the record

A claim is easy to make.
Proving what stood behind it later is the hard part.

When something you published is questioned — weeks or months on — the work is reconstruction: what exactly was said, what evidence supported it, who signed off, where it went, and what changed afterwards. RexUnda builds that record while the work happens, instead of assembling it from folders and memory once someone asks.

the artifact

One claim, and everything that stood behind it.

A claim record is a single object: the exact words, the moment they were fixed, the evidence attached, the conditions that were missing, and every later revision linked rather than overwritten.

01 · stated
The exact words
Captured verbatim and fixed to the moment they were made, on the database's own clock rather than a caller's.
02 · supported
What backed it
The material relied on, attached to the claim rather than filed near it.
03 · missing
What it still needs
Unmet conditions named before publication, not discovered after. A claim that cannot be assessed says so.
04 · after
What changed
Revisions and withdrawals link to the original. The first version is never rewritten.

A record being complete is not the same as a claim being right. RexUnda keeps the first and never asserts the second.

why trust the keeper

A miss weighs exactly what a hit weighs.

RexUnda runs its own measurement book in public — signals scored, then graded against what the market actually did. The grading is the point, and the misses are published on the same terms as the hits.

Hit
Counted once, in the window it was called for.
Miss
Counted once, on the same scale, and it stays on the record.
the standing number

Where the book stands today.

32.6%
n=1682 resolved · 24h window · 95% interval 30.4–34.9%
measured against 26.4% — the rate this book's own rows produce by chance, derived every time it is shown

A hit here needs two things at once: the direction right, and a move past a calibrated floor. Both together is a far harder test than either alone, which is why the comparison shown is the rate this book's own rows produce by chance — derived from them, not assumed.

the sealed test · no. 2

The rules were fixed before the data existed.

One question, written down and locked in advance — the measure, the window and the threshold all declared before a single row of it accrued. Accrual begins on the seal date below; the window runs long after that, and nothing about the outcome is readable until it closes.

accrual begins 2026-09-01
earliest read 2027-03-01 — not before