Vibe Coding 小騎士,駕馭各種模型打造稱手的工具

Taiwan
Joined June 2010
Introducing SKY-001, the first Engineering World Model. On CAD and BIM tasks, it beats GPT-6 Astra and every frontier model we tested: -> 51-125x faster - > 150-210x cheaper -> Higher IOU than all frontier models SKY-001 lets you dream a physical object before you build it. 18 months of work. Built by the team out of Maluuba's deep learning lab (acq. by @MSRResearch). The first ‘AI General Engineer’ will be the backbone for Autonomous Manufacturing.
789
2,214
604
15,437
4,008,492
Marcus retweeted
1/ Introducing the Cua Spaces SDK. Every big lab now ships a bot with its own computer: Grok Bot, Muse, dots. None let you bring your own. With Cua Spaces, now you can. Put Claude Code, Codex or any agent on compute you choose, and run jobs across all your computers at once.
152
213
90
3,101
389,012
I made an app where you can drop any PDF and you get 3D instructions. Built with Opus 5.5 + Jig Of course my first target is @IKEA so I turned their top 20 manuals into 3D. The prompt is just a PDF and "/assembly". That's it. Either I get sued or we finally get the instructions we deserve. Check it out here, you can also make your own (it's free!) ask.jig.com/#section-assembl…
56
99
23
1,232
67,812
DeepSeek V4.1 Flash 1017 just matched Opus 5.5 on long horizon browser use. 19x cheaper 👀 > 75/100, same as Opus 5.5, almost Astra level > $0.22/task vs $2.35/task > Sol it still SOTA, but it's close The new pareto frontier 🔥
71
101
27
1,498
120,043
Marcus retweeted
I was using GPT for browser use, but now I just use deepseek… it’s 8x faster, 10x cheaper, and so much more pleasant
DeepSeek V4.1 Flash 1017 just matched Opus 5.5 on long horizon browser use. 19x cheaper 👀 > 75/100, same as Opus 5.5, almost Astra level > $0.22/task vs $2.35/task > Sol it still SOTA, but it's close The new pareto frontier 🔥
34
69
4
2,113
55,883
Here is a 6 minute call within JOSIE-Voice on a MacBook Pro M4 Pro using: Qwen-3 0.6B TTS/ASR 4bit and Deepseek 4.1 Flash on ollama cloud. it does 100+ Tool calls with agentic work in the real world. No cuts or speedup's.
46
45
8
756
50,444
Marcus retweeted
Here’s a prototype of the Kids version: “How does a pump fill a tyre?” I’m experimenting with different iterations to find the simplest way to help kids understand how things work, while still keeping enough detail to make the experience engaging, interactive, and educational
47
69
12
1,023
229,174
Marcus retweeted
Love this! If I had kids, I’d love to teach them this way. I’ve been exploring learning through play too! A Claude skill for making hand-drawn, interactive everyday objects. Here’s my little washing machine :) colorfigures.vercel.app
Here’s a prototype of the Kids version: “How does a pump fill a tyre?” I’m experimenting with different iterations to find the simplest way to help kids understand how things work, while still keeping enough detail to make the experience engaging, interactive, and educational
99
568
39
5,802
360,454
兄弟们,Codex + Magpie,才是真正的 Token 自由! 我愿称 Magpie 为 AI 圈最强路由器,没有之一。 WorkBuddy AI + OpenCode,把免费的 DeepSeek V4.1 Flash 接入 Codex。 GPT-6 Astra 负责烧脑,免费模型负责干活,Magpie 负责调度。 最强智囊团 + 最强牛马,这套组合拳打下来,生产力得有多恐怖? 以前是 Token 不够用,现在是活根本干不完啊。
80
140
8
985
160,777
Marcus retweeted
magpie果真是薅羊毛神器。把能想到的账号全部添加进去,用一个路由来管理,所有设备都指向这一个地址,高强度地猛蹬了一整天,睡前一看竟还有这么多token。
magpie的这个实时路由动态做的太帅了,两个屏幕一边看着agent干活,一边看着路由的调度,感觉自己在指挥千军万马。
28
57
6
562
74,164
Marcus retweeted
SpaceXAI engineer, Lauren made all 54 of her skills public, and now you can add them to Grok Bot in one click absolutely amazing... the full list is in the image: · 3 to start: /poteto-mode runs everything else · 8 to build: arena, swarm, architect, blast-radius · 6 to review: interrogate, no-comments, verification · 8 to understand: how, why, teach, bro · 5 to write: unslop, technical-writing · 24 principles her agents read before every decision save the list, then add it to your Grok Bot
Grok Bot just got a serious brain upgrade With Claude Opus 5.5 entering the stack, one Bot could run an entire AI team that researches, writes code, uses tools, verifies results, and delivers finished work Here's how to build it in 10 steps:
40
97
5
717
67,635
For 30 years, technology has moved in one direction: turning information into intuitive interfaces Somehow, the smartest software we’ve ever built took us back to a textbox We’ve spent the last 2 years at @openuidev trying to fix that. 6 months ago we open-sourced it: 10K+ GitHub stars, 6M+ downloads ChatGPT shipped Intelligent UI and showed everyone what it looks like Today we’re releasing OpenUI 1.0 bringing it to every other agent
241
350
245
2,230
2,170,685
magpie 确实好用,直接将 workbuddy 国际版的 免费 deepseek v4.1 flash 接入 dsh 中使用,不用自己反代了 workbuddy 这个 harness 本就不适配 deepseek v4.1 flash,可以说完全是负优化的,接入 dsh 后感觉 v4.1 flash 都好用了不少了
25
24
7
208
43,815
In July we open-sourced a dynamic graph of everything a child learns in primary school. 1,590 concepts, 3,221 dependencies. Now we’re turning every concept into interactive experiments kids can play and learn from. First one, age 7: heat and water. Everything that used to be a boring textbook can now be an engaging experience, potentially tailored to every child. Kids education just got a massive upgrade.
209
836
105
8,016
425,924
Marcus retweeted
Jev过后,强烈建议大家狠狠关注这个刚发布的大模型Glimpse,成功帮助我在小王子游戏开发和维护阶段大幅提高bug排查效率,绝对的vibe coding开发验收神器!! 该模型由Elorian AI(@ElorianAI)打造,主打视觉理解与推理,这意味着其能最原生、最直接的理解图片和视频中各元素的位置、结构与空间关系,并以此来分析问题 这次我有幸提前拿到了 Glimpse 的内测机会。结合当下的场景,我第一时间就想到用它来检查小王子游戏中的各种潜在 bug:只要提供实际游玩的录屏,Glimpse 就能根据画面判断游戏设计是否合理,并为我梳理出各类潜在bug 这极大程度地解决了我在开发维护阶段「验收容易遗漏」的痛点 因为一直是我一个人测试游戏,难免会漏掉一些 bug;而且我在实践中发现,就算用 Astra、Opus 做验收审查,也总有部分细节问题难以发现 而 Glimpse 凭借超强的视觉推理能力,基本能一次捕获录屏中绝大部分细节问题,并按视频时间戳清晰罗列出来,方便我逐一核验 下面视频中,我也用酒馆章节的实机画面做了实测,效果真的惊艳! 在没有任何游戏上下文的情况下,Glimpse 梳理出的bug问题是 Astra 的 3 倍。更实用的是,点击时间戳就能在右侧直接跳到对应画面查看 Glimpse现在也开放了API,我们完全可以将其接入Codex、CC来完善开发后的验收审查流程 当然,除了 vibe coding 验收,我认为 Glimpse 还非常适合视频生成、图片生成等视觉相关的场景:借助强大的多模态理解能力,帮助规范和约束提示词,降低 AIGC 视频和图片生成的抽卡概率!从商业角度看,完全值得大家关注起来!!
Spent years watching models ace a single image and completely whiff on anything that moves or changes angle. Today we’re fixing that. Try Glimpse now, our first visual thinking model with competitive frontier performance compared to Astra on key public visual reasoning benchmarks.
52
109
6
759
54,357
Your whole home as a live 3D model on your iPhone 🏠 Tap a lamp in 3D and it turns on. Dim it, change its colour, set the AC, switch scenes. Works with Home Assistant, Hue, Apple Home and Smart Life. Synced with your Mac. Open source, on GitHub 👇
Reverse engineered my IoT devices and vibe coded my whole apartment into the menu bar. Click a lamp in the 3D model, the real one turns on. ACs too. Why stop at one AC.
27
39
3
574
31,381
Marcus retweeted
You can pretty much reverse engineer any device in your home and build a mac app around it. Applied to lights, doors, ac and even the future robot helper. Here's the skill: github.com/morluto/rea It's crazy what you can build nowadays with a skill and a few prompts.
Reverse engineered my AC, and vibe coded a lil status bar app so I don't have to find the remote every time. No first world problem will remain unsolved in 2026.
62
212
16
2,798
140,298
Marcus retweeted
Magpie 现在全面插件化了 🧩 以前社区插件只能接入新的订阅;现在 Magpie 自己的几乎每一层都向社区开放,一个 npm 包就能扩展: • 订阅 / Provider:登录、刷新、模型、请求、用量全由插件负责 • 网关中间件:改写经过 Magpie 的请求和响应 • 路由策略:自己决定请求怎么在多个账号、多个模型之间分配 • onUsage:每一条用量记录插件都能收到 • 工具:插件写的工具通过 Magpie 的 MCP 服务器,给每一个接入的 Agent 用 • Agent:让 Magpie 支持一个新的 Agent • 页面:插件在 Magpie 窗口里有自己的页面,跑在沙盒里 入口:Magpie › 插件 › 发现,每类插件各有分区;点「编写插件」直接打开文档。 文档:usemagpie.ai/docs/plugins 社区仓库:github.com/magpie-community/… 三个示例已经发到 npm:tools-calc(精确计算、单位换算、日期计算工具)、page-usage(用量一览页面)、middleware-reserve(留余量的路由策略)。 magpie plugin add @magpie-community/tools-calc 一行装好。来写插件吧 👇
66
45
8
447
53,543
Marcus retweeted
I’m calling JEV 4 times a second and telling it to find 5 menus of items. Insanely fast and cheap option to identify objects in frame.
95
131
31
1,854
178,013