TrajectoryLock

THIS IS: research prototype / auditable geometric test. Compatibility vs declared official line. Independence groups so copies don't inflate certainty. CLI + local workbench + JSON API. THIS IS NOT: a certified forensic instrument; substitute for scene reconstruction, medical findings, lab exam; shooter/intent/guilt/narrative identifier; automatic detection of invisible projectiles; face recognition. Match probability is P(match | declared model), not P(official account is true). Synthetic example results must never be represented as real-case findings. No private case facts. Author: Aziel Eliab.

How close is this line to the claimed line? Load JSON. Run check. See the result. Three separate numbers — they are not the same thing. They do not name a shooter, intent, or guilt.

Load a JSON case (or tap Load example / Import).
Tap Run check.
Read the three numbers. Then Export if you want a receipt.

—compatibility
how close is this line to the claimed line

—match chance
P(match | declared model) inside frozen tolerances

—how strong is the evidence
independent groups, not copies

Frozen tolerances: load JSON, then Run check. Do not retune after seeing the result without a new hypothesis revision.

Case JSON

Result

Load JSON, then Run check. Synthetic example results are never real-case findings.

Guardrail always shown. This is not a certified forensic instrument. It does not identify a shooter, intent, or guilt.