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