The Canon, made runnable
Foundational textbooks taught this field. But you can’t run a book. So we turn each core technique into a single runnable notebook, live in your browser and built on our own data: nothing to install, nothing to pay. Every lab links back to the exact chapter, so the moment an idea clicks you can go deeper. Learn the concept here; read the masters for the theory.
How the whole Canon fits together
Nineteen books, one shelf. Here is how they fit together. You start with probability, stand on the foundations, build the core models, then specialize and layer inference and the modern frontier on top. The same handful of ideas reappears in every book, so the shelf is friendlier than it looks.
1 PROB: probability"] --> FOUND["Foundations
2 LA: linear algebra
3 ALG: algorithms"] FOUND --> MODELS["Core models
4 ISL, 5 ESL,
6 DL, 7 PRML"] MODELS --> SPEC["Specialize
8 RL, 9 FPP,
10 IML, 11 BAN"] MODELS --> ENG["Engines and limits
12 CVX: optimization
13 IT: information"] MODELS --> INF["Inference
14 AOS, 15 CI,
16 BDA, 17 CASI"] MODELS --> FRONT["Modern frontier
18 PML
19 LLM: agents"] INF --> FRONT
The nineteen books, grouped by where they fit.
Start here. 1 PROB (probability) is the on-ramp every other book assumes.
Foundations. 2 LA (linear algebra) is the ground it stands on, and 3 ALG (algorithms) is how it runs.
Core models. 4 ISL builds them, 5 ESL explains why they work, 6 DL unfolds the neural net, and 7 PRML puts a probability on everything.
Specialize. 8 RL (reinforcement learning), 9 FPP (forecasting), 10 IML (interpretability), and 11 BAN (bandits).
Engines and limits. 12 CVX (convex optimization) solves the fit, and 13 IT (information theory) draws the outer limits.
Inference. 14 AOS shows it is real, 15 CI asks why, 16 BDA is the Bayesian workflow, and 17 CASI is computer-age inference.
Modern frontier. 18 PML is where deep learning and Bayesian uncertainty meet, and 19 LLM builds language models and agents from scratch.
Loading the canon…