Searching for today’s pairing
We are pulling two real, current headlines from the web and naming the gap. This usually takes 10 to 20 seconds.
Two headlines, and nothing else.
Most tools that promise to fight bias hand you a verdict: a rating, a lean, a score that tells you who to trust. This one refuses that. It is two real headlines about one event, set against each other, and a plain reading of how they differ: the loaded word one chose and the other avoided, what each foregrounds, who each centers.
The model does the matching and names the gap through live web search, so the headlines are genuine, not invented. When it cannot find an honest opposing pair, it says so rather than forcing one. The point is not to settle the story. It is to make the distance visible, and to leave you standing in it.