GCC sets a 15-line limit on AI-generated contributions

The GCC steering committee announced an AI policy and the thread is, as one commenter put it, 'the full spectrum of personalities and the hottest of takes.' The policy lands at a 15-line limit on AI-generated code contributions, with the reasoning that a human needs to be able to understand and vouch for any change. Commenters broadly called it a reasonable middle ground, noting you cannot ban LLM usage entirely and blanket prohibition has real downsides for legitimate productivity gains.

The 15-line threshold is interesting as a concrete operationalization of 'human oversight.' It is small enough that a reviewer can genuinely understand it, but it does not pretend that AI assistance never happened. The debate in the thread is less about the number and more about whether tests count, and how you actually enforce it.

This is the first major open-source project governance policy on AI contributions that sets a quantitative limit rather than a qualitative stance. Other projects will watch whether it holds.