Hobby programming communities such as OSDev, LangDev, TxtDev, EmuDev, RLDev, the demoscene, and code golfers have expressed strong opposition to the use of large language models (LLMs), according to a blog post published on August 4. These communities emphasize the value of mastering difficult technical fields through hands-on experience rather than relying on LLM assistance, viewing the process itself as the core product.
The blog post by Fogus highlights a GitHub thread related to chess engine development that sparked reflection on why these niche groups are hostile toward LLMs. Early engagement with LLMs in some communities was initially positive but quickly soured due to a perceived lack of deep understanding by LLM practitioners and the emergence of vitriolic reactions. The sentiment is that using LLMs bypasses the hard-fought knowledge and skill acquisition that define these communities.
This resistance contrasts with broader tech industry trends where LLMs are increasingly integrated into development workflows. The hobbyist groups prioritize the learning journey and craftsmanship over simply producing working code, which they see as a 'nice-to-have' rather than the main goal. Their stance underscores a cultural divide between professional and hobbyist programming approaches to AI tools.
The blog post serves as a detailed reflection on the cultural dynamics shaping LLM adoption in specialized programming circles. It was published on August 4, 2026, on blog.fogus.me, providing a rare insider perspective on the friction between AI advancements and traditional programming communities.