Part 8 · Keep Learning

Going deeper (the sources)

Stanford CS336, “Language Modeling from Scratch” is the direct inspiration and the natural next step: five assignments that build a tokenizer and transformer, a Triton FlashAttention-2, a scaling-law fit, a Common Crawl pipeline, and an RL fine-tune. Reach for it when you want the implementation rather than the derivation.

The cheat sheet is every formula on one printable page.