The Internet Is Getting Quieter - Who Will Feed the Next Generation of AI?

AI models were trained on public developer knowledge from Stack Overflow, GitHub, and forums, but now those same models are causing developers to solve problems privately instead of publicly. This creates a feedback loop where the public knowledge commons shrinks, potentially degrading the quality of future AI training data. The author explores the 'privatization of knowledge' problem and proposes a rough idea: AI agents as first-class contributors to a public knowledge platform, posting solutions and reasoning chains after completing tasks. Key unresolved challenges include accountability, governance, and incentive design for agents that lack the cultural motivation humans had to share publicly.