Kaien Yang

Check out the motorized couch we built at @Stanford moonshot club!! @karpathy, any suggestions for FSD? We’d love to give you a ride in it too 😉 It currently also has a semi-autonomous feature that follows a person wearing a particular shirt 🚀

@schen_7 thanks Sarah!! do lmk if you're ever in NYC :D

@kaien_yang Congrats!

Palm Drive to Park Ave!! 🌲➡️🏙️ Just graduated this quarter from Stanford with my Math BS and CS MS. Thrilled to join Citadel as a QR in NYC, specializing in fixed income and macro. Stanford was all love, now it's time to serve 🚀

It’s always Gemini season 💻

@deedydas @ChrSzegedy would love to see how Gauss @mathematics_inc performs on IMO25!!

🚨 A startup by the founder of Robinhood just dropped how they got a gold medal at the IMO 2025, the elite math contest. 4 teams have done this. Harmonic Aristotle, unlike OpenAI and DeepMind, uses formal Lean-based search methods and a geometry solver like Bytedance SeedProver.

Fully packed first class! Excited to TA Stanford CS329A, Self-Improving AI Agents, with @achowdhery @Azaliamirh @boson2photon. Here's to an amazing quarter!

After our team's presentation on AI-driven TPU circuit design, Jeff chatted w/ me to congratulate us, pick my intuition about key results, and brainstorm scaling to larger problems. I was a Student Researcher. He's the head of Google AI. That kind of investment and generosity is a big part of what makes him a legend 🐐

New webpage! Thanks Gemini 🫡 kaienyang.com

@jxmnop @OwenDugan your turn! :)

i was surprised to learn how much of modern AI was developed by physicists: - scaling laws - alphafold - RLHF (dario got his phd in biophysics) - diffusion models - normalizing flows - VAEs - neural tangent kernel - hopfield networks - information bottleneck theory - neural style transfer - imagenet (apparently feifei has a bachelors in physics) - chinchilla optimality

Behind every great chip is a great mind. Huge shoutout to the best internship mentor! Congrats on the Ironwood TPU and a legacy etched in silicon :)

@LawtonSkaling @jclin22009 @cathyzbn @ScottHickmann Lawton the MechE Maestro strikes again!! Absolutely incredible 🙌

We won the best hardware hack at TreeHacks - for the 2nd year in a row! Shoutout to my insanely cracked team: @jclin22009 doing the front end, @cathyzbn on document processing and formatting, and @ScottHickmann doing printer controls. We're going to develop our documentation in the coming weeks, in the hopes that anyone with a 3D printer and a need for braille can affordably print it at a fraction of the cost!

For TreeHacks, we built Braille Bot. It transforms any document—text, images, or websites—into printed braille at a fraction of the cost. It can be easily installed on an existing 3D printer for just $15, significantly cheaper than anything currently available.

New paper alert!🚨 What do LLM reasoning, diffusions, & jailbreaks have in common? 🤔 All exhibit critical windows📈--a sudden formation of distinctive features during sampling, e.g. correctness or toxicity. We present a unifying theory of critical windows for diffusion & LLMs.

@AkshGarg03 @karpathy homie maximized the loss

stanford midterms be like #cs231n @karpathy

When the GOAT comes to The Farm @rogerfederer 🐐🎾

Thank you TreeHacks for awarding us best hardware for our book digitizing machine! This was the most fun and rewarding hackathon. This wouldn’t have been possible without my teammates Lawton, Jason, and Kaien, so many thanks to them as well! #treehacks #books #hackathon

Shoutout @ScottHickmann @LawtonSkaling @KaienYang for being so awesome to build with, and @hackwithtrees for an incredible experience! :)

TreeHacks, 26 hours later

@tonyzzhao @Stanford Congrats on the amazing work @tonyzzhao! Thanks for letting me experience the fun of beta testing ALOHA :)

Introducing ALOHA 🏖: 𝐀 𝐋ow-cost 𝐎pen-source 𝐇𝐀rdware System for Bimanual Teleoperation After 8 months iterating @stanford and 2 months working with beta users, we are finally ready to release it! Here is what ALOHA is capable of:

How can we design architectures that can process or transform neural networks? We introduce a framework of *Neural Functionals* that process NN weights while respecting their permutation symmetries. Paper: arxiv.org/abs/2302.14040 1/9