All 20 phases of the ai-engineering-from-scratch repo, from math to agent swarms, with lesson counts and what you can do after each one

The companion to the reel. PAGE 1 — what the repo really is (523 lessons in 20 phases, MIT license, 60K+ stars on GitHub), the one command that runs lesson 1 with plain Python and no libraries, the real output we got (last line: This is literally what a neural network layer does), and honest notes: the README estimates about 342 hours, so this is a full curriculum, not a weekend. PAGES 2-4 — all 20 phases in order with lesson counts and a plain-English line on what you can do after each, grouped in five bands (foundations, perception and language, large language models, agents, production and capstone), with the reel path 01, 03, 07, 14, 16 highlighted. PAGE 5 — the AI tutor setup from the README (npx skills add, /start-learning in Claude Code or start-learning in Codex, the placement quiz that writes LEARNING.md, and the focused MCP and Agent Skills paths). We did not run the installer. Checked on GitHub on September 29, 2026. Independent guide from Hyperautomation Labs — not affiliated with the repo's author.

By submitting, you'll be subscribed to the Hyperautomation AI Report newsletter — three practical AI emails a week. Unsubscribe anytime.

Free. No spam. Unsubscribe anytime.