How judging works

Blind rank-order

You are shown a complete sealed session (sketches, descriptors, uploads) and a set of candidate images. One of them is the session's actual target. You are never told which one.

Your task is to rank the candidates from best to worst correspondence with the session. That is the whole task. Rank on what the transcript actually says, not on what it might have meant. A ranking is a measurement, not a compliment.

Why you never judge your own session

• You know your own intent, so you cannot read your own transcript blind.
• Scoring your own work retrains the assumption habit the practice tries to remove.
• An independent judge sees your output as data and, over time, reads your strengths and weaknesses better than you can.

In Viewer 001 this is a hard invariant. The system never assigns you your own session.

Tokens

A session costs one token, paid from your balance. Your onboarding funds your first three sessions; after that, one fresh judgment opens each session. Tokens are earned by judging: one delivered judgment earns one token and opens one session. Judgments must be genuine work; a judgment completed too quickly to be an actual reading of the session earns nothing and opens nothing.

Before starting a new session, you judge one from the queue. A judgment is spent by the session it opens; revoking a session returns its token but not the judgment. If the queue is empty, the requirement is waived and one token is granted, at most once per day.

Tokens cannot be bought, at any price. The only way in is judging, because every sealed session needs judges. You give reading to receive reading.