I think this is the best structured class out there that teaches how to scale LLMs . Hope the 17 year old knows linear algebra. Building an intuition for the shape of the matrices is important. A lot of understanding the 'building from scratch' means understanding choices like why RoPE instead of the original frequency based positional encoding. Start with Karpathy and then go to CS 336

The views of Terrance Tao on this is a great read https://arxiv.org/pdf/2608.16753

Great project for kids interested in EECS. I wonder if a miniature version can we done at a fraction of the cost as a cool high school project.

I’m watching the Frontier Systems videos - https://cs153.stanford.edu/

The one issue I had with the CS336 course was the delivery of the RL components. I liked Lectures 5 & 6 from CME 295 better

https://cme295.stanford.edu/syllabus/

I’ve heard good things about the diffusion models class as well - CME 296. Seems like a good next step.

https://cme296.stanford.edu/syllabus/

I brought a group together to do this class using the YouTube videos and course materials available online. It is challenging but rewarding. We tackled it one lecture video per week. Started with over 30 learners and by last session we were down to 8.

The repo - https://github.com/harvard-edge/cs249r_book?tab=readme-ov-fi...

What I like most are that most libraries are lively without being distracting. There is some movement but it’s not excessive. I’ve found it a great place for focused study and deep reading.

This is a position paper.

poetic justice (pun intended)