Everywhere else, ranking optimises engagement, and engagement is maximised by conflict. Purple Party ranks by how far agreement travels across the divide, adjudicates factual claims with a panel rather than a crowd, and keeps the record of who changed their mind.
Almost nobody changes their mind in public any more, because the cost of doing it is humiliation and the reward is nothing. Purple Party exists to invert that price.
The mechanism is not persuasion and it is not moderation. It is bookkeeping: every claim carries whether it has been challenged, every essay carries the objections raised against it, every member carries the record of what they got wrong and fixed. Where other networks make certainty cheap and revision expensive, this one is built to do the opposite — and the whole product is downstream of that one trade.
Build the place where people who can change their minds can find each other, and give them the instruments to argue well.
A network where the highest-status members are the ones with the longest correction records, and where that is unremarkable.
If this works, a member reads an essay and checks the objections before the thesis. A claim with no challenges against it looks thin rather than clean. Someone announces they were wrong for nine years and the response is interest rather than a screenshot. The test is not whether people agree more — it is whether they can say precisely where they disagree, and what evidence would move them.
The failure mode is a network that merely feels civil. Civility is easy to manufacture and worth very little. What is being attempted here is harder and narrower: making it cheap to be wrong out loud.
Four rules that decide the arguments the rest of the product has.
Each one is live. Click into anything.
Ranked by how far agreement spreads rather than how much of it there is. Every post that makes a factual claim carries its adjudication status in the open — held up, contested, corrected by the author, or never examined at all.
Open the feedA plain chronological record of members changing their positions, with the reason attached to each one. Nothing else is on the page — no reactions, no reply counts, no ordering by anything but time. The unremarkableness is the point.
Open the updatesLong-form arguments with discussion attached to individual sections, and a standing that reflects what survived challenge. Wikipedia converges on one neutral text; this keeps the disagreement, attributed and navigable.
Open the libraryAnswer first, then the crosstabs unlock. Calibration polls ask questions with a true numeric answer, and reliably find that everyone is wrong — across every group. That is the strongest bridging mechanic in the research.
Open the pollsAssessment results as bands rather than points, because a point would be more precise than the evidence. Each dimension is published one at a time, by choice, alongside things the member actually did — corrections made, Turing tests passed, steelman ratings from people who disagree with them.
Open a profileEverything here runs in the browser. There is no server, no account, and nothing is sent anywhere — actions that would hit a backend say so instead of pretending. What is real is the shape: the ranking language, the adjudication ledger, the response gate, the disclosure controls, and the copy that has to carry all of it.