AlpineBSD
A new project called AlpineBSD aims to combine the NetBSD kernel with an Alpine Linux–style userland while enforcing a strict “as little AI-generated code as possible” policy. Commenters debate whether such AI-free software ecosystems are principled safeguards against legal, ethical, and quality risks or impractical tribalism that will isolate maintainers from useful tools and upstream updates. The thread also explores NetBSD’s real-world hardware and userland limitations, the fragmentation of open source communities over AI use, and how future licensing and training-data questions could affect both proprietary and free software.
Project scope and goals
- AlpineBSD combines the NetBSD kernel with an Alpine‑inspired userland and s6 as init (with plans for OpenRC).
- The stated goal is to minimize AI‑generated (“tainted”) code in base and main repos, pinning some core tools (e.g., gcc, binutils, gdb, perl) to last known non‑LLM versions.
- Several commenters find the mission statement unclear: is it aimed at native hardware vs. containers, and why use this instead of plain NetBSD? Others see it primarily as an attempt to build a like‑minded community.
Reactions to the AI‑free policy
- Supporters welcome a “LLM‑free” space, compare it to Amish‑style tech abstention, and argue it’s valid harm reduction against LLM slop PRs, deskilling, environmental impact, and murky copyright.
- Critics label the stance tribal or “backwards,” arguing that:
- LLMs are already useful for research, prototyping, and bug‑finding.
- Banning any tainted code will force heavy maintenance (pinned toolchains, fewer upstream updates) and may be unsustainable as AI use spreads.
- Some differentiate between:
- Rejecting LLM‑generated contributions into one’s own project (seen as reasonable), vs.
- Refusing to consume any software that contains any LLM output (seen by some as ideological overreach).
Licensing and legal uncertainty
- A recurring concern is that LLMs trained on GPL or other non‑permissive code might produce outputs that inherit those licenses, threatening projects that ingest such code.
- NetBSD’s existing guidelines treating LLM output as “tainted” are cited as precedent for cautious policies.
- Others argue infringement is hard to prove, lawsuits are rare, and legal regimes may adapt to normalize AI training and use.
NetBSD vs Linux and userland choices
- Multiple comments debate NetBSD’s “spotty” hardware support: it often boots widely but may lack robust Wi‑Fi, graphics, power management, and laptop peripherals compared to Linux.
- Some praise Alpine’s small, simple userland and see it as more approachable than NetBSD’s default userland and pkgsrc workflow, which several describe as hard to install, configure, and upgrade.
Viability and ecosystem impact
- Many predict limited longevity for such a niche, anti‑AI BSD spin, pointing to prior Linux/BSD hybrids that faded.
- Others see value in having “pristine” human‑written codebases as long‑term training and comparison baselines amid growing AI‑generated code and increasing fragmentation of FOSS into AI stance “camps.”