Drowning Test is a longitudinal protocol by Lee Sharks for measuring the compositional survival of an entity or work in generative search.
The test begins with a fixed Holographic Kernel whose claims are itemized and deposited. At each measurement cycle, a panel of queries is run across selected platforms. Each answer is checked for the kernel claims, producing a gamma score between zero and one.
A time series can show gradual loss, recovery, stability, or abrupt changes in platform treatment. A shift of at least 0.3 between adjacent cycles is provisionally called a silent state change.
Gamma is not a quality score. An answer may preserve the kernel and still be misleading, or omit the kernel while otherwise answering well. The specification recommends cross-platform panels and explicit classification rules to reduce observer discretion.