This seems only possible to the extent that you can cut human content creators out of the loop. At some point, the models will need to be trained on something.

This seems only possible to the extent that you can cut human content creators out of the loop. At some point, the models will need to be trained on something. That something is the words of actual intelligent people who are saying things that existing models aren’t intelligent enough to write. So ultimately unless you’re content with being a consumer of the LLM regurgitation content cesspool, then you’ll just need to find those trusted intelligent people and subscribe to their content, probably via something extraordinarily simple such as RSS.