Comment by pc86

"You can lessen your dependence on a specific LLM implementation by not using LLMs" is certainly a take but it doesn't really address the root issue of models getting nerfed to save resources after they've gained wide adoption.