A 'Born Against' movement is solidifying within the high-stakes enclaves of hobbyist programming. In domains like OSDev, LangDev, and chess engine architecture, the use of Large Language Models (LLMs) is no longer seen as a productivity hack—it is increasingly branded as an intellectual default. These communities are reacting with open hostility toward AI-assisted contributions, and for good reason. As Fogus notes in a recent analysis, for groups involving EmuDev, TxtDev, and the demoscene, the process of struggling with complex systems is the primary product. Functional code is merely a byproduct; the true objective is the deep domain knowledge forged in the fire of manual debugging.
The friction arises from a fundamental clash between raw efficiency and genuine expertise. While the industry at large treats LLMs as a force multiplier, Fogus argues that in learning-focused circles, these tools act as a surrogate that robs practitioners of their craft. Early experiments with AI in these niches have already 'poisoned the well.' Shallow contributors, lacking foundational understanding, use LLMs to simulate competence, which veterans view as a fraudulent attempt to bypass the years of curiosity required to earn social capital. In an environment where respect is currency, AI-generated solutions look like counterfeit notes.
This gatekeeping is not a Luddite reflex; it is a survival mechanism for engineering integrity. When the price of an error is a total system failure—common in system-level engineering—superficial AI logic becomes a liability. The risk extends beyond individual skill atrophy to the structural pollution of unique knowledge bases with synthetic, 'hallucinated' garbage. Fogus reminds us that even seasoned experts possess no natural immunity against the subtle deceptions of LLM outputs. By barring AI, these communities ensure that the 'how' and 'why' of engineering remain central, preventing the devaluation of hard-fought knowledge in an era of cheap, automated shortcuts.