HOW IT WORKS

VERDICT is a weekly practice. You write a confession. You predict what strangers will call you. They judge. You see the gap between your prediction and theirs. Over time, the mirror accumulates into something only you can see.

The loop

Write a confession. 300 characters. No names, no links.
Predict your label. What will strangers call you?
Lock. No edits. No deletes. No defense.
Strangers judge. One label per confession.
Verdict reveals. The distribution of what they called you.
Your mirror. The gap between your prediction and theirs — accumulating, week by week.

What this browser stores

This browser

Your confessions, your judgments, your verdicts, and a random identity token — all in IndexedDB on this device. Cleared when you clear your browser data. Never sent to the server except as content or counts.

The server

Anonymous confession text (decays after 90 days), verdict counts per label, and rate-limiting hashes (purged after 7 days of inactivity). No accounts. No emails. No names. No IP logging. No cookies.

What VERDICT does not collect

No email, phone, or name — ever
No cookies or tracking pixels
No device fingerprinting
No IP address logging
No analytics or behavioral tracking
No cross-site identifiers
No social graph

Five compartments

The server stores data in five compartments that are never joined. Each compartment is sealed at the schema level — no query joins identity-adjacent data with content data.

Contains	Never joined with
Confession text, verdict counts	Rate token hashes
Verdict distributions	Any device identifier
Rate-limiting token hashes	Confession content
Vote dedup fingerprints	Labels, text, identity
Moderation queue entries	Rate tokens, device info

Even a complete database dump cannot reconstruct who confessed what or who voted how.

Data lifecycle

7 days Inactive rate-limiting tokens are purged.
14 days Confessions close to new votes.
90 days Confession text is deleted. The verdict distribution remains — counts, not words.

The pattern outlives the words.