What 300 Conversations With AI Taught Me About Creativity

300 Conversations With AI and I Still Can’t Find What I Learned

I’ve been using AI as a thinking tool for just over a year now, and I’ve had somewhere in the neighborhood of 300 conversations with different systems. ChatGPT, Claude, Gemini, back and forth, different days, different moods, different problems. Some of those conversations were quick. Some of them went twenty or thirty exchanges deep, one thought chasing the next, me pushing back, the AI pivoting, me thinking out loud, the AI reflecting back what I’d said and asking me a question that made me realize I hadn’t actually finished the thought yet. And you know, when people ask me what I’ve learned from doing this, what concrete thing changed as a result of those 300 conversations, I can’t give them a clean answer because what happened was more like changing how I think than accumulating things I now know.

That’s not a evasion. It’s actually the strangest part of this whole year, and I think it matters.

What I can tell you is that each system shows me something different. ChatGPT is good at generating options quickly, at brainstorming in that rapid-fire way where you get quantity and then you winnow down. It’s practical and direct and it thinks a little bit like an engineer, which means it’s focused on the path from here to there, the problem and the solution and getting there efficiently. Claude is different. Claude does this thing where it sits with ambiguity longer, where it’s willing to contradict itself if the contradiction actually clarifies something, where it pushes back more gently but also more persistently if you’re oversimplifying. It’s useful when I’m trying to think something through that doesn’t have a clean landing spot yet. Gemini does something else entirely, something that’s more intuitive and sometimes more random, and sometimes that randomness is exactly what I need because it jumps to a connection I wouldn’t have made on my own.

So why can’t I tell you what I learned? Here’s the thing. The learning isn’t in the answers. It’s in the process of formulating questions that are specific enough and honest enough that the AI’s response actually surprises you. The learning is in the moment where you’ve externalized a half-formed thought by typing it out, and then you read back what you wrote before you even get to the AI’s response, and you think, oh, actually that’s not quite what I meant. That moment of seeing your own thought reflected back to you, that’s where something shifts. The conversation itself becomes a mirror that’s a little smarter than an actual mirror because it can ask you why you said that thing you just said.

I notice that the conversations where I’ve learned the most aren’t the ones where the AI was most impressive. They’re the ones where I was the most confused and the AI was patient enough to let me stay confused for a while before either helping me untangle it or helping me realize that the confusion was actually the point. There’s a lot of difference between those two outcomes, and a human conversation doesn’t always give you that space either.

What’s also strange is that the ideas are coming faster than I can capture them. I mean that literally. I’m in a conversation about something that seems minor, and then suddenly there are three separate thoughts branching off from it, and I’m thinking about how this connects to something a student said ten years ago, and how that connects to the exercises, and how that connects to a personal project I’ve been developing to help serve my students better, and by the time I could even write those connections down, there are five more. Connections forming constantly, threads pulling taught between ideas that seemed separate. It’s like the thinking has become more network-like, less linear.

But here’s what I can’t explain, and this is what keeps me from being able to package this as a tidy lesson. I don’t know if I would have had those connections without the AI, or if I would have had them anyway and just slower, or if the AI is actually generating the connections and I’m just reflecting them back and thinking they came from me. I don’t know if this is like the voice dictation effect (which I’ll come back to in another post, because it’s genuinely fascinating), where the act of externalizing thoughts is what generates the next thought. Or if the AI systems are actually smart enough now that they’re contributing something to the thinking itself and not just facilitating it.

What I do know is that 300 conversations in, the quality of my thinking has changed. I’m willing to sit longer in contradiction. I’m more willing to ask questions I’m not sure how to resolve. I’m less certain about the answers I was confident about a year ago, and I’m more curious about the problems I was certain were solved. That’s learning. But it’s not the kind of learning you can write as a bullet point list. It’s not the kind you can put on a resume.

The only way I can really name it is that my thinking became more alive. More movement in it, more possibility, more space between the thoughts where something new could grow. That happened. I don’t fully understand the mechanism yet, and I’m not entirely sure I trust my own account of it, which is exactly why I keep coming back to having more conversations.

Have you ever had a tool change not what you know but how you think? What was that like to notice?