When I started using ChatGPT as a brainstorming partner, I noticed that the kind of help I wanted depended on what we were discussing. Sometimes I needed a careful, analytical response, someone willing to follow a problem through its assumptions and awkward exceptions rather than settling too quickly on an attractive answer. At other times I wanted the opposite: a practical, down-to-earth conversation that went straight to the point. A simple question could disappear beneath layers of explanation, and I wanted a way to bring it back into view without losing the value of the deeper thinking.
So I decided to experiment with giving the AI different conversational personalities. One would be more analytical, inclined to test an idea and notice what might have been overlooked. I called that version D1. The other would be more direct and plainspoken, someone who might listen to a complicated proposal and ask why we were making it so complicated in the first place. That one became D2. The names gave us a convenient way to refer to the contrasting approaches, and a little room to have fun with them. They weren’t two separate people, or two minds hiding inside a computer. They were different ways of guiding the same AI tool to respond.
Looking back, I can see why the two voices became more than an amusing device. One represented the pleasure of following every thread of a difficult question. The other represented the relief of remembering what the question was for. I needed both, even though they were only playful ways of describing approaches to problem-solving.
The important part was that neither voice had to win permanently. A simple answer could be a blessing on one day and a dangerous shortcut on another. Detailed reasoning could save a project from a hidden mistake or bury a small request beneath an elaborate plan. The work became more enjoyable when I could move between the two without treating either as a character flaw. The joke gave me a language for that movement.
The little D1-and-D2 experiment began partly because a single style of response could never do every job equally well. Sometimes I wanted someone to test the logic and point out the unexamined assumption. Sometimes I wanted the shortest useful answer, without a whole parade of qualifications. Giving those styles names let me ask for what I needed without composing a small essay about tone every time.
The joke was important too. It made the whole arrangement less solemn. I could imagine the two voices disagreeing over how many paragraphs a problem deserved. Changing conversational style didn’t change who was responsible for deciding what mattered. It simply gave me another way to explore an idea.
Before long, the names became characters in our running conversations. I could imagine D1 carefully explaining every possible complication while D2 looked at the whole arrangement and wondered whether we couldn’t just put the button somewhere sensible. The humour mattered because it let me explore the differences without having to declare one style superior. Each could be useful, each could be exasperating, and neither needed to become the official voice of reason.
That is why D1 and D2 appear in the pages ahead. I created them as an experiment in getting more useful conversations, but their contrasting approaches began giving me something else to think about: how I reason through a problem, why I sometimes make it harder than it needs to be, and what happens when two very different kinds of thinking are allowed to help each other.
