Thanks for your reply. My issue is that if I use a prompt that is more than just a one-liner the assistant is messing up the date completely, claiming the current date to be a random number in October 2023. I'm sorry, but I hope you understand that such an outcome fundamentally undermines my confidence in the system as a whole. I wouldn't even know where to begin addressing this, as I have far too little insight into how AI actually works—especially considering that the technology is still in its infancy. Perhaps the English-speaking side of things is more advanced. However, in order to achieve a halfway acceptable German-speaking assistant, I find myself needing to provide extensive instructions in the prompt—something I hadn’t expected. For example, if the prompt is a single line, the date seems to be output correctly. Anyways, when the prompt becomes a multiline instruction (due to the complexities of my requirements), the date handling suddenly fails. This raises not only the question of how this happens but also why a longer prompt would affect the accuracy of something as basic as date formatting. For instance, it's baffling that an assistant might confirm an appointment five times during a single call—despite clear instructions in the script to avoid redundancy—only to completely forget moments later that the appointment was ever made. Such glitches make the system impossible to recommend in good conscience. What puzzles me even more is that this doesn’t seem to be a widespread concern. Am I really the only one encountering these issues? If so, it feels like winning an unfortunate lottery. Or perhaps we’ve simply grown accustomed to AI hallucinating occasionally—accepting it as long as it "talks to the caller," while the actual content of what it says seems to take a back seat. What I also find deeply disappointing is the lack of any official response from the VAPI team regarding my concerns.