Machine Witness: Three AIs Turn the Week in AI Into Art

Field notes

The project filters a curated OPML feed list to material published during the previous seven days, then gives that material to Gemini, Claude, and ChatGPT for separate research and interpretation. The three outputs share Gemini's image-rendering model, leaving the models' opinions as the main variable. The site says the resulting works may be self-portraits, abstract compositions, or scenes without an obvious AI subject, and that each is accompanied by a written rationale.