May 23, 2026

Math, models, and margin meltdowns

AI Engineering from Scratch

Build your own AI the hard way — if the internet can stop arguing about the website

TLDR: A free GitHub course wants to teach people to build AI step by step from basic math to advanced projects, with hundreds of lessons and no paywall. The comments were split between hype and total distrust, with people debating whether it’s a serious learning tool or just more AI-written fluff with bad mobile design.

A giant new GitHub course called AI Engineering from Scratch is promising the full boot-camp fantasy: 435 lessons, 20 phases, no paywall, no sign-up, and no magic shortcuts. The pitch is simple enough for non-experts: don’t just press buttons in fancy software, actually learn how the machine works from the math upward, then build the parts yourself before touching the popular tools. In theory, it’s a dream for anyone tired of shallow tutorials and flashy demos they don’t really understand.

But the real show? The comments turned into a mini civil war. One camp was instantly suspicious, with people side-eyeing the slick presentation and muttering about “vibe-generated” courses — internet code for something that feels assembled by artificial intelligence instead of carefully crafted by humans. Another commenter went straight for the jugular, calling the project “AI generated garbage,” while someone else complained that modern AI docs are all written in the same bloated, repetitive robot voice. Ouch.

Then came the wonderfully petty subplot: the website margins drama. Yes, while the course promises to teach the future of artificial intelligence, one user was busy roasting the layout for looking bad on small screens. That somehow became the perfect symbol of the whole debate: if you’re teaching people to build smart machines, shouldn’t your own page at least fit on a phone? Still, not everyone was cynical — a few commenters said pairing this course with AI tutoring tools could be an insanely effective way to learn. So the mood is clear: half the crowd sees a bold, open-source masterclass, and the other half sees robot-written homework wearing a GitHub hoodie.

Key Points

  • The article presents an AI engineering curriculum made up of 20 phases and 435 lessons.
  • It teaches algorithms from raw mathematics before introducing frameworks such as PyTorch.
  • The curriculum covers topics ranging from linear algebra to autonomous swarms.
  • Each lesson follows a fixed process: understand the problem, derive the math, write code, run tests, and preserve the resulting artifact.
  • The curriculum is free, open source, hosted on GitHub, and includes runnable code in Python, TypeScript, Rust, or Julia.

Hottest takes

"I don't trust a vibe-generated course" — meeton
"all the docs around ai tooling are also written by ai" — ulcer
"AI generated garbage" — jatins
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.