Born Against, or why hobby programming communities are against LLM usage
Hobby programming communities are increasingly clashing with large language model (LLM)–assisted coding, arguing that these tools undermine the core appeal of the craft: slowly mastering hard problems, understanding code deeply, and sharing that expertise with peers. Critics also point to practical harms, from low-quality “AI slop” and plagiarism risks to a flood of half-baked projects and documentation that make community spaces harder to maintain. Supporters counter that LLMs are just another power tool, enabling people with limited time or experience to explore ideas, automate tedious work, and build software they otherwise never could, suggesting the field may split into “handcrafted code” spaces and LLM-enabled ones.
Hobby vs. Outcome-Oriented Programming
- Many distinguish between programming as a hobby (enjoying the process) vs. programming to get software built (caring mainly about the result).
- For “process people,” LLMs feel like using a car for a running hobby or a computer to solve a sudoku: it removes the point.
- Others say they’ve always been outcome-focused and LLMs finally let them ship long‑desired tools and experiments within limited time.
Role of LLMs as Tools
- Some hobbyists happily use LLMs as power tools: boilerplate generation, configs, documentation, refactors, test harnesses, quick PoCs.
- Others use LLMs only for tedious or low‑learning tasks; they still want to hand‑write core logic to preserve the learning/joy.
- Several describe workflows where they architect, interpret results, and polish, while LLMs explore search spaces or write draft code.
Community Norms, Gatekeeping, and Fairness
- Niche communities (OSDev, emulators, chess engines, IF, etc.) often value mastery itself; working code is secondary to understanding.
- In those spaces, LLM‑generated code is seen as “cheating,” akin to using engines in chess or CNC in a hand‑tool contest.
- Some defend this as legitimate “rules of the game” rather than toxic gatekeeping; others call it status protection and exclusionary.
Code Quality, Maintenance, and “Slop”
- Many complain that LLM‑assisted novices produce harder‑to‑review code with subtle bugs, increasing technical debt and cleanup work.
- “Vibe coding” and agentic systems are criticized for creating impressive‑looking but fragile, low‑understanding code and documentation.
- Others report genuine productivity gains, especially for small personal tools, and argue good engineering still matters on top of LLMs.
IP, Licensing, and Plagiarism Concerns
- Heated debate over using LLMs on GPL/AGPL or other engines: is LLM‑rewritten code a derivative work or just “lifting ideas”?
- Some argue algorithms and layouts aren’t copyrightable; others stress that license‑laundering via LLMs is legally risky and morally “a dick move.”
Status, Learning, and Broader AI Effects
- Several note fear of skill devaluation: years of hard‑won expertise vs. a teenager plus an LLM.
- Others highlight reduced human‑to‑human help, rise of AI‑generated spam, and mistrust when people suddenly have flawless AI‑polished language.
- Experiences vary: some neurodivergent users find LLMs empowering; others find them distracting and harmful to deep focus.