跳到正文
6月26日周五
  1. Orange AI41

    用户指出刚上线 Cola 的 Seed 2.1 Pro 模型(自称原生多模态、多模态最强,相比 2.0 增强 coding 和 Agent 能力)在推理时上下文精度极差:常搞错人物、性别、时间。用户指出错误后模型频繁道歉,态度端正但问题明显。

    引用Orange AI@oran_ge

    刚刚 Cola 上线了最新的 Seed 2.1 Pro 模型 这个模型是原生多模态模型,是目前的多模态最强模型。 相比 2.0 版本,增强了 coding 能力和 Agent 能力,具体的评测可以参考藏师傅的文章 体验地址 http://colaos.ai

  2. jason38

    Jason Liu 用两小时语音备忘录替代键盘操作,将对话转录输入 Codex,后者直接生成了产品公告、发布视频脚本和公关材料。引用推文补充,此前他与 @guinnesschen 进行了两小时漫谈,涵盖功能想法、设计哲学、AI 未来等,原始转录交给 Codex 后产出的 PR 和创意作品几乎完美捕捉了对话意图。

    引用Guinness Chen@guinnesschen

    Yesterday @jxnlco and I had a fantastic two-hour conversation that wandered through feature ideas, design philosophy, the future of AI, and dozens of tangents. We fed the raw transcript to Codex, and it turned it into PRs and creative work that captured our intent almost perfectly. The twists and turns are what make the prompt good. Please don't edit them out.

6月25日周四
6月24日周三
  1. PixVerse23

    使用AI为你的游戏制作过渡过场动画?太棒了。整个场景充满夏日清新感。期待新项目完成 @patata1216

    引用美咲@patata1216

    ぜんざいのアトリエ開発迷走中😅 海エリアはやっぱり釣りでしょう、とかなんとか言うことで、釣りミニゲームを組み込み。 AI開発らしく、Pixverseで作ったAI動画を差し込んで作ってみたよ。 やってみると、こういうバリエーション動画を作るのはやっぱキャンバス便利だね #pixversecpp @PixVerse_

  2. OpenBMB65

    🥳感谢分享,@MLBoy_DaisukeMajima 🚀 MiniCPM-V 4.6 在设备上以这样的速度运行,实在令人印象深刻——尤其是在 Apple Core AI 上以不到 2B 参数跑出。 干得漂亮,推动高效多模态 AI 向前发展。🫡

    引用MLBoy_DaisukeMajima@JackdeS11

    📸 MiniCPM-V 4.6 —— 2B 参数以下最强的视觉模型之一 —— 现在通过 Apple Core AI 在 iPhone 17 Pro 上以约 51 tok/s 运行。 🤗 http://huggingface.co/mlboydaisuke/MiniCPM-V-4.6-CoreAI 🛠️ http://github.com/john-rocky/coreai-model-zoo

    推荐理由:社区把 MiniCPM-V 4.6 搬上 iPhone 17 Pro,跑出 51 tok/s,还给了代码和模型,做端侧多模态的可以直接跑起来了。

  3. MiniMax (official)39

    用户@whosamberella为学习仿人机器人知识,利用MiniMax M3进行调研,并在OpenCode中搭建了一个仿人机器人目录网站(humanoid-index-one.vercel.app)。其SVG原型不仅能展示,还能帮助学习不同仿人机器人的特征。MiniMax将其过程概括为:研究→综合→构建。

    引用Amber Shen@whosamberella

    I wanted to learn about humanoid robots 🤖, then decided the most reasonable thing to do was build myself a humanoid catalog to learn from. https://humanoid-index-one.vercel.app/ @MiniMax_AI M3 researched the space, then built the catalog in @opencode . The SVG prototype doesn’t just look good. It actually teaches you the distinct characteristics of each humanoid.

  4. OpenBMB63

    面壁智能(OpenBMB)的扩散式 TTS 模型 VoxCPM-0.5B 已通过 Apple Core AI 完全部署至 iPhone 端侧,无需联网。该模型整合了 MiniCPM4 语言模型、LocDiT flow-matching 和 AudioVAE,每一层均运行于 Neural Engine 和 GPU 上。模型权重和部署代码已开源至 HuggingFace 与 GitHub。

    引用MLBoy_DaisukeMajima@JackdeS11

    VoxCPM now runs FULLY on-device on iPhone — via Apple's Core AI. @OpenBMB's diffusion TTS (MiniCPM4 LM + LocDiT flow-matching + AudioVAE), every layer on the Neural Engine / GPU. No network. 🤗 https://huggingface.co/mlboydaisuke/VoxCPM-0.5B-CoreAI 📦 https://github.com/john-rocky/coreai-model-zoo

  5. Rohan Paul41

    NVIDIA Rubin AI 服务器用 45°C 水-乙二醇冷却液直接冷却芯片和网络部件,取代传统空气冷却。在适宜气候下,干式冷却器可替代冷却塔,设施冷却水用量从约 260 万加仑/MW/年降至接近零。液冷为闭环循环,不持续蒸发水。一个 50MW AI 设施每年可节省超 400 万美元冷却能源和水费。全液冷 Rubin 服务器还将系统从 6 个机架单元缩减至 2 个,在相同空间内容纳更多计算。引用 NVIDIA 数据称,数据中心用水仅占美国日常用水量的 0.2%,液冷却正大幅降低水耗并创造热量再利用机会。

    引用NVIDIA@nvidia

    Water usage has been a hot topic in the AI data center world, but the numbers may surprise you. According to the Manhattan Institute, data centers use 0.2 percent of daily water usage in the U.S. and that number has dramatically decreased in the past few years due to a new method: liquid cooling. By moving to 45°C liquid cooling, AI factories in favorable climates can use dry coolers instead of conventional cooling-tower-based systems, cutting facility cooling water use from roughly 2.6M gallons per MW per year to near zero. Liquid cooling enables AI factories to be both water and energy efficient, while creating opportunities for heat reuse and dispersal to local communities, allowing these factories to become energy grid assets. Learn more below ⬇️ https://nvda.ws/4gAVGoJ

  6. Rohan Paul60

    Atomic Chat 集成 Cline,使代码智能体能在本地 AI 模型上运行。Cline 作为智能体层,可读文件、编辑代码、运行终端命令、检查错误并请求批准,已被 800 万+开发者信任。Atomic Chat 作为模型层,支持 1000+ 模型离线本地运行,并暴露本地 API 供其他工具调用。整个过程私有、免费、开源,体现了强大智能体回归本地私有设备的趋势。

    引用atomic.chat@atomic_chat_hq

    Run Cline on Local AI models with Atomic Chat! @cline is a coding agent trusted by 8M+ developers. Write, refactor, ship code securely on your own hardware with local models powered by @atomic_chat_hq — no cloud, private, free and open-source

6月23日周二
  1. Nathan Lambert32

    Nathan Lambert 为新书新增 7.4 小时讲座视频,内容涵盖从 2015 年 Hinton 知识蒸馏论文到当下多教师 on-policy 蒸馏(OPD、MOPD、OPSD)。视频重点讲解了使 on-policy distillation 适配主流 RL 框架所需的 3–4 项核心公式改动,并回顾了合成数据如何逐步占领训练后数据研究领域。此外还介绍了 Constitutional AI、AI 反馈以及将评分准则作为奖励等主流方法。时间线:00:00 合成数据兴起,10:50 师生蒸馏背景,24:47 on-policy 蒸馏,37:11 Constitutional AI,45:50 评分准则与结论。

    引用Nathan Lambert@natolambert

    New lecture for the book! Nominally about synthetic data, but mostly is a walk through of the distillation literature from the Hinton 2015 paper to multi-teach on-policy distillation of today! At 7.4 hours of video in my post-training brain dump and counting :) It was fun to stare at the math long enough and talk through the 3-4 core changes that needed to be made to the original formulation to have on-policy distillation be ready for the mainstream like it is today (and in RL frameworks). Otherwise, I include a bit of a history lesson for how synthetic data generally slowly took over all post-training data research (it wasn't always the case)! Then I do some 101 review on constitutional AI, rubrics, and other popular methods. 00:00 The emergence of synthetic data 10:50 Background on teacher-student knowledge-distillation 24:47: On-policy distillation (OPD, MOPD, and OPSD) 37:11 Constitutional AI & AI Feedback 45:50 Rubrics as rewards & conclusions Ofc, watch on YouTube etc.

  2. PixVerse26

    用 GPT image 2 制作故事板,再通过 PixVerse 的 Seedance 2.0 转为电影级短片——一位教练的最后礼物,一段世界杯决赛的 AI 故事。主推文称赞道:用 Pixverse 把这个故事呈现得很棒。

    引用𝐌@Strength04_X

    COACH'S LAST GIFT ⚽ An old pair of boots, a quiet tunnel, and a coach holding a faded photo from the stands What happens next is a World Cup final story I created entirely with AI Built the storyboard with GPT image 2 and turned it into a cinematic film with Seedance 2.0 on @PixVerse_ Didn't expect an AI football story to hit me this hard honestly Full video below 👇