A growing divide exists between AI hype and software engineering reality, as prominent open source projects increasingly ban LLM-assisted contributions. Rakshit Yadav's review of 120 projects found 37 enforce total AI bans, including GCC, QEMU, SDL, Gentoo, Zig, and Ghostty. Platforms like Codeberg, Sourcehut, and Flathub prohibit AI-generated code, documentation, and bug reports. The Linux kernel permits AI use only with attribution. Debian developers are currently voting on a proposal to ban AI across code, docs, translations, and bug reports.

Maintainers argue LLM output appears professional but often contains subtle flaws that waste expert review time. Senior developers face a flood of AI-generated submissions from contributors lacking the expertise to verify correctness, creating information asymmetry. As one observer noted, humans trust confident, eloquent output but are poor at fact-checking. Projects requiring human-in-the-loop policies still demand submitters filter all LLM output before human exposure.

The irony is stark: AI datacenters run on Linux/Debian, and models train on open source codebases, yet maintainers reject AI contributions to protect quality and human efficiency. The debate centers on whether LLMs genuinely increase productivity or merely generate plausible-looking "slop" that degrades software reliability.