Learn Nix Faster with an AI-Powered Interactive Sandbox
The idea came from a frustration I had with my own Nix learning journey. You read the Flake docs, you understand outputDrv, you can cobble together a shell.nix — and then you try something real and nothing works. The error messages are opaque, the state is invisible, and there is no one to ask except a forum thread from 2019. LabCraft puts you inside a Linux VM with a monitored shell so the mentor can see exactly where your command sequence diverges from the intended path and offer targeted guidance.
Right now I am building out a ~48-part Nix and NixOS course. Part 1 is live for all users today, covering foundational concepts with interactive exercises. The roadmap spans flakes, module system configuration, service management, and deployment workflows. Each part pairs reading material with a playground task where the AI mentor observes your commands in real time and steps in when you are stuck rather than just letting you wander.
What makes this different from a typical tutorial site is the feedback loop. The curriculum, the playground instrumentation, and the mentor logic are all iterated on together based on what learners actually do — not what they say they understand. If a particular exercise causes confusion across multiple users, the team can adjust the task, the hints, or the explanation without rebuilding the whole thing.
I am looking for feedback on the experience itself. How clear are the progression steps? Does the mentor help or get in the way? What would make you keep coming back after the first session? If you have spent time with Nix or NixOS — whether you are a beginner wrestling with your first flake or an experienced user who thinks the whole ecosystem is worth learning — I would love your take on where LabCraft can improve.