Transformers
Self-attention
Every word in a sentence quietly looks at every other word and asks 'how much do you matter to me?'
Brick · Coming soon
Concepts Library · Free & growing
One brick at a time.
A free, ever-growing library of single-concept explainers. Each entry: the intuition, the math that matters, one tiny mini-lab, and a link to the course that builds the full castle around it.
Library opens with the launch · Preview below
Sampling
Temperature
One number that decides whether your LLM is precise or creative. We'll show you what's happening inside the softmax.
Brick · Coming soon
Tokens
BPE tokenization
Why your model sees 'unbelievable' as three pieces; and why that matters for cost, accuracy, and prompts.
Brick · Coming soon
Representations
Embeddings
Words as vectors. Vectors as geometry. Geometry as meaning. The hidden grammar of every modern AI system.
Brick · Coming soon
Systems
RAG (retrieval-augmented generation)
How to give an LLM a memory it doesn't have; without hallucinations, and with citations.
Brick · Coming soon
Tools
Function calling
The bridge from chatbots to actual systems. How LLMs decide when to call a tool, with what arguments.
Brick · Coming soon
Evaluation
LLM-as-Judge
Using a strong model to grade a weaker one; when it works, when it lies, and how to keep it honest.
Brick · Coming soon
Security
Prompt injection
The number-one security failure in LLM apps. We'll show you the attack and the defenses that actually work.
Brick · Coming soon
Retrieval
HNSW vector index
How modern vector databases find a needle in a billion-vector haystack; in milliseconds.
Brick · Coming soon
Want a single concept added to the library next?
Tell the lab