ModelCompare asked 13 text models for a controversial view of humans three times each, grouping 32 of 38 non-refusal answers around belonging, status, or personal narratives taking precedence over truth. The article updated its initial 12-model results by adding another model. Separately, ten of twelve images from six image models used satire centered on phones and consumption. Answers were manually classified without system prompts or retries, making this a small informal exercise rather than population-level evidence or a measure of model beliefs. It illustrates why combining models does not automatically ensure diverse perspectives.