AI资讯","大模型","Agent"],
  href: "/blogs/ai-daily-2026-03-24"
}

export default article

const article = {
  title: "AI资讯日报：Anthropic发布架构师认证 微信接入OpenClaw",
  publishedAt: "2026.03.24",
  tags: ["AI资讯","大模型","Agent"],
  href: "/blogs/ai-daily-2026-03-24"
}

export default article

article.tsexport default

publishedAt2026.03.24

const article = definePost()

AI资讯日报：Anthropic发布架构师认证微信接入OpenClaw

2026年3月23日AI资讯日报：Anthropic发布Claude架构师认证考试，腾讯微信接入OpenClaw智能体，马斯克启动Terafab超算芯片工厂，SQL可直接训练AI分类器，Arc发布BioReason-Pro蛋白质模型，AI编程或引发软件业裁员潮

AI资讯大模型Agent

readMore()

const article = {
  title: "DeepSeek核心工程师郭达雅离职 曾参与V2/V3/R1研发",
  publishedAt: "2026.03.23",
  tags: ["DeepSeek","大模型","AI工程师"],
  href: "/blogs/ai-daily-2026-03-23"
}

export default article

const article = {
  title: "DeepSeek核心工程师郭达雅离职 曾参与V2/V3/R1研发",
  publishedAt: "2026.03.23",
  tags: ["DeepSeek","大模型","AI工程师"],
  href: "/blogs/ai-daily-2026-03-23"
}

export default article

article.tsexport default

publishedAt2026.03.23

const article = definePost()

DeepSeek核心工程师郭达雅离职曾参与V2/V3/R1研发

DeepSeek核心工程师郭达雅被曝离职，他是V2、V3、R1等模型的核心作者，参与了Coder、Math、Prover等专项模型研发。其博士毕业后第三天即完成毕业论文要求，多次获得腾讯广告算法大赛冠军，还参加综艺《燃烧吧！天才程序员》展现编程实力。

DeepSeek大模型AI工程师

readMore()

const article = {
  title: "北大彭宇新团队提出TARA方法：注入类别树知识提升分层视觉识别",
  publishedAt: "2026.03.22",
  tags: ["多模态大模型","计算机视觉","分层识别"],
  href: "/blogs/ai-daily-2026-03-22"
}

export default article

const article = {
  title: "北大彭宇新团队提出TARA方法：注入类别树知识提升分层视觉识别",
  publishedAt: "2026.03.22",
  tags: ["多模态大模型","计算机视觉","分层识别"],
  href: "/blogs/ai-daily-2026-03-22"
}

export default article

article.tsexport default

publishedAt2026.03.22

const article = definePost()

北大彭宇新团队提出TARA方法：注入类别树知识提升分层视觉识别

北京大学彭宇新团队在CVPR 2026发表新研究，提出分类感知表征对齐方法TARA，将类别树结构知识注入多模态大模型，解决分层视觉识别中同层判别性差、跨层一致性差、新类泛化性差三大难题。

多模态大模型计算机视觉分层识别

readMore()

const article = {
  title: "2026年3月20日AI资讯：Qwen3.5 Max数学全球第三",
  publishedAt: "2026.03.21",
  tags: ["AI资讯","大模型","AI编程"],
  href: "/blogs/ai-daily-2026-03-21"
}

export default article

const article = {
  title: "2026年3月20日AI资讯：Qwen3.5 Max数学全球第三",
  publishedAt: "2026.03.21",
  tags: ["AI资讯","大模型","AI编程"],
  href: "/blogs/ai-daily-2026-03-21"
}

export default article

article.tsexport default

publishedAt2026.03.21

const article = definePost()

2026年3月20日AI资讯：Qwen3.5 Max数学全球第三

今日AI要闻：阿里Qwen3.5 Max数学能力冲上LMSYS全球第三，Cursor发布Composer 2编程代理，Devin开启多智能体编排。科研前沿：潜空间学习揭示物理规律，MIT无线信号穿透成像，NanoGPT训练效率提升十倍。

AI资讯大模型AI编程

readMore()

const article = {
  title: "轻舟智航于骞：世界模型+强化学习是通向物理AI的必经之路",
  publishedAt: "2026.03.20",
  tags: ["物理AI","世界模型","强化学习"],
  href: "/blogs/ai-daily-2026-03-20"
}

export default article

const article = {
  title: "轻舟智航于骞：世界模型+强化学习是通向物理AI的必经之路",
  publishedAt: "2026.03.20",
  tags: ["物理AI","世界模型","强化学习"],
  href: "/blogs/ai-daily-2026-03-20"
}

export default article

article.tsexport default

publishedAt2026.03.20

const article = definePost()

轻舟智航于骞：世界模型+强化学习是通向物理AI的必经之路

轻舟智航CEO于骞在德国慕尼黑汽车论坛演讲，阐述物理AI是全球AI下一浪潮，自动驾驶是打开物理AI大门的钥匙，世界模型+强化学习是实现物理AI的关键技术路径。

物理AI世界模型强化学习

readMore()

const article = {
  title: "MiniMax M2.7发布：多智能体协作与代码能力跃升",
  publishedAt: "2026.03.19",
  tags: ["AI大模型","MiniMax","多智能体"],
  href: "/blogs/ai-daily-2026-03-19"
}

export default article

const article = {
  title: "MiniMax M2.7发布：多智能体协作与代码能力跃升",
  publishedAt: "2026.03.19",
  tags: ["AI大模型","MiniMax","多智能体"],
  href: "/blogs/ai-daily-2026-03-19"
}

export default article

article.tsexport default

publishedAt2026.03.19

const article = definePost()

MiniMax M2.7发布：多智能体协作与代码能力跃升

MiniMax发布M2.7模型，支持多智能体原生协作，代码能力追平GPT-5.3，在复杂任务处理和Agent团队协作方面有质的飞跃，已上线开放平台。

AI大模型MiniMax多智能体

readMore()

const article = {
  title: "GPT-5.4一周狂赚10亿美元ARR",
  publishedAt: "2026.03.18",
  tags: ["OpenAI","GPT-5.4","大模型"],
  href: "/blogs/ai-daily-2026-03-18"
}

export default article

const article = {
  title: "GPT-5.4一周狂赚10亿美元ARR",
  publishedAt: "2026.03.18",
  tags: ["OpenAI","GPT-5.4","大模型"],
  href: "/blogs/ai-daily-2026-03-18"
}

export default article

article.tsexport default

publishedAt2026.03.18

const article = definePost()

GPT-5.4一周狂赚10亿美元ARR

OpenAI GPT-5.4上线一周日处理5万亿token，年化收入10亿美元。在ARC-AGI测试中效率提升32倍，成本仅0.37美元/任务，83%概率胜过人类工作。

OpenAIGPT-5.4大模型

readMore()

const article = {
  title: "零跑，首次全年盈利！",
  publishedAt: "2026.03.17",
  tags: ["人工智能","AI","大模型"],
  href: "/blogs/ai-daily-2026-03-17"
}

export default article

const article = {
  title: "零跑，首次全年盈利！",
  publishedAt: "2026.03.17",
  tags: ["人工智能","AI","大模型"],
  href: "/blogs/ai-daily-2026-03-17"
}

export default article

article.tsexport default

publishedAt2026.03.17

const article = definePost()