الميزان · The Balance
Mizan
A public, preregistered trial of the Qur'an's own testable claims. Candidate tensions are pressed adversarially, recorded under fixed rules, and exposed for human review.
The instrument is designed to report failure, uncertainty, and bias as plainly as survival. Every result links back to its verses, reasoning, provenance, and governing procedure.
Coherence queue4,949 / 20,448Live ledger connecting...
Current scientific standing
Instrument-bias flag active
The negative rate is below the preregistered floor. This run is not reported as a Qur'anic result.
Live measurement
What the instrument currently says
Queue progress
20,448 pairsVerdict distribution
absolute countsCalibration floor
negative verdictsCurrent 0.83%; required 5.00%.
Human audit coverage
reviewed versus provisionalHighest-priority review
Findings that need human scrutiny
A model verdict is a lead for review, not a settled contradiction. These records are shown so scope, speaker, translation, and context errors can be found rather than hidden.
Loading priority records...
How a pair reaches the ledger
A visible procedure, not a black box
- CandidateA seed citation or shared corpus feature proposes a verse pair.
- TriageA tension score prioritizes work but cannot determine the verdict.
- AdjudicationProsecutor, defender, and neutral roles operate under sealed prompts.
- LedgerThe decision summary and provenance bind to the preregistration hash.
- Human auditCorrections append to history instead of erasing prior output.
Before the final verdict
Stand as a recorded witness
Your declared starting point is committed before the trial completes, without putting your identity on chain.