When was the last time you used a LLM? This is not an issue with current models.