Hysen Labsai-engineering-from-scratch
Learn it. Build it. Ship it for others.
AI Engineering From Scratch: build it instead of watching it
A free, MIT licensed curriculum with 503 lessons across 20 phases that asks students to ship a reusable artifact in every lesson, from prompts to MCP servers.
The build first pitch
The tagline on the GitHub description is learn it, build it, ship it for others. The README gives the curriculum a striking scale: 503 lessons, 20 phases, and around 320 hours across Python, TypeScript, Rust, and Julia. It also cites a motivating statistic, claiming 84 percent of students already use AI tools while only 18 percent feel prepared to use them professionally. Closing that gap is the stated purpose of the course.
Every lesson leaves something behind
The design rule is simple: every lesson ships a reusable artifact, which can be a prompt, a skill, an agent, or an MCP server. The curriculum is described as free, open source, and MIT licensed, with the idea that you do not just learn AI but build it end to end by hand. That emphasis on artifacts is what separates the course from a typical list of reading materials.
How a learner starts
Getting started begins with a ten question placement quiz that maps existing knowledge to a starting phase and saves a personalized study plan. From there a learn command teaches one lesson per session, covering concept, math, code, and quiz. A course guide jumps to the exact lesson covering whatever the learner is stuck on. The structure suggests the course was designed around the reality that people do not study in a straight line.
Agent friendly and translated
The curriculum works with coding agents that read a SKILL.md directory, listing Claude Code, Cursor, Codex, OpenClaw, and Hermes, with an installer asking which agents to set up. Users without an agent can read the same lessons on the project website. The content is available in multiple languages through translated landing pages, with English described as canonical and lesson pages machine translated on a translations branch. A usage stat line cites a large number of readers and page views in the last 30 days.
Editorial conclusion
The curriculum's wager is that shipping small reusable things, prompts and skills and agents, teaches better than watching lectures. The placement quiz and agent integration suggest it targets people already living inside AI tools.
Community notes