[jgalt212]

If a standard prompt gives you the wrong or spurious answer, the model is not a good model. In 2026, we should not be having to engineer a summarize prompt.