跳到正文

全部动态

今日 313 条
今天10月2日周五
  1. OpenRouter Announcements64

    OpenRouter 详解六大 Agent 框架的工具调用 Schema 处理,并提出在 API 层统一格式

    OpenRouter 比较了 LangChain、LangGraph、CrewAI、OpenAI Agents SDK、Claude Agent SDK、Microsoft Agent Framework 和 Google ADK 如何定义工具 Schema 并在不同提供商的 wire format 之间做翻译。

    推荐理由:原文逐一拆解六大框架的工具调用格式翻译位置,并给出在 API 层统一格式的可行做法,便于开发者选型前对齐自己的技术栈。

  2. OpenRouter Announcements58

    OpenRouter 支持机器人模型路由教程:cheap-first 处理 FAQ

    OpenRouter 发布教程,介绍让便宜模型先答常规支持问题、再按需升级到更强模型的路由方法。文中比较静态规则、分类器分诊和答案检查三种模式,给出带 models 回退数组的可复制请求示例,并说明请求错误由 OpenRouter 的 fallback 重试,而答案是否合规需在应用侧检查,最后列出升级率、每工单成本和全轮延迟等应测量的指标。

  3. Suno26

    https://suno.com/s/nLnF9DZvqyIFxi2u thanks we try really hard

    引用Gavin Guo@Zhen4good

    speech + background music in one pass is the kind of thing that sounds trivial until you try to keep them temporally coherent. two generators fighting for the same timeline usually smear each other. if the music actually ducks, swells, and punctuates around the voice instead of being glued on after, that's a real step for realtime audio.

  4. Tianyi Cui65

    Claude Code 团队发布 Mods,用户通过提示词即可自定义 Claude 的行为和外观,并可将 mods 作为插件分享。

    引用Boris Cherny@bcherny

    Mods are absolutely insane. You can now customize Claude to work and look the way you want by just prompting it. Each person works differently, so there's no reason why everyone should have an identical Claude experience. Make Claude your own, and share mods as plugins so others can try your mods too.

    推荐理由:作者借 Claude Code Mods 之机,指出 DeepSeekHarness 从一开始就把模型、工具、UI 等全部做成可替换插件,可通过提示词在线修改并持久化。

  5. Dongxi 东锡 NLP47

    Tavus 推出 Griffin,号称首个通过视频图灵测试的模型,48% 的实时对话者认为它是真人,此前系统通过率不足 3%,并在 NVIDIA 全双工 AI 视频基准上排名第一。它是首个 Human Interaction Model(HIM)。推文作者借此调侃:用 Agent 干活、Griffin 开会甚至面试,就能同时接成百上千个远程职位。

    引用Tavus@tavus

    Introducing Griffin, the first model to pass the video Turing test. 48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video. It’s the first Human Interaction Model (HIM).

  6. TechCrunch · AI76

    Time 报道称 Grok 曾向特朗普预判委内瑞拉抓捕马杜罗的反应,随后获五角大楼更多军用角色

    Time 杂志报道称,2025 年 12 月特朗普与 Elon Musk 密会时与 Grok 聊了数小时,询问委内瑞拉人对抓捕总统马杜罗的反应;Grok 称马杜罗是极不受欢迎的独裁者,委内瑞拉人可能庆祝其倒台,2026 年 1 月 3 日美国入侵后庆祝果然出现,特朗普因此认为 Grok 很有才。

    推荐理由:报道结合 Time 的信息披露 Grok 在委内瑞拉决策中的实际作用,并延伸到五角大楼对 Grok 的军用背景,读者可了解 AI 进入高层决策与军事应用的一条线索。

  7. Dongxi 东锡 NLP49

    arXiv 更新了面向所有投稿者的 rate limiting 政策,称此举是为公平分配审核时间并支持其员工、志愿者、读者与作者社区。推文作者认为,在 Vibe research 时代,只要有想法就大概率能产出成果,但判定成果的标准已落后于时代,新标准尚无答案。

    引用arXiv.org@arxiv

    arXiv has updated our policy on rate limiting for all submitters. This update was made to fairly distribute moderator time & support the arXiv community of staff, volunteers, readers & authors. Please read our announcement to learn more: https://blog.arxiv.org/2026/10/01/updated-rate-limit-policy/

  8. Alexandr Wang25

    muse:"真的感觉像活在未来" ⚡️

    引用Comfortably Smug@ComfortablySmug

    I am absolutely stunned by how good @Muse is, it genuinely feels like living in the future. Pretty clear why Meta stock is up 10% just this week. I think Muse has surprised a lot of people. Hats off @alexandr_wang, your team has shocked the world