Linux From Scratch Is Back on the Front Page — and the "Cookbook vs. Principles" Fight Is Really About Learning in the AI Era
A plain link to linuxfromscratch.org pulled 316 points, timed to the project's LFS 13.1 release ("Bruce Dubbs - 2026/09/01: The Linux From Scratch community announces the release of LFS Version 13.1"). LFS is the 25-year-old book that walks you through building a complete Linux system from source. The discussion immediately became a referendum on whether its pedagogy works — and the dissent is the most-upvoted take:
"I suppose I hoped enduring the pain of waiting for the packages to compile would somehow make me smarter. It didn't. I didn't learn much. It was like following a baking recipe." — rando103747202
Their constructive complaint is sharper than the usual "it's tedious": the ordering is backwards. "You start by building a cross compilation toolchain but… why? And you add the bootloader dead last, even though it's the first thing you see when you boot a Linux machine?" They propose discovery-driven construction — get a minimal boot working first, then add components as you feel the need for them. sixtyj concurs: "LFS is known but it is really more a cookbook than explanatory principles," pointing to OSTEP and Robert Love's kernel book as the actual principle-teachers.
The defense: it's muscle memory, and the AI angle
The defenders' case is that the recipe is the point — you come out the other side able to do things (bootstrap compilers for dead platforms, diagnose deep build failures, build your own distro) that degaart lists in detail. And smitty1e lands the argument's real stakes:
"LFS is about the muscle memory; the documentation; the cohesiveness. I always recommend it to people and consider it foundational to my understanding of How Stuff Works. Whether that amounts to a hill of beans in the AI world remains to be seen." — smitty1e
That AI question cuts both ways in the thread. kaashmonee admits the new tools invert the pedagogy: "But with AI, i can't help but admit i find myself indulging in building things backwards, now that i can try a couple different approaches simultaneously for it" — i.e., agentic tooling finally makes the discovery-first approach rando103747202 asked for affordable. Meanwhile binarin makes the strongest practical case for the old way: he learned most from LFS "when I had no internet," with mismatched Slackware/RedHat CDs — "Since then I can build anything and anywhere, even if it's not cooperating." That's exactly the capability that atrophies when every failure is one prompt away from a fix. Alternatives surfaced: Oasis ("Everything being statically compiled makes it less magic" — myaccountonhn), stagex.tools (LFS-in-containers, LLVM-first), and KISS.
Why it matters
This is the same argument the community is having about every form of rote-vs-principled skill in the agentic era (see Goodhart's Law Hits the Professions: "AI is breaking our proxies for expertise"): LFS's compile-and-copy ritual either builds durable systems intuition or wastes a weekend, and the thread splits cleanly by generation of experience. The telling detail is that the critics' proposed fix — build backwards from a booting system, add complexity only when felt — is now trivially scriptable by an agent, which may finally make the good pedagogy cheap.