Introducing LabCraft

Nix does something almost no other tool does: it lets you describe a machine's entire state, version it, and rebuild it exactly, on demand. The first time that clicks, it's hard to unsee. Reproducible from the ground up stops being a slogan and becomes the way you wish everything worked.

The catch is getting there. Nix splits people in two: those who take the time to understand what it really is and come out fluent, and those who never get up its steep learning curve. The documentation is terse, the answers are scattered across two decades of forum threads, blogs, and tutorials, and almost every example assumes context you don't have yet. The power is obvious from day one. But the path from "I installed it" to "I actually get this" is not. LabCraft is an attempt to build that path.

Why Nix is hard, and worth it

Nix isn't hard by accident. It touches shells, package management, build systems, and the operating system at once, and it asks you to think differently about all of them. That's exactly why it's worth learning, and why reading about it only gets you so far. How far depends on what you already bring to it. You can follow every word of a flakes explainer and still have no idea what to do when your own build won't resolve.

AI hasn't changed that, though it's tempting to think it has. A model will write you a flake in seconds, with total confidence, whether or not the expression is any good. If you can tell a sound flake from a broken one, that's a real accelerant. If you can't, you're just shipping broken systems faster. The judgment to know the difference is the part that doesn't come from a model, and it's the part the docs can't hand you either.

You learn Nix by breaking it

You get fluent by building a real machine, breaking it, and putting it back together. A rebuild that boots into an unusable generation. A flake input that won't resolve no matter what you try. A service that comes up green and still doesn't work. On LabCraft you cause those on purpose, on a real machine, and then dig yourself out. Sit in the failure once, fix it yourself, and it stops being frightening. The next time you meet it, on your own system, at 2am, your hands already know the move.

That's the whole idea. You will break things here on purpose. That's the point, and that's how the judgment gets built.

What you actually get

A real machine to break, not an exercise that checks whether you typed the expected command. You can push a Nix system until it won't boot and then bring it back, the kind of failure you'd never risk on a setup you care about, made safe to cause on purpose. Each step builds on the one before it instead of dropping you at a blank prompt, and when you finish, the check reads the actual state of your system to decide whether you got there, not a transcript of what you typed.

The labs are held to a standard too. Nix moves fast, and when a tutorial quietly goes out of date, the real cost isn't wasted time. It's not being able to tell whether the failure is your mistake or the tutorial's. That's why the labs get tested against the versions they teach and are meant to break on our side before they break on yours.

Why Nix first

Nix is as powerful as it is hard: a new language wrapped around an operating system rebuilt from the ground up, and the tool people most often want to learn and most often give up on. If LabCraft can teach it well, it earns the right to teach the rest. So it starts here, and starts deep. Where it goes next depends on what people want.

LabCraft isn't open to everyone yet, but it won't be long. If you want in on the first wave, put your name down below and we'll reach out as seats open.