Understand AI.
Design systems.
Make better decisions.
Learn AI from first principles through explanations, interactive visualizations, engineering decisions, experiments and real system challenges.
282 pages and 243 build exercises.
Tokens attend to tokens
Q · Kᵀ / √d → softmax → weighted values, computed in the page.Q · Kᵀ / √d scores normalized with softmax
Understand it. See it work. Make the decision.
AI can write more of the implementation than ever. Your advantage is understanding what the system is doing, knowing where it fails, choosing the right design, and verifying the result.
- UNDERSTAND — Build the mental model before writing code.
- EXPLORE — Change inputs, adjust parameters, and watch the system respond.
- DECIDE — Choose between real engineering trade-offs under hardware constraints.
- BUILD — Implement the parts worth understanding directly in Python.
- VERIFY — Test assumptions, find failure modes, and decide whether it ships.
The control loop of a graph-based agent. Interactive version
Seven sections, in reading order. Start wherever you need to.
Foundations
Python, systems basics, and the math the rest of the site leans on.
19 pages
02Models
Classical ML, deep learning, vision, NLP, speech, LLMs, graph ML, and RL.
86 pages
03Agents & Applications
Agents, harness engineering, and AI applied to science, healthcare, and other domains.
37 pages
04Systems & Infrastructure
Designing, shipping, and running ML systems: MLOps, databases, frameworks.
72 pages
05Safety & Evaluation
Evaluation, security, alignment, and interpretability.
22 pages
06Research & Build
Reading papers, building models from scratch, and full projects.
28 pages
07Career
Interview prep, roadmaps, and resources.
17 pages