You Cannot Prompt Your Way to Something Original
Ask a model a question in one sentence and it hands you the most average possible answer. Not a bad answer. An average one, assembled from everything ever written on the subject. That is not a defect you can prompt around, it is what the machine is built to do, and “be more creative” is not an instruction that changes it.
Most people stop there and conclude the thing cannot be creative. A lot of them settle on calling it fancy autocomplete and move on.
What actually works is stranger and more useful. You go do the creative work yourself, the human way, with a pen if it comes to that. You pay attention to what you actually did, which is almost never one move. It is four or five, in an order, with a decision in the middle of it. Then you hand the model that process instead of the question.
That is the whole idea here. The technique is the thing you are teaching it, and you cannot teach a process you have not run yourself.
So every post in this series comes in two halves. First a human creative technique, the kind of thing people were doing on paper decades before any of this existed. Forcing two unrelated things together. Giving yourself an arbitrary rule. Taking the problem apart before trying to solve it. Answering as somebody else. Letting something random decide. Walking away from it.
Then the same method run through a model, with what changes, what breaks, and where the research says it works or does not.
The human half comes first and it stands on its own. If you have no interest in AI at all, those techniques still work, because they worked on people long before there was anything to prompt. And if you do use AI every day, the human half is the part that teaches you what to actually ask for.
Read in order. The first post sets the frame the rest of them pay off.
