Earlier this month we announced the coming launch of GDELT 5.0. In creating the background image for the launch announcement, we iterated through numerous designs and ideas using generative image creation models. Here are a few iterations we went through on our way to the final image. This sequence demonstrates the generative ideation process in which we crafted an initial prompt, tried different overall prompts, then as we began to narrow in on designs we liked, iteratively refined the prompt based on each image it produced, ranging from visual elements to color schemes to the level of abstractness. At the same time, we encountered the very real limits of modern image generators with respect to their more limited visual range, consistency of visual symbolism and difficulty being truly "creative" versus merely mimicking what they've seen before.