How we reviewed the files

Simple labels, source-backed caveats.

This site is built from the repo's normalized release inventory, direct media package index, DVIDS metadata, measurements, dossiers, and written findings. The public experience is intentionally light, but the labels come from the evidence boundary in the repo.

Sufficiency score

  • 0: catalog or context record only.
  • 1: single still, sketch, or summary with weak time/location context.
  • 2: direct media or paired document exists, but key telemetry is missing.
  • 3: enough date/location/media context for partial external correlation.
  • 4: full-motion evidence plus timestamp, location, platform, and sensor context.

LLM boundary

LLMs can help draft plain-language summaries and organize metadata. They do not decide whether a clip shows anomalous performance. Conclusions are based on source records, deterministic measurements, visual review, and explicit missing-data limits.

What stronger claims would require

Exact timestamp Exact location Platform motion Range Sensor model Field of view Unredacted frame sequence External correlation

Repo artifacts