Proposed readings
Every few years, someone reads it.
Nine announced readings of the script, from 1921 to 2019, and the three positions that organise the field. For each: what was claimed, what a reading of that kind would have to do, what this project measured on that family, and the recorded decision. The measurements open their source file and hash.
01 / The bar
What counts as a reading.
The project fixed this rule before searching for any key, and every claim below is held to it. A short list of plausible words is not a reading. A key that works on the words it was fitted to is not a reading. The test is what happens on pages the key never saw.
02 / Nine announced readings
The claims, and what the record shows.
In order of announcement. Where the project did not measure the family directly, the card says so and gives the nearest measured ground.
03 / Positions, not readings
Three camps the book engages.
04 / Behind the names
Thirteen cipher families, each tried.
Most announced readings belong to one of these families. Each was run against held-out text and matched controls in the June batteries. The wording in the last column is the recorded note; the label restates it within the book's current limits.
| family | recorded outcome | recorded note | key number |
|---|
05 / What remains open
The door the measurements cannot close.
The book's current position is stated plainly: the restricted search produced no passing candidate, and that supplies no general exclusion of encipherment. These are the named residues.
06 / Try it
Run the same tests yourself.
Pick a period language, let the browser hunt for a table, and watch the same hunt on a shuffled copy.
↗ Computed in your browserTest a cribAlign a fragment with a proposed reading and apply the key elsewhere.
↗ Recorded experimentThe current searchAll 45 manuscript runs, their controls, stability and outputs.
↗