A Reddit user proposes crowdsourcing conversations with closed-source LLMs like Claude and ChatGPT to create a training dataset for an open-source model. The community-funded approach aims to distribute a locally runnable model free of charge, eliminating reliance on proprietary APIs.