Skip to content

Curriculum

The 10 canonical tracks from pracha.me/curriculum. Every subject has the same shape: concepts/, authors/, arcs/, builds/. See the structure spec for the v2 design contract.

Current state — autonomously maintained

71 substantive concept pages across all 10 tracks · 10 newly auto-seeded stubs awaiting the next cycle · 6 retrospective entries on the backlog. The system writes, reviews, and expands itself — every cycle a retrospective agent reflects on what regressed and seeds the next sprint with the gaps it found. See the latest retro for what the agents themselves identified as weak this round.

The 10 tracks

# Track Overview
01 AI Overview
02 Generative Modeling Overview
03 Representation Learning Overview
04 Neural Networks & Deep Learning Overview
05 Statistical & Probabilistic ML Overview
06 Reinforcement Learning Overview
07 Attention, Memory, Reasoning, Continual Overview
08 Causal & Statistical Inference Overview
09 Algorithms & Systems for AI Overview
10 Complexity, Cognition & Natural Intelligence Overview

Each row links to a subject overview where you'll find the concepts, authors, arcs, and builds for that track.

The four artifact types

Type What it is Where it lives
Concept Self-contained Olah/Distill-grade explainer core/<track>/concepts/
Author Person-anchored reading guide (Pearl, Karpathy, Olah…) core/<track>/authors/
Arc Learning path through N concepts in order core/<track>/arcs/
Build MVB recipe — runnable, persona-tagged core/<track>/builds/

Pages converge: a concept lives inside an arc, links to its key authors, and ends with a build you can ship.