Post #1282724

@simulo This is difficult to put into words, but re: Synthetic users: On the face of it, it seems like such a disconnect, resting on a "the map is not the territory" fallacy. I'm guessing an LLM will have been trained on transcripts/reactions etc. to generate a plausible reaction to a product test. But the transcript is not the user - it's the result of the user meeting something and trying to formulate their thoughts into words. How can that be a sufficient data foundation for synthetic users?