NP THE EVIDENCE LOCK
— SIGNALS LOGGED

PUBLIC EXPERIMENT // 01

A claim is a lock.
Evidence is the key.

Two research passages sit side by side. Your job is not to solve P versus NP. It is smaller, stranger, and currently more useful: decide whether the support actually fits the claim.

6 rounds · about 8 minutes · no specialist credentials required

THE PREMISE

Most verification asks a few experts to be right. This experiment asks many imperfect readers to be independently specific.

BEFORE THE LOCK // ANONYMOUS

Tell us what kind of eyes are reading.

No name. No email. Broad background lets us see where interpretations split.

Boolean logic familiarity
SAT / CNF familiarity

Stored: these four answers, six puzzle judgments, duration, and score. Not stored in the study database: name, email, or IP address.

LOCK 01 / 06

Read the two fragments.

ALL SIX LOCKS ATTEMPTED

Seal the field report.

Your score has two halves: agreement with a frozen machine diagnostic and consistency between your own answers. Agreement is not truth. Disagreement is not failure.

—/ 100

FIELD REPORT ACCEPTED

You are a —.

Your signal is now part of the experiment.

Diagnostic alignment—
Rule coherence—
Calibration probes—

Diagnostic alignment compares you with a frozen machine mapping that is explicitly not human gold. The experiment is interested in where people disagree with it—and with each other.

RECEIPT—
METHOD // READ BEFORE TRUSTING

This game measures signals, not truth by applause.

Each run contains four quotations from two papers on Boolean functional dependency and two hidden calibration probes. Cases are randomized. Compiler mappings and probe identities never enter the browser payload.

The score

50% diagnostic alignment: agreement with a mapping frozen before public play. It is a machine prior, not an answer key.

50% rule coherence: whether your final verdict follows the evidence kinds, scope, and assumptions you selected.

The boundary

Public stats disclose submission and judgment counts only. Raw answer distributions remain sealed during collection to avoid social influence. The database stores no names, emails, free-form demographics, or IP addresses.

Source quotations: ICCAD 2007 and IEEE TC 2010.
THE TWELVE KEYS

Choose the direct fit. If every key is merely adjacent, choose “No key fits.”