Ask one AI what happened in the news today and you get one answer, delivered with great confidence. Ask ten, and you find out how much of that confidence belongs to the model.

That is the idea behind Poly Trends, our own project. Every day it takes a news story and gives it to ten AIs. Each one writes what happened, where it sees a dispute, and what a reader could do about it. Then the answers are scored out of ten, and the page shows an average for the American and the Chinese models. On the story of India’s women’s cricket team winning its first match under a new captain, the American AIs averaged 7.5 and the Chinese AIs 7.2. The score rates how good that day’s answers are, not who will guess the future.

Where they disagree. The interesting part is not the average, it is the gap between answers: one model stresses the debut, another the opponent’s weak batting, a third the change of captain. A reader sees the disagreement instead of an invented consensus (see all the answers on that trend).

What it is useful for.

• Checking how stable a story is: if ten AIs split, treat any single AI summary with care.
• Comparing regions: the page shows how American and Chinese models rate the same answers.
• Following a topic: the AI tag collects stories about artificial intelligence itself.

New trends appear through the day, in 11 languages.