How I partnered with AI to write this article.

1) When I first read about the GPDval study in Ethan Mollick’s work, I sensed something about the implications for knowledge work. But I couldn’t fully wrap my head around it. I used Claude Opus to help me clarify what I was trying to say, acting as a thought partner and asking me sharp questions to refine my argument.

2) After I outlined and wrote the article, I was still lacking tangible examples. I was stuck on the wine-maker sommelier metaphor I used in some of my earlier work. So I used ChatGPT, Gemini, and Claude’s deep research feature. I gave them my outline along with key literature on desirable difficulties and perceptual learning and asked for a senior-researcher “horizon scan” to surface analogies from relevant fields. Then I used this research to nail down examples that would work in this setting (e.g., radiology).

3) Finally, after I had the full article with examples, I uploaded my entire article into Gemini + my NotebookLM (“The Science of Learning”) and asked it to fact-check any weak arguments where science might contradict my reasoning and what I needed to look into further. It put me back into a loophole of really understanding perceptual learning, as I had made some wrong inferences.

4) After I updated the article to better reflect science, I let ChatGPT act as my developmental editor and help me improve word choice, pacing, etc. I have an entire subset of prompts for this editing stage.

5) Lastly, I used a combination of Grok, ChatGPT, Claude, and Gemini for headline and subheadline brainstorming, again, an entire subset of prompts for this.

How do you partner with AI in your field of work?

Thanks to Mark Myers whose question prompted me to write about this how I work with AI partnership.