NDE Accounts
Self-reports are treated as medical narrative data, not as automatic proof of any mechanism.
Structured Intake
The questionnaire captures event context, observations, timing claims, and phenomenological detail.
Locked AI Scoring
Advanced reasoning models apply the same multi-stage prompt to every selected case.
Case Profiles
Outputs preserve uncertainty through claim-level fields and a case-level headline signal score.
Cross-Model Method
Agreement and drift can be measured across model families, prompt versions, and future runs.
Example output language
From "dossiers" to structured case profiles.
"Case profile" keeps the clinical and evidentiary tone without implying a legal file or a formal physician-authored case report. The site can use this public term while the underlying schema remains precise and auditable.
Case #7572
Cardiac arrest
"I watched them apply the paddles. I counted three shocks."
CPR observed
Defibrillation
Personnel identified
Signal Score
78 / 100
Standards-aware, mechanism-neutral
The claim is methodological before it is statistical.
The scoring instrument does not infer survival, nonlocal perception, anti-materialism, or metaphysical truth. It asks a narrower medical-evidentiary question: how strongly does a report suggest organized conscious experience during a medically bounded interval in which ordinary waking consciousness would be unexpected?
Separate the claims
Event verification, content verification, timing certainty, and medical plausibility remain distinct.
Red-team the ordinary
Prior knowledge, inference, post-event discussion, media tropes, and memory reconstruction are scored explicitly.
Compare model runs
A locked prompt creates an auditable basis for convergence, disagreement, and calibration over time.