I was recommended Stanford CS336 (Spring 2026) as a way to learn the fundamentals behind LLMs and get some hands-on experience. I’ve finished the lectures and am now working through the assignments. This post is where I keep my notes to organize my understanding and for future reference.

This class reminds me of a joke from college (year of 2020):

Q: “What’s the most important class a CS student should take?”
A: “Intro to Deep Learning. Not enough to become a deep learning researcher, good enough to know who’s bluffing.

Building a Transformer - Pre-training

Reinforcement Learning - Post-training

Stanford CS336: Language Modeling from Scratch - What I've learned

Building a Transformer - Pre-training

Reinforcement Learning - Post-training

Inference and Evaluation

Scaling Laws

System Parallelism

Data

Multimodal