X

让 Codex 自己做了一条视频介绍了一下这个视频生成方案 藏师傅的 PPT Skill 负责美学、版式、动效 HyperFrames 负责时间线和渲染、字幕 Listenhub Skill 负责配音 即梦 CLI 负责 前端无法生成的演示和短 B-roll Video

藏师傅的 PPT Skill+Codex+Heygen HyperFrames 这个组合太顶了！ 可以直接基于问当生成带动效的解释视频 而且 Codex 居然可以在聊天里面直接预览视频，这个挺厉害的。 再加上即梦 CLI 补几个真实视频片段，用来做一些产品更新介绍之类的一点问题没有。 Video

CItadel的CEO和创始人肯·格里芬对AI的看法发生了重大转变： “首先，在过去的几个月里，人工智能工具包的生产力发生了飞跃式的变化。它比九个月前强大得多。对我们 Citadel 来说，这使我们能够开发出更广泛的人工智能应用场景。 坦白说，看着人工智能代理在几个小时或几天内就能完成我们通常需要金融硕士和博士花费数周甚至数月才能完成的工作，这真的很有意思。这些不是中层白领工作，而是极其高技能的工作，我姑且用一个词来形容，它们被智能人工智能自动化了。 说实话，有一次周五回家，我其实挺沮丧的，因为你能预见到这将对社会产生多么巨大的影响。当你亲眼目睹这一切，当你看到过去需要数年才能完成的工作在几天或几周内就完成了，你会觉得，哇，这简直太不可思议了。 这是我第一次亲眼见证人工智能在我们公司内部产生的真正影响。” #AI #金融革命 #Citadel Video

说个所有AI创业者都不愿意承认的事实： 现在做一个AI工具的门槛已经降到了地板， 普通人做一个AI工具都只需要一天， 但学会用它干成一件事，却至少得一个月， 感觉像是AI时代的一个悖论😅 5.7M 阅读 23 万点赞的这条推，表面看是游戏圈在自嘲， 视频展示的是一颗树莓 237 万个高斯点，做一筐扔进游戏直接 2 FPS， 但如果把游戏开发四个字去掉，你会发现这是 2026 年所有 AI 工具用户的共同故事。 我先先把这个梗讲透： 原推作者 @DanyBittel 用 90 组焦点堆栈、每组 68 张照片，重建出来这颗树莓，总共 237 万个高斯点， 这是一种叫 3D Gaussian Splatting 的新型 AI 重建技术，简称 3DGS， 视觉效果有多吓人呢？ 每一颗小果粒的绒毛、表面光泽、半透明的果肉质感全都纤毫毕现，在 RTX 3060 Ti 这种中端显卡上还能跑 100+ FPS，前提是只有这一颗🙃 @nazbowling102 的笑点在这里，老哥迫不及待想看哪个独立游戏开发者一激动，把一整筐这种树莓当道具扔进游戏里，然后纳闷为啥游戏跑 2 FPS🤣 我觉得这个吐槽之所以 5.7M 阅读，是因为它戳中了游戏圈的集体回忆—— Monster Hunter Wilds 一颗八角茴香用了 2048 乘 4096 的纹理直接卡帧，Cities Skylines 2 给行人建了高精度牙齿模型，全都是一个小道具毁全局的真实事故。 但这条推真正让我深入研究的还不是游戏开发的事，虽然我是个游戏爱好者，但对于游戏开发是个小白。 ayi干货输出开始！ 咱们把游戏开发四个字去掉，这个故事正在所有 AI 工具领域都能同步上演， AI 生成的图，单张精美绝伦，但批量做长素材时质量瞬间崩溃， AI 生成的视频，10 秒钟惊艳，1 小时长片的管线一团乱， AI 生成的代码，单个函数完美，扔进项目跑起来一堆隐藏依赖， 共性是同一条规律： 新工具让做出来这件事的门槛降了 100 倍， 但用得动、跑得稳、能交付这件事的门槛反而升高了 10 倍。 过去做不出来是因为没人能做，现在做出来是因为工具太好用， 但优化、压缩、整合、降本的脏活累活没人帮你干，AI 工具时代真正稀缺的不再是创造力，而是生产工程能力。 所以我觉得这条树莓推真正的价值，不是教育游戏开发者怎么做 LOD， 是给所有正在被新工具喂得满嘴流油的人一个提醒： demo 级和生产级永远隔着一条河， AI 让前者变得免费， 后者还是要你自己游过去的！ Video

I spent nearly 30 years in school to earn my PhD in agronomy. Classrooms, exams, field trials, thesis writing — the whole formal grind. Then AI taught me the one lesson no professor ever fully delivered: Your future depends less on what you were taught… and way more on how well you can teach yourself in a world that changes every single month. School gave me structure, patience, and deep knowledge of soil, roots, crops, and ecosystems. But it never handed me a playbook for turning real field data into web apps. It never showed me how to use open-source projects, prompts, and code to solve actual problems for farmers, researchers, and environmental teams. So I had to learn all over again. No grades. No syllabus. No one telling me what chapter comes next. Messy prompts, broken code, late nights, failed prototypes… and slowly, real clarity. That’s why I’m here as Mai Xiaomai — bridging two worlds. The patience I learned from the soil and the speed I’m getting from AI. I share practical tools, GitHub finds, small experiments, and honest notes from someone living between agronomy and artificial intelligence. The soil is rich. If you work in agriculture, environmental science, climate, food systems, or you’re simply curious about real-world AI, come grow with me. What’s one skill you had to teach yourself that school never prepared you for? #MyPiece

x.com/i/article/205365581387…

the most underrated part of being at @aiDotEngineer singapore is realizing that the whole charade about needing to be in SF to build in AI is rapidly fading you can still go to SF sure - but it’s not like people are not building elsewhere from hangzhou to sz to sg to blore

What is GBrain? My open source project is a knowledge system, not RAG in a box. It gives agents 8 layers that work together to improve memory in a way that makes your already smart OpenClaw or Hermes Agent feel clairvoyant about who you are. Personal AI becomes possible.

A big pivot from Ken Griffin on AI: “Number one is, in the last few months, there has been a step change in the productivity of the AI toolkit. It is profoundly more powerful than it was just nine months ago. And for us at Citadel, that has allowed us to unleash a much broader array of use cases for AI. And it has been really interesting to watch, to be blunt, work that we would usually do with people with masters and PhDs in finance over the course of weeks or months being done by AI agents over the course of hours or days. These are not these are not mid-tier white collar jobs. These are like extraordinarily high skilled jobs being, I'm going to pick a word, automated by agentic AI. And I gotta tell you, I went home one Friday actually fairly depressed by this because you could just see how this was going to have such a dramatic impact on society. When you witness it in your own four walls, when you see work that used to be man years of work being done in days or weeks, it's like, wow, like that's the first time I've seen real impact in our four walls.” This echoes my own experience with agents and the conversations I am having with students, friends & clients. The toolkit has dramatically transformed and it feels like in finance, for the first time, AI is real. Video

很多推友问我商单收益怎么样？怎么接商单？ 这是我最近的累计商单收入， 赚了$4736.03，差不多3万多块人民币， 比X平台的创作者收益可香多了， 也比做闲鱼和小红书投入产出比高很多， Tutti是一个X商单平台，目前开放了一批名额， 粉丝2000以上的宝子戳下方链接加入： tutti.so/join?ref=5VN7KP 领域不限：AI、新能源、机器人、泛科技、游戏、美妆、金融、设计都可以，只要内容是「自己真正在发」的。 最后跟所有做 X 的兄弟说一句， X目前依旧是蓝海，全球最好的创作者平台， 你认真敲的每一个字，输出的每一个观点， 本来就值这个价， 别让你的努力和流量，白白浪费了 #X创作者 #副业 #变现

To prevent "programmatic use", Claude Code may now request webcam access to assure user is present when prompting

说个反直觉的真相，你在Obsidian里建的文件夹越多，你的笔记系统就越没用！ Obsidian CEO Steph Ango分享了他自己的笔记工作流，和大多数人用的完全不一样， 我们90%的人都把Obsidian用错了， 他们创始人自己几乎不用文件夹，也很少手动加标签， 整个工作流的核心只有三个东西，模板，属性，内部链接， 他创建一篇完整的会议笔记， 整个过程不到十秒钟， 大多数人拿到Obsidian第一件事就是建一堆复杂的文件夹， 然后花几个小时整理分类，最后越用越乱，干脆放弃， 喵的我就是这样做的🤣 Steph Ango说，好的笔记系统应该拥抱混乱和懒惰， 你不应该为了整理而整理， 应该先把想法快速记下来， 结构会在后续的连接中自然涌现， 他的方法很简单， 几乎每一个笔记都从模板开始， 模板自动填充所有需要的元数据， 日期，人物，主题，地点，评分， 这些属性不是静态的标签，是可以计算，可以过滤，可以生成任意视图的数据库， 你完全不需要把会议笔记放进会议文件夹， 只需要给它加上一个会议的属性， 系统会自动把所有会议整理成一个动态表格，按日期和人物排序， 再配上几个全局热键， 新建笔记，插入模板，打开每日笔记，插入内部链接， 整个过程不需要碰鼠标， 很多人花了几百个小时折腾各种插件和主题， 却从来没有真正建立起一个能长期坚持的笔记习惯， 因为高摩擦的系统最终一定会被废弃， 只有低摩擦的系统才能产生知识复利， 几年后你的Vault不会变成一堆死文件， 它会变成一个活的外部大脑， 能一键查出所有和某个人相关的会议， 所有你给过七分以上评分的书， 所有在某个城市发生过的事， 这才是第二大脑真正该有的样子！ #Obsidian #PKM #第二大脑 Video

Marking this as a moment convincing @swyx to bring @aiDotEngineer to India next year with @sanjeed_i @udayan_w Exciting times!! 🥳

Cant wait for an indie dev to accidentally put a carton of these in his game as a prop and wonder why his game runs at 2fps

微信读书Cli安装和配置教程。 1. 官方指令，复制发给Codex或Claude Code 下载 cdn.weread.qq.com/skills/wer… 安装 skill 2. 或者用 X 上网友 @eviljer 做的优化版Skill npx skills add jerlinn/jerlin-weread 3. 如果需要API key，访问微信官方这个页面获取 weread.qq.com/r/weread-skill… 4. 使用案例 直接跟AI说：“调用微信读书skill 查看被讨厌勇气的高亮划线”

Chinese students are buying GPT-5.4/5.5 and Claude API access from Xianyu/Taobao proxy sellers for almost 96-97% cheaper People are apparently burning 100M+ tokens a day for like $1 and vibecoding nonstop.

Terence Tao says the math behind today’s LLMs is actually simple. Training and running them mostly uses linear algebra, matrix multiplication, and a bit of calculus, material an undergraduate can handle. We understand how to build and operate these models. The real mystery is why they work so well on some tasks and fail on others, and why we cannot predict that in advance. We lack good rules for forecasting performance across tasks, so progress is largely empirical. A key reason is the nature of real-world data. Pure noise is well understood, perfectly structured data is well understood, but natural text sits in between, partly structured and partly random. Mathematics for that middle regime is thin, similar to how physics struggles at meso-scales between atoms and continua. Because of this gap, we can describe the mechanisms but cannot yet explain capability jumps or give reliable task-level predictions. That mismatch, simple machinery versus hard-to-predict behavior, is the core puzzle. ---- Video from 'Dr Brian Keating' YT Channel (Link in comment) Video