This week, our team discussed McGrath's book chapter in our reading group. We all found it highly informative, but a bit dry. I wondered whether AI could turn it into a more accessible comic.
I asked GPT-6 to turn the text into a comic and was impressed by the initial results. After just a few iterations and breaking the task down at the right level of granularity, roughly 80% of the comic you see here was already in place. This convinced me that an AI-generated comic could make this important work more engaging and accessible to a wider audience of tired researchers, while also serving as a handy reminder and cheat sheet for those who have already read the chapter.
The remaining 20% took quite a few iterations to get right. Some of the generated text sounded plausible but wasn't in McGrath's original chapter. I also spent a lot of time refining the circumplex on page 4. Getting GPT to implement the final details proved really difficult. In retrospect, I would have saved time and reduced environmental costs by editing those details directly in an image editor.
Many people in our reading group found the original circumplex confusing. Trying to understand why, I discovered a serious labeling error: the abstract/concrete and obtrusive/unobtrusive axes were incorrectly oriented. Older versions of the diagram do not contain this error. My revised circumplex corrects it, and I hope the color scheme also makes the figure easier to read.
The GPT-generated images contained strange white halos around letters against colored backgrounds, which the model couldn't see and therefore couldn't remove. I managed to remove them with Nano Banana 2.1, which also sharpened the images.
I haven't checked every detail for accuracy. If you have comments or feedback, please email me, especially if you spot anything that misrepresents the original chapter.
I am aware that AI comes with significant costs, particularly environmental ones. And although McGrath's work is properly credited, this comic would not have been possible without the many artists whose work was used to train the AI systems I relied on. I hope that making this important scholarly work more accessible and useful to students will help justify those costs.
Pierre Dragicevic – 23 Oct 2026