UAP Codex · open research

79 years without an official explanation.

There's something institutions couldn't — or wouldn't — explain since 1947. A compendium of the 331 best-documented institutional cases — the ones that survived military, congressional, and journalistic filters. Not a sightings list. The evidence that doesn't explain away easily.

Follow

The scope

It's not theory — it's a documented evidence collection.

331

Institutional cases

Survived military, congressional, and journalistic filters

79

Years of phenomenon

Longer than the entire Cold War (1947–1991)

62

Countries with records

Every continent except Antarctica — not a US-only phenomenon

The corpus answer — in one sentence and in numbers

UAP are not one single thing. They are several distinct things mixed under the same label — and that's exactly what makes them hard to explain.

Most point to classified military programs — we've known since the U-2 (1950s) and the F-117 (1980s). A non-trivial portion involves non-human entities we don't yet know how to categorize. Some may be rare natural phenomena (plasma, sprites), though multi-sensor military evidence rules that out for most Tier S cases. Misidentifications get filtered out first — what remains are the 331 cases that survived that culling.

How the corpus's 331 cases classify among the narratives — they sum to 100%

Misidentificación: 52.0%Fenómeno natural: 13.3%Indeterminado: 11.2%No-humano: 8.5%Posible fraude: 7.9%Tecnología humana clasificada: 4.2%Tecnología adversaria: 3.0%52.0%13.3%11.2%8.5%7.9%
331cases
  1. Misidentification52.0% · 172 cases
  2. Natural phenomenon13.3% · 44 cases
  3. Indeterminate11.2% · 37 cases
  4. Non-human8.5% · 28 cases
  5. Possible hoax7.9% · 26 cases
  6. Classified human technology4.2% · 14 cases
  7. Adversary technology3.0% · 10 cases

Forced, navigable classification — each case counts once in its most-likely narrative (argmax); mundane opened into misid/natural/possible hoax, non-human consolidated. Incidents are classified by the object and document cases by their content's lean; 'Indeterminate' gathers what cannot be decided. The expected-value split —comparable, spreading the uncertainty— is on /calidad.

Temporal distribution

Reports don't distribute randomly.
Four peaks mark institutional ruptures.

1947 Roswell — USAF creates Project Sign · 1973 Pascagoula — Senate hearings · 2004 Nimitz — first official ATFLIR video · 2026 PURSUE — first presidential disclosure. Each peak is a moment when an institution could no longer keep denying.

Corpus por año · 19472026

331 casos · 79 años

47

1950s

42

1960s

50

1970s

30

1980s

31

1990s

27

2000s

24

2010s

44

2020s

Walk the 331 cases chronologically →