Ryan Lee

Building a Better World with AI …

[Ryan Lee @ryanleexai]

0/30 of Stanford CS336 Language Modeling from Scratch.

[Chris Lattner @clattner_llvm]

I'm excited to share that Qualcomm is acquiring Modular: this will accelerate our path to unifying accelerated compute with an open platform. This will also mark a new era in open software development for Qualcomm. 👇

[Ryan Lee @ryanleexai]

Ultrasound真的可以做到比MRI更清晰吗？有点怀疑。

[Midjourney @midjourney]

A technical dive inside our new "Midjourney Scanner"

[Ryan Lee @ryanleexai]

cmu的database system 和国内大学的数据库课程区别大到可以几乎说是两个完全不同的东西了。

另外Andy Pavlo讲课真的很有激情。

[Eremus @Eremus426531]

最近在看cmu的数据库课程，才发现他们上的课跟我学校里学的数据库差别太大了，基本cmu两三节课就把学校一学期的数据库内容讲了一大半，而且��讲的比学校好

[Ryan Lee @ryanleexai]

好玩儿。

[Chris Thomas @chrisvtom]

Walnut is my new favourite home robot

A tiny quadruped with sub 100$ BOM. It has all the compute it needs for a natural gait, high quality voice control, and to feel the world around it.

@victoroldensand and I built this little guy in three weeks to introduce a new type of social robot, one that doesn’t talk too much, and can explore and interact with the world on its own.

Join our discord to learn more about it. https://t.co/2tdZ5wbMUQ

[Ryan Lee @ryanleexai]

管理是对人的Harness.

[Ryan Lee @ryanleexai]

@YuLin807 google的技术可以学，google的产品用不得。

[Ryan Lee @ryanleexai]

夏季昼夜温差大就是这么神奇，伦敦/巴黎/洛杉矶夏天白天温度可以冲到30度以上，但是夜间温度能降到10几度。

不仅蚊子少，自己种菜的话，各种病虫害都比国内要少。

[花果山大圣 @shengxj1]

最近北京开始有蚊子了 😭
英国缺点无数个 但是没蚊子的夏天确实是��优点

[Ryan Lee @ryanleexai]

⁠ ...and with a bunch of ugly, hardcoded if/else branches inside.

[Arpit Bhayani @arpit_bhayani]

Fuck! AI agent is just an expensive while loop.

[Ryan Lee @ryanleexai]

千万别让国内公司的CEO看到了。

[CuiMao @CuiMao]

@MaxForAI meta那���多人，一个个裁过来太慢了，不如都拉去做标注，走的比裁的快。

[Ryan Lee @ryanleexai]

我们CUDA Boy的岗位也要保不住了吗？

[How To Prompt @HowToPrompt__]

ByteDance has published a paper that should make every NVIDIA investor sweat.

They trained an AI that writes CUDA better than humans experts.

They call it CUDA Agent.

And it completely rewrites the economics of AI hardware.

They built a massive agentic reinforcement learning loop. The AI writes a kernel, compiles it, profiles the hardware, analyzes the bottlenecks, and rewrites the code until it's flawless.

It learned how to optimize memory access patterns and hardware tiling strategies that traditional compilers miss.

The results are staggering.

On the industry-standard KernelBench, CUDA Agent completely destroyed traditional compilers.

It delivered code that runs up to 3.2x faster than PyTorch's native execution.

On the hardest, most complex models, it beat the strongest proprietary models in the world—including Claude Opus 4.5 and Gemini 3 Pro, by 40%.

It didn't just match human experts. It started discovering optimizations that static compilers literally cannot see.

Here is why this is a massive threat to NVIDIA.

NVIDIA's dominance relies on the fact that CUDA is incredibly hard to master. Developers get locked in because optimizing code for other chips is too painful.

But if an AI agent can autonomously generate hyper-optimized hardware kernels...

You don't need a team of $500k a year CUDA engineers to build world-class infrastructure.

And if an AI can autonomously master CUDA, it can master AMD's ROCm. Or custom silicon.

The impenetrable software wall protecting NVIDIA's monopoly just got breached by a reinforcement learning loop.

If anyone can automatically squeeze maximum performance out of any chip...

Hardware becomes a commodity.

[Ryan Lee @ryanleexai]

@turingou 农业现场，智能眼镜这种设备会不会更方便？

[Ryan Lee @ryanleexai]

Nvidia 的RTX Spark芯片看上去挺靠谱，就是Windows 操作系统太烂了。

Apple的M系列芯片挺好，就是AI开发者生态太差了，MLX 框架太封闭了，性能调试完全靠猜。

[NVIDIA @nvidia]

At #MSBuild, @Microsoft and NVIDIA unveiled an end-to-end stack for developers scaling agentic AI, from Windows devices to cloud to local.

One unified stack. Every layer accelerated.

Highlights include:

💻 RTX Spark: Windows PCs purpose-built for personal agents

💡 NVIDIA open models on Microsoft Foundry, with Anthropic’s Claude models running natively on NVIDIA GB300 Blackwell Ultra systems on Azure

🚀 NVIDIA-accelerated Microsoft Fabric Data Warehouse with up to 6x faster SQL execution

🦞 NVIDIA OpenShell in GitHub Copilot: secure-by-design agent runtime

📣 Fairwater Wisconsin AI factory is live and validated for the NVIDIA Vera Rubin platform

🔗 Read the full blog: https://t.co/n2buTyByjD

[Ryan Lee @ryanleexai]

这个对于行业外的��来说，很反常识，一般默认用户会在搜索框里面输入关键词进行搜索。

但是其实从一开始，搜索引擎的搜索框里，很多时候收到的是提问。

比如：夏天皮肤油腻怎么办？父母总是吵架怎么办？身材不好，怎么挑衣服？等等。

这也是搜索引擎公司，比如Google/Baidu很早就做机器学习和AI的原因。

[AB Kuai.Dong @_FORAB]

这篇 Reddit 营销版块上的同行讨论，或许预示着为何传统搜索正在被淘汰。

大量有需求的用户，跑到店铺页面搜索时，并不是直接查商品名，而是像朋友一样询问哪款适合夏天用？有什么产品更好？

但传统的搜索，不会反馈结果，这导致有很多客户流失。

[Ryan Lee @ryanleexai]

@MinLiBuilds @ZenMuxAI 手势识别最后选了哪个开源库？

[Ryan Lee @ryanleexai]

Based on my personal growth and observations of those around me, Andrej Karpathy's perspective is spot on: one of the greatest significances of AI lies in its ability to affordably provide personalized tutoring for everyone, thereby elevating the intellectual level of society as a whole.

[Dwarkesh Patel @dwarkesh_sp]

Over the last 200 years, we've automated away a lot of hard physical labour. But people still go to the gym.

Indeed, many people today are more physically capable than people in the past. We can train systematically for whatever physical goal we want, and it's more fun than hard labour on a pre-modern farm.

@karpathy's hope is that, in the future, the same will be true of learning.

AI tutoring that's tailored to each person will make learning easy, and more people will want to do it. We will be able to go much further than our ancestors.

[Ryan Lee @ryanleexai]

AI 投资确实大，连Google都要融资了。

[华尔街日报中文网 @ChineseWSJ]

谷歌母公司Alphabet周一宣布，计划进行800亿美元股权融资，以为其在人工智能竞赛中的巨额资本支出筹集资金。该公司表示，已达成协议向伯克希尔哈撒韦公司出售100亿美元的股票。https://t.co/ZpctsMXPGI

[Ryan Lee @ryanleexai]

HBO 出品的记录片 狂野太空(Wild Wild Space)不错。

第一次知道除了SpaceX之外，还有这么多航天创业公司。

Space is the next frontier。

[Ryan Lee @ryanleexai]

@aehyok 回来发现大家都涨粉不少，都成大V了😃

[Ryan Lee @ryanleexai]

上次张小珺和谢赛宁的访谈里面，也提到Google 的TPU难用，他带的学生花费了大量时间在解决TPU的各种使用问题上。

Google的很多云产品也是，经常号称性能吊打竞品，但是��须按照他们提供的各种奇葩接口和编程范式来使用。

结果性能有没有提升不知道，开发成本倒是陡增。

[Max For AI @MaxForAI]

Midjourney的创始人David表示TPU真难用啊。

使用谷歌的TPU来训练模型，可能让他们的研究落后了大约一年。

如果他们从一开始就完全使用 Nvidia 的方案，情况本可以好得多，不太理想！

要是能回到过去，他会从一开始就尽量只用 Nvidia卡。

但他反驳了他们被谷歌骗了说法：
我不认为有人欺骗了我们，可能是我们自己欺骗了自己，但我们确实在早期使用它们时获得了很多好处。

无论如何，我们许多最成功的模型都是在这些芯片上制作的，只是总体上可能拖慢了我们的研究进度。

冷知识：Midjourney这个名字来自于一个中国的古代谚语