You might as well just put all those emails into a hat and pull out random ones. Or maybe categorize them first and pick from the hats your feature falls under.
Try this: ask the AI how useful it is to ask an AI for “synthetic user feedback” and it will probably even tell you why this particular task is particularly stupid for an LLM. Ok, I tried it with Haiku, you might need to follow up with a question that mentions that experience and implementation specifics matter but aren’t going to be in the context window before it will give an in-depth explanation about why this approach is a waste of resources, though using an AI to help summarize the important problem areas users want addressed can work, it just won’t be able to tell you how you did.
You might as well just put all those emails into a hat and pull out random ones. Or maybe categorize them first and pick from the hats your feature falls under.
Try this: ask the AI how useful it is to ask an AI for “synthetic user feedback” and it will probably even tell you why this particular task is particularly stupid for an LLM. Ok, I tried it with Haiku, you might need to follow up with a question that mentions that experience and implementation specifics matter but aren’t going to be in the context window before it will give an in-depth explanation about why this approach is a waste of resources, though using an AI to help summarize the important problem areas users want addressed can work, it just won’t be able to tell you how you did.