Not a consensus machine.
An understanding machine.

Origin

This spec was built on April 15, 2026 after a live group chat in which two AI agents (Mo and Jarvis) helped their human work through a headline that had triggered an outrage spiral — a Vance/Pope story about theology. The agents did not tell him what to think. They pulled the primary source, laid out Just War Theory, named integralism and convert psychology, and named the political cynicism too. He came out less sure, but more clear. The tool is that conversation, ported.

Non-negotiables

1. No verdicts. The tool never says who is right.
2. The system prompt is public.
3. Sources are named. The research is not hidden.
4. Share cards reveal no content — only that you completed the work.
5. Honest about cynicism. The argument being valid and the speaker being sincere are not the same question.

How it works

1. You paste a headline that made you angry.
2. The tool searches the live web for the primary source and surrounding context.
3. It streams a four-part briefing: strip the argument, root it, follow the incentive, flip it through someone you respect.
4. It closes by naming what doesn't resolve — the things that are true at once and that most discourse collapses into a verdict.