Mathematics
StartThe math foundation for understanding the transformer architecture and reading modern AI papers β intuition first.
0 / 4 chapters passed
Start βA free, interactive course Β· math β transformers
Every line of an LLM paper is built from math you met once and half remember. This course rebuilds it β vectors to attention to LoRA, one interactive lesson a day β until the equations read like code.
One new lesson drops every day:
Fig. 1. Attention scores every pair of tokens with exactly this number. Drag either arrow β watch what agreement does to it.
You build things and call model APIs, and you'd like to know what's actually happening inside. Every symbol here gets introduced before it's used.
Still in school, or it just never stuck β that's fine. This starts at βwhat is a vectorβ and assumes nothing beyond basic algebra.
You follow the announcements and skim the abstracts. The whole point of this course is that you can open the paper itself and follow it.
A new lesson comes out daily, with a short video. Everything already released stays open, so you can catch up whenever.
Lessons are interactive β drag the vectors, change the numbers, watch the formulas respond. Reading alone doesn't build intuition; poking at things does.
Pass the lesson's quiz to unlock the next one; chapter exams need 90%. It sounds strict, but it's the reason the later chapters feel manageable.
Vectors β calculus β probability β optimization β the transformer, block by block β training a small LLM β fine-tuning with LoRA and RLHF β the modern stack: MoE, FlashAttention, Mamba.
2 courses96 lessons8 chapter examsevery lesson interactive
The math foundation for understanding the transformer architecture and reading modern AI papers β intuition first.
0 / 4 chapters passed
Start βBuild a real understanding of the transformer architecture, then train and fine-tune LLMs β theory and practice.
0 / 4 chapters passed
Every expert you follow started with vectors.
Start Lesson 1 β it's free