Tuesday, August 18, 2026

an experiment in satellite image editing/comparing with Claude chat

We have enterprise access to the Claude AI model world, and I'm experimenting with both the Claude code features and Chat/Co-work options. I thought something interesting would be a comparison of the city light pollution before and then a few days after the large derecho/tornado event that still has 100,000 people without power a week after the storm.



spots that were less bright on the 13th compared to the previous image. Ignore the broad fainter red marks -- this is because the previous image had plenty of moonlight. I had thought about posting the entire conversation, as interacting with LLMs have some interesting failure modes, but not for today.