Why / the reasons

Why this exists

The surest way to understand a machine is to build a small one. Hacker News has spent years finding the tutorials that teach you how. This site puts them in order and keeps the evidence.

The gap

• HN's best build-it-from-scratch tutorials are scattered across submissions going back to 2007. The same tutorial is often rediscovered many times, and each thread has its own verdict.
• build-your-own-x is a great list, but it has no ranking or order. HN.Academy (2019) ranks online courses by how often they're mentioned. Neither tells you which tutorial to start with, which one comes next, or what HN said about it.
• Each course here is a path: steps in learning order, the HN points behind each one, and the comments that praised or warned about it.

Why it matters more now

• When compilers arrived, most programmers stopped writing assembly, and software did fine. A smaller group kept that knowledge, and it mattered more because they were fewer. Agents are likely to repeat the pattern one layer up.
• This time is harder. Compilers removed a layer; agents remove the practice: the debugging, the dead ends, writing the second version. That struggle is where judgment comes from, and judgment is what you need to review an agent's code or catch it when it's confidently wrong.
• The tutorials will mostly stay online. The habit of working through them is what may fade. This site is for the people who keep it.
• The people building AI still learn this way: Karpathy's Let's build GPT and llama3 from scratch are steps in the GPT course.
• Search results are filling up with generated tutorials. The ones people wrote by hand, tested against HN's commenters, are worth finding first.

For the record

• Most of these tutorials are one person's work on a personal site, and personal sites go away. Ghuloum's compiler paper lost its host; coding-geek.com went down. Where a host is gone, the step links an archived copy.
• The discussion is part of the record too. The site keeps the scores from all 1,889 HN submissions behind its sources and quotes the comments verbatim, with the author, year and a link, so what HN thought of a tutorial stays next to the tutorial.
• Every number comes from the public HN API, and each course page shows the date its data was pulled.

Where it came from

• On October 31, 2026, Andrej Karpathy's HackerTube gathered the videos HN recommends into one place, and a run of similar sites for books, articles and more followed. This is that idea applied to tutorials, with two differences: the steps are ordered for learning, and the criticism stays in.

How it's made

• The site was built with the help of an AI coding assistant. Point totals come from the public HN API, and a script checks that every quote appears word for word in the comment it links to, written by the person it names.
• Which sources are in, the order, the advice and the time estimates are editorial. How we count has every rule.